CN107340298A - Balance car system monitoring method based on camera pavement detection - Google Patents
Balance car system monitoring method based on camera pavement detection Download PDFInfo
- Publication number
- CN107340298A CN107340298A CN201710515182.XA CN201710515182A CN107340298A CN 107340298 A CN107340298 A CN 107340298A CN 201710515182 A CN201710515182 A CN 201710515182A CN 107340298 A CN107340298 A CN 107340298A
- Authority
- CN
- China
- Prior art keywords
- mrow
- msub
- mtd
- mtr
- value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8887—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges based on image processing techniques
Abstract
The invention discloses a kind of balance car system monitoring method based on camera pavement detection, the method that the LBP textural characteristics statistical information collected in pavement image is carried out to gradient difference using camera obtains pavement texture distribution, coordinate the front and rear contrast difference of filtering of inertial sensor, the two weighted sum is asked for the estimation of a normalized pavement disease degree, it is eventually used for improving the inertial sensor calculating system of balance car, real-time adaptive changes Kalman filter and realizes that pre- measuring angle is more accurate, motor control strategy is selected according to cam feedback value simultaneously.Using technical solution of the present invention, it can estimate that measuring and calculating draws road bump degree of disease, convergence and the precision of Kalman Prediction are optimized after introducing vision, draw more stable motor control strategy, simultaneously, the visual information that vision sensor collects, which can also preserve, to be met manual analysis or carries out early warning processing needs, is strengthened man-machine interaction, is improved overall security.
Description
Technical field
The present invention relates to balance car measuring and calculating and control technology field, more particularly to a kind of putting down based on camera pavement detection
Weigh car system monitoring method.
Background technology
As balance car market develops rapidly, the market product of inferior quality also pours out, and security incident constantly takes place frequently, and improves flat
The vehicle control robustness that weighs and usage range are most important.One of key technology during vehicle body attitude measuring and calculating controls as balance,
Decide whether vehicle can reach accurate balance.
Presently, the manned balance car or the manned balance car of research unit's exploitation that in the market is sold all are in tradition
The modification adjustment of minor details is carried out on the basis of inertial sensor, single-sensor can not be broken through all the time in measuring and calculating control signal
On limitation, before not only result in the stagnant step of manned balance car technical research not, also left in safety problem a bulk of
Clear area.
The content of the invention
For the deficiencies in the prior art, the present invention solves the problems, such as it is to improve balance car control accuracy, it is ensured that whole
Body security.
In order to solve the above technical problems, the technical solution adopted by the present invention is a kind of balance based on camera pavement detection
Car system monitoring method, the LBP textural characteristics statistical information collected in pavement image is subjected to gradient difference using camera
Method obtain pavement texture distribution, coordinate the front and rear contrast difference of filtering of inertial sensor, the two weighted sum asked for
The estimation of one normalized pavement disease degree, it is eventually used for improving the inertial sensor calculating system of balance car, in real time
Adaptively changing Kalman filter realizes that pre- measuring angle is more accurate, while selects motor control according to cam feedback value
Strategy, including process in detail below:
Process 1 is acquired data using accelerometer and gyroscope, realizes that gathered data is believed using Kalman filter
Number fusion, denoising draws real-time inclination angle Pitch information of the car body in the i.e. front and rear sides direction of Y direction, the receipts after being predicted
The angle information held back, finally with the integration angle progress difference of gyroscope, comprise the following steps that:
(1) system for introducing a discrete control process, with a linear machine differential equation (Linear Stochastic
Difference equation) describe:
X (k | k-1)=AX (k-1 | k-1)+BU (k) (1)
According to this formula, gyroscope current time measured value is substituted into, is worked as afterwards plus the drift value of gyroscope
The predicted value Angle of preceding angle, being write as matrix form is exactly:
(2) prediction of covariance matrix, need to define two input values in calculating process, be respectively:Gyro sensors
The drift noise of device and the angle noise of acceierometer sensor,
P (k | k-1)=AP (k-1 | k-1) A'+Q (3)
Q wherein in formula is vectorCovariance matrix, i.e.,:
Because drift noise also have angle noise be it is separate, then;
Cov (Q_bias, Angel)=0 (5)
Cov (Angel, Q_bias)=0 (6)
Definition procedure angle noise covariance parameter and measurement noise covariance parameter in program are respectively:
Q_angle=0.001;Q_gyro=0.003;R_angle=0.5;
(3) gain coefficient of Kalman filter is calculated, this is that a bivector gain coefficient is set toFor diagonal
The amendment of degree and angular speed, expression formula are:
Kg (k)=P (k | k-1) H '/(HP (k | k-1) H '+R) (7)
The constant R occurred in formula refers to the noise coefficient of acceleration measurement, coefficient original definition in program
For:R_angle=0.5;
(4) kalman gain coefficient amendment predicted value, the angle drawn using two sensors of accelerometer and gyroscope
Difference regard error amount Angle_err, the amendment of value, formula are predicted using the product of kalman gain coefficient and error
It is as follows:
X (k | k)=X (k | k-1)+Kg (k) (Z (k)-HX (k | k-1)) (8)
Update prediction error value and angle value simultaneously:
Angel+=K_0*Angel_err (9)
Q_Bias+=K_1*Angel_err (10)
(5) this renewal is carried out to matrix covariance P matrixes, the covariance being mainly used in during next iteration it is pre-
Survey, formula is as follows:
P (k | k-1)=(I-Kg (k) H) P (k | k-1) (11)
(6) difference, and every 20 times are done into the angle measurement X at k moment (k | k-1) and k moment angle predictions
It is average that an angle mean error measurement is counted in angular surveying,
Process 2 draws the inclination data of real-time angular speed and X-axis of the car body barycenter on Z axis by gyro sensor,
The roll (i.e. the real-time inclination angle in left and right sides direction) of car body barycenter horizontal direction X-axis is drawn by Kalman's fused filtering, and
Data are normalized with map function and draws the mathematical statistics feature that pavement disease degree characterizes;Comprise the concrete steps that:
(1) calculate Z axis acceleration information using gyroscope and accelerometer calculates X-axis angle information and carries out Kalman Prediction
Fusion, obtains the posterior estimate for converging on an exact value θr,
(2) by calculate in X-axis to angle Roll go to ask for its sine value sin θr, and this sine value is multiplied
Vehicle wheel base length D between upper body left and right wheels, obtaining h values can be to be expressed as the height that car body barycenter is propped up by uneven road surface
Degree,
H=D*sin θr;
(3) the height value h divided by the height 2d of chassis raised the car body that this is calculated by uneven road surface is obtained
To the slope Q of whole car, this Q value is finally characterized as road surface for balance car tilting of car body degree, when value is 1,
Then reach the limiting safe state that vehicle body can bear maximum inclination, also reach normalization effect, formula is as follows:
Q=D*sin θ/2d (13);
Process 3 constantly obtains balance car car body road ahead image information by imaging sensor, and to image information
Gray processing, image segmentation, LBP characters extraction statistics and normalization conversion process are carried out, is comprised the following steps that:
(1) counted step by step such as using COMS cameras collection road surface picture data signal a (t), extraction LBP features
Under:
1) it is converted into gray-scale map for RGB color image;
2) floating-point operation Structural Transformation is optimized;
3) image digital signal is subjected to vertical divided in equal amounts, divided the image into as 1*10=10 sub-regions;It is so every
The pixel resolution of sub-regions is 640*48;
(2) local binarization;Choose threshold method and carry out binaryzation, realize the accurate extraction of target texture;Tool
Body extraction is as follows step by step:
1) during image traversal, 3*3 greylevel window is extracted every time as object, and selected window center gray scale
It is worth the threshold X as this binaryzationi,j, by 8 adjacent pixels respectively with this threshold XijIt is compared, if being more than this
Individual threshold value is just entered as 1, on the contrary then be 0;
2) it is numbered from the upper left corner, thus produces the binary number of one 8 and (be followed successively by [bi-1, j-1 bi-1, j
Bi-1, j+1 bi, j+1 bi+1, j+1 bi+1, j bi+1, j-1 bi, j-1]);This 8 binary number is just in this
The LBP values of heart window pixel point;
3) 8 binary eight kinds of rotations arrangement modes are all included, selects the LBP of minimum as center window
The LPB values of pixel characterize the texture information of this point;
(3) LBP histogram gradient distribution statisticses are counted, the normalization of pavement disease degree estimate, are comprised the following steps that;
1) LBP statistics with histogram is carried out to every sub-regions figure, 0 to 255 distribution histogram is divided into 16 sections
Counted, the pixel number for counting each section accounts for global proportionality, obtains a set array Rnm (1 for including 16 elements
≦n≦10;1≦m≦16);By this 16 element difference absolute value computings, caused one new sequence rnk (1≤n≤10;1
≤ k≤8), LBP histograms probability density distribution gradient is characterized with this;
rN,2*K=RN,2*K-RN,2*K-1
2) its array rnm summation Cn (1≤n≤10) is counted;
CNFixed interval is concentrated on, minimum value is just 0 (now pavement texture distribution is minimum), and the meaning that it is characterized is road
Face state the most flat, on the contrary this Cn value nearer it is to 1 (texture contrast is maximum), then shows that the degree of disease on road surface is tight
Weight;
The pavement disease degree quantized value Cn and P that are obtained in imaging sensor and inertial sensor are weighted by process 4
Summation, the two comprehensive confidence level final decision draw the introduced measurement variance of a measurement pavement disease degree, that is,
Sensor noise value, current car body X-axis inclination angle are θ, estimation noise variance I, and formula is as follows:
Process 5 carries out debugging method according to respective sensor accuracy and reliability.
First method:The weight K1 of feature is determined with experience debugging method, K2, the value of K3 three, three kinds are united
Meter feature is weighted summation, and three parameters of regulation are optimal until car body balance, and parameter regulation rule is as follows:
K1+K2=1;
K3<1
Second method:Carried out according to the corresponding optimal tuning parameter I of measured Cn, Q and P value under different road surfaces
Least square fitting carries out parameter tuning under matlab, fits the function of first order of suitable space-time, draws parameter K1,
K2 and K3, function are as follows:
Process 6 will be pre- as Kalman of lower a moment above for the introduced measurement noise estimate of variance I of pavement disease
The input value of the initial value of survey predicts the angle value of balance car car body of lower a moment, comprises the following steps that:
(1) Kalman filter implementation process in reference process 1, by the definition procedure noise covariance parameter in program
For:
Q_angle=0.001;Q_gyro=0.003;
(2) it is real-time that the introduced Noise Variance Estimation value I in road surface is exported to the calculating system for giving balance car as a result
R_angel is updated, realizes that Kalman filter interface adaptive changes, Q_angle=0.001 and Q_gyro values are constant, formula
It is as follows:
R_angel=I;
(3) according to complete calculating process in process 1, show that roll angle (pitch) is used as the final angle of inclination of balance car,
It is sent into electric machine controller and is calculated, draws simultaneously actuating motor controlled quentity controlled variable.
Process 7 is according to the measuring and calculating estimate of camera road pavement degree of disease, comparison threshold value I, to select ensuing electricity
Machine control strategy, this programme sets I=0.6 (value is not quite similar between different balance cars), when detected value is less than threshold value I
Wait, it is believed that flat during road surface, now motor control strategy selection fuzzy-adaptation PID control, if detected value is more than threshold value I,
It is cas PID control from motor control strategy, while the angle after optimization, angular speed measuring and calculating value is substituted into PID controller,
Draw simultaneously actuating motor controlled quentity controlled variable:
Comprise the following steps that:
(1) after calculating road bump degree estimate I, preserve and be eventually used for the decision-making foundation of motor control strategy.
(2) I is worked as(t)During less than I, system thinks that road surface is relatively flat, and now sensor measurement signal is steady, and noisy
Amount is few.Select fuzzy-adaptation PID control, according to its model possess preferable signal follow-up capability and the faster respond of system come
The operational ton that user provides is adapted to, also there is fuzzy controlled quentity controlled variable to perform accurate, simple operation and other advantages, specific formula in addition
It is as follows:
Parameter regulation formula is as follows:
Kp=Kp0+Δkp
Ki=Ki0+Δki
Kd=Kd0+Δkd
(3) I is worked as(t)During less than I, system thinks that road surface is more jolted, and now sensor measurement signal contains non-artificial shake
It is larger, and noisy amount is more.Cas PID control is selected, cascade PID has its simple structure relative to monocyclic PID, and control is accurate, increases
The strong anti-interference of system, but because its structure diversification causes integral term excessive and caused by system control hysteresis, this is for top
Road surface of winnowing with a dustpan is applicable just, and we require system appropriate blunt point of spike burr to caused by bumpy road, are so more favorable for
User security operates.
(4) after completing a controlling cycle, the arrival of waiting system clock, circulating repetition process 1 to process 7.
Using the available beneficial effect of technical solution of the present invention:
(1) under different pavement conditions, introduce visual cooperation inertial sensor and carry out data fusion, automatic detection road surface disease
Evil planarization is simultaneously quantified.
(2) under conditions of the complexity of road surface, the information of road surface that vision is drawn accesses Kalman's forecast model, implementation model
Mobilism is adaptive, can accelerate the followability of convergence rate and quickening prediction signal during filter recursion.
(3) information of road surface that vision is drawn is subjected to decision-making to select motor control method (cascade PID, fuzzy), blocked
The control signal that Germania predicts is input to PID controller, obtains being more suitable for the motor control amount on road surface.
(4) the visual information that vision sensor collects can also preserve, it is necessary to when can transfer carry out people from road surface
Early warning processing is analysed or carried out to work point, strengthens man-machine interaction, improves overall security.
Brief description of the drawings
Fig. 1 is particular flow sheet of the present invention;
Fig. 2 is adaptive input Kalman Prediction and fixed input Kalman Prediction convergence rate comparison diagram;
Fig. 3 calculates to adapt to input Kalman Prediction and fixed input Kalman's prediction scheme to 10 degree of signals and associated noises
Details enlarged drawing;
Fig. 4 is adaptive input Kalman Prediction and fixed input Kalman prediction scheme effect contrast figure.
The step response curve comparison diagram of Fig. 5 positions cascade PID and fuzzy.
Embodiment
The embodiment of the present invention is further described below in conjunction with the accompanying drawings, but is not the limit to the present invention
It is fixed.
Fig. 1 shows particular flow sheet of the present invention, a kind of Pavement Evaluation side combined based on LBP features with inertial sensor
Case and improved balance car angle meter method, the LBP textural characteristics collected in pavement image are counted using camera
The method that information carries out gradient difference obtains pavement texture distribution, coordinates the front and rear contrast difference of filtering of inertial sensor, by two
The estimation of a normalized pavement disease degree is asked in person's weighted sum, and the inertial sensor for improving balance car is surveyed
Calculation system, finally using pavement disease degree estimate as control strategy selection gist, and then optimize motor control, including it is following
Detailed process:
Using accelerometer and gyroscope, real-time angular speed and Y-axis of the collecting vehicle constitution heart in X-axis incline process 1 respectively
Angular data, the data fusion of gathered data signal is realized using Kalman filter, and denoising show that car body is i.e. front and rear in Y direction
The real-time inclination angle Pitch information of two side directions, the convergent angle information after being predicted, finally dived with the integration of gyroscope
Angle carries out difference;Comprise the following steps that:
(1) system for introducing a discrete control process, with a linear machine differential equation (Linear Stochastic
Difference equation) describe:
X (k | k-1)=AX (k-1 | k-1)+BU (k) (1)
According to this formula, gyroscope current time measured value is substituted into, is worked as afterwards plus the drift value of gyroscope
The predicted value Angle of preceding angle, being write as matrix form is exactly:
(2) prediction of covariance matrix, need to define two input values in calculating process, be respectively:Gyro sensors
The drift noise of device and the angle noise of acceierometer sensor,
P (k | k-1)=AP (k-1 | k-1) A'+Q (3)
Q wherein in formula is vectorCovariance matrix, i.e.,:
Because drift noise also have angle noise be it is separate, then;
Cov (Q_bias, Angel)=0 (5)
Cov (Angel, Q_bias)=0 (6)
Definition angular speed noise and gyroscope noise proportional coefficient in program are respectively:
Q_angle=0.001;Q_gyro=0.003;
(3) gain coefficient of Kalman filter is calculated, this is that a bivector gain coefficient is set toFor diagonal
The amendment of degree and angular speed, expression formula are:
Kg (k)=P (k | k-1) H '/(HP (k | k-1) H '+R) (7)
The constant R occurred in formula refers to the noise coefficient of acceleration measurement, coefficient original definition in program
For:R_angle=0.5;
(4) kalman gain coefficient amendment predicted value, the angle drawn using two sensors of accelerometer and gyroscope
Difference regard error amount Angle_err, the amendment of value, formula are predicted using the product of kalman gain coefficient and error
It is as follows:
X (k | k)=X (k | k-1)+Kg (k) (Z (k)-HX (k | k-1)) (8)
Update prediction error value and angle value simultaneously:
Angel+=K_0*Angel_err (9)
Q_Bias+=K_1*Angel_err (10)
(5) this renewal is carried out to matrix covariance P matrixes, the covariance being mainly used in during next iteration it is pre-
Survey, formula is as follows:
P (k | k-1)=(I-Kg (k) H) P (k | k-1) (11)
(6) difference, and every 20 times are done into the angle measurement X at k moment (k | k-1) and k moment angle predictions
It is average that an angle mean error measurement is counted in angular surveying,
Process 2 draws the inclination data of real-time angular speed and X-axis of the car body barycenter on Z axis by gyro sensor,
The roll (i.e. the real-time inclination angle in left and right sides direction) of car body barycenter horizontal direction X-axis is drawn by Kalman's fused filtering, and
Data are normalized with map function and draws the mathematical statistics feature (inclined degree) that pavement disease degree characterizes;Specific steps
It is:
(1) calculate Z axis acceleration information using gyroscope and accelerometer calculates X-axis angle information and carries out Kalman Prediction
Fusion, obtains the posterior estimate for converging on an exact value θr,
(2) by calculate in X-axis to angle Roll go to ask for its sine value sin θr, and this sine value is multiplied
Vehicle wheel base length D between upper body left and right wheels, obtaining h values can be to be expressed as the height that car body barycenter is propped up by uneven road surface
Degree,
H=D*sin θr;
(3) the height value h divided by the height 2d of chassis raised the car body that this is calculated by uneven road surface is obtained
To the slope Q of whole car, this Q value is finally characterized as road surface and jolted degree for balance car car body, when value is 1,
Then reach the limiting condition that vehicle body can bear maximum inclination, reach normalization effect, formula is as follows:
Q=D*sin θ/2d (13);
Process 3 constantly obtains balance car car body road ahead image information by imaging sensor, and to image information
Gray processing, image segmentation, LBP characters extraction statistics and normalization conversion process are carried out, is comprised the following steps that:
(1) counted step by step such as using COMS cameras collection road surface picture data signal a (t), extraction LBP features
Under:
1) it is converted into gray-scale map for RGB color image;
2) floating-point operation Structural Transformation is optimized;
3) image digital signal is split vertically, divided the image into as 1*10=10 sub-regions;So per height
The pixel resolution in region is 640*48;
(2) local binarization;Choose local auto-adaptive threshold method and carry out binaryzation, realize the accurate extraction of target texture;
Specific extraction is as follows step by step:
1) during image traversal, 3*3 greylevel window is extracted every time as object, and selected window center gray scale
It is worth the threshold X as this binaryzationi,j, by 8 adjacent pixels respectively with this threshold XijIt is compared, if being more than this
Individual threshold value is just entered as 1, on the contrary then be 0;
2) it is numbered from the upper left corner, thus produces the binary number of one 8 and (be followed successively by [bi-1, j-1 bi-1, j
Bi-1, j+1 bi, j+1 bi+1, j+1 bi+1, j bi+1, j-1 bi, j-1]);This 8 binary number is just in this
The LBP values of heart window pixel point;
3) 8 binary eight kinds of rotations arrangement modes are all included, selects the LBP of minimum as center window
The LPB values of pixel characterize the texture information of this point;
(3) LBP histogram gradient distribution statisticses are counted, the normalization of pavement disease degree estimate, are comprised the following steps that;
1) LBP statistics with histogram is carried out to every sub-regions figure, 0 to 255 distribution histogram is divided into 16 sections
Counted, the pixel number for counting each section accounts for global proportionality, obtains a set array Rnm (1 for including 16 elements
≦n≦10;1≦m≦16);By this 16 element difference absolute value computings, caused one new sequence rnk (1≤n≤10;1
≦k≦8);
rN,2*K=RN,2*K-RN,2*K-1
2) its array rnm summation Cn (1≤n≤10) is counted;
CNFixed interval is concentrated on, minimum value is just 0, and the meaning that it is characterized is road surface state the most flat, on the contrary
This Sn value nearer it is to 1, then shows that the degree of disease on road surface is serious;
Process 4 by the pavement disease degree quantized value Cn, the inclined degree Q that are obtained in imaging sensor and inertial sensor and
P carries out linear programming combined weighted summation, and comprehensive three's confidence level final decision draws the institute of a measurement pavement disease degree
The measurement variance of introducing, that is, sensor noise value, current car body Y-axis inclination angle are θ, estimation noise variance I, and formula is as follows:
The experience debugging method that process 5 is carried out according to each sensor accuracy and reliability.
First method:The weight K1 of feature is determined with experience debugging method, K2, the value of K3 three, three kinds are united
Meter feature is weighted summation, and three parameters of regulation are optimal until car body balance, and parameter regulation rule is as follows:
K1+K2=1;
K3<1
Second method:Carried out according to the corresponding optimal tuning parameter I of measured Cn, Q and P value under different road surfaces
Least square fitting carries out parameter tuning under matlab, fits the function of first order of suitable space-time, draws parameter K1,
K2 and K3, function are as follows:
Process 6 will be pre- as Kalman of lower a moment above for the introduced measurement noise estimate of variance I of pavement disease
The input value of the initial value of survey predicts the angle value of balance car car body of lower a moment, comprises the following steps that:
(1) Kalman filter implementation process in reference process 1, by the definition procedure noise covariance parameter in program
For:
Q_angle=0.001;Q_gyro=0.003;
(2) it is real-time that the introduced Noise Variance Estimation value I in road surface is exported to the calculating system for giving balance car as a result
R_angel is updated, realizes that Kalman filter interface adaptive changes, Q_angle=0.001 and Q_gyro values are constant, formula
It is as follows:
R_angel=I;
(3) according to complete calculating process in process 1, show that roll angle (pitch) is used as the final angle of inclination of balance car,
It is sent into electric machine controller and is calculated, draws simultaneously actuating motor controlled quentity controlled variable.
Process 7 is according to the measuring and calculating estimate of camera road pavement degree of disease, comparison threshold value I, to select ensuing electricity
Machine control strategy, this programme sets I=0.6, when detected value is less than threshold value I, it is believed that flat during road surface, now motor control
Policy selection fuzzy-adaptation PID control processed, it is cas PID control from motor control strategy if detected value is more than threshold value I,
The angle after optimization, angular speed measuring and calculating value are substituted into PID controller simultaneously, draw simultaneously actuating motor controlled quentity controlled variable:
Comprise the following steps that:
(1) after calculating road bump degree estimate I, preserve and be eventually used for the decision-making foundation of motor control strategy.
(2) I is worked as(t)During less than I, system thinks that road surface is relatively flat, and now sensor measurement signal is steady, and noisy
Amount is few.Select fuzzy-adaptation PID control, according to its model possess preferable signal follow-up capability and the faster respond of system come
The operational ton that user provides is adapted to, also there is fuzzy controlled quentity controlled variable to perform accurate, simple operation and other advantages, specific formula in addition
It is as follows:
Parameter regulation formula is as follows:
Kp=Kp0+Δkp
Ki=Ki0+Δki
Kd=Kd0+Δkd
(3) I is worked as(t)During less than I, system thinks that road surface is more jolted, and now sensor measurement signal contains non-artificial shake
It is larger, and noisy amount is more.Cas PID control is selected, cascade PID has its simple structure relative to monocyclic PID, and control is accurate, increases
The strong anti-interference of system, but because its structure diversification causes integral term excessive and caused by system control hysteresis, this is for top
Road surface of winnowing with a dustpan is applicable just, and we require system appropriate blunt point of spike burr to caused by bumpy road, are so more favorable for
User security operates.
(4) after completing a controlling cycle, the arrival of waiting system clock, circulating repetition 1---7 processes.
Fig. 2 is shown in process 5, works as K1, and in the case that K2 is 0, simple camera sensing device data feedback comes adaptive
The change Kalman Prediction of input and fixed input Kalman Prediction convergence rate contrast should be adjusted, is balanced in this programme for two kinds
Car angle estimates measuring and calculating scheme and carries out a contrast, and its Plays input angle is 10 degree of signal, as a whole two kinds of angles
Degree measuring and calculating scheme can keep the followability for measurement angle well, and for the convergence of accurate angle.It is but right
From the point of view of variance than result in two kinds of measuring and calculating schemes, adaptive input Kalman Prediction scheme improves 6 on traditional infrastructure
Percentage point.
Fig. 3 shows that two kinds of angle meter schemes for adaptive input prediction and Classical forecast are entered to 10 degree of signals and associated noises
The details amplification of row measuring and calculating, two kinds of measuring and calculating schemes all show consistent estimated performance, in identification in iteration initial time
It is more excellent that whom does not show.
Fig. 4 shows the Kalman Prediction of adaptive input and fixed input Kalman's prediction scheme Contrast on effect, with
Time integral, after iterations increase, the two shows certain difference, especially when significantly jumping occurs in measurement signal
Wait, adaptive prediction shows preferably convergence and less accumulated error.
Fig. 5 is the step response curve comparison diagram of cascade PID and fuzzy.It can be seen that fuzzy-adaptation PID control is responding
On time and static error is better than cas PID control, and this quick response may be not suitable for the control of bumpy road, and for
Time lag existing for cascade PID, which is used for bumpy road, to obtain more true controlled quentity controlled variable in the trend that the positive and negative counteracting of noise be present.
Using the available beneficial effect of technical solution of the present invention:
(1) under different pavement conditions, introduce visual cooperation inertial sensor and carry out data fusion, automatic detection road surface disease
Evil planarization is simultaneously quantified.
(2) under conditions of the complexity of road surface, the information of road surface that vision is drawn accesses Kalman's forecast model, implementation model
Mobilism is adaptive, can accelerate the followability of convergence rate and quickening prediction signal during filter recursion.
(3) information of road surface that vision is drawn is subjected to decision-making to select motor control method (cascade PID, fuzzy), blocked
The control signal that Germania predicts is input to PID controller, obtains being more suitable for the motor control amount on road surface.
(4) the visual information that vision sensor collects can also preserve, it is necessary to when can transfer carry out people from road surface
Early warning processing is analysed or carried out to work point, strengthens man-machine interaction, improves overall security.
Embodiments of the present invention are made that with detailed description above in association with accompanying drawing, but the present invention be not limited to it is described
Embodiment.To those skilled in the art, without departing from the principles and spirit of the present invention, these are implemented
Mode carries out various change, modification, replacement and modification and still fallen within protection scope of the present invention.
Claims (8)
- A kind of 1. balance car system monitoring method based on camera pavement detection, it is characterised in that:It will be gathered using camera The method that LBP textural characteristics statistical information into pavement image carries out gradient difference obtains pavement texture distribution, coordinates inertia The front and rear contrast difference of the filtering of sensor, estimating for normalized pavement disease degree is asked for into the two weighted sum Meter, is eventually used for improving the inertial sensor calculating system of balance car, and real-time adaptive changes Kalman filter and realizes prediction Angle is more accurate, while selects motor control strategy, including process in detail below according to cam feedback value:Process 1 is acquired data using accelerometer and gyroscope, realizes that gathered data signal melts using Kalman filter Close, denoising draws real-time inclination angle Pitch information of the car body in the i.e. front and rear sides direction of Y direction, convergent after being predicted Angle information, finally carry out difference with the integration angle of gyroscope;Process 2 draws the inclination data of real-time angular speed and X-axis of the car body barycenter on Z axis by gyro sensor, passes through Kalman's fused filtering draws the roll i.e. real-time inclination angle in left and right sides direction of car body barycenter horizontal direction X-axis, and to data Map function is normalized and draws the mathematical statistics feature that pavement disease degree characterizes;Process 3 constantly obtains balance car car body road ahead image information by imaging sensor, and image information is carried out Gray processing, image segmentation, LBP characters extraction statistics and normalization conversion process;The pavement disease degree quantized value Cn and P that are obtained in imaging sensor and inertial sensor are weighted and asked by process 4 With the two comprehensive confidence level final decision draws the introduced measurement variance of a measurement pavement disease degree, that is, passes Sensor noise figure, current car body Y-axis inclination angle are θ, estimation noise variance I, and formula is as follows:<mrow> <mi>I</mi> <mo>=</mo> <mrow> <mo>(</mo> <msub> <mi>K</mi> <mn>1</mn> </msub> <mo>*</mo> <mi>Q</mi> <mo>+</mo> <msub> <mi>K</mi> <mn>2</mn> </msub> <mo>*</mo> <msub> <mi>C</mi> <mi>n</mi> </msub> <mo>)</mo> </mrow> <mo>*</mo> <mi>&theta;</mi> <mo>+</mo> <msub> <mi>K</mi> <mn>3</mn> </msub> <mo>*</mo> <msub> <mover> <mi>P</mi> <mo>&OverBar;</mo> </mover> <mi>i</mi> </msub> </mrow>Process 5 carries out debugging method according to respective sensor accuracy and reliability;Process 6 using above for the introduced measurement noise estimate of variance I of pavement disease as Kalman Prediction of lower a moment The input value of initial value predicts the angle value of balance car car body of lower a moment;Process 7 is according to the measuring and calculating estimate of camera road pavement degree of disease, comparison threshold value I, to select ensuing motor control System strategy, when detected value is less than threshold value I, it is believed that flat during road surface, now motor control strategy selection fuzzy control System, be cas PID control from motor control strategy if detected value is more than threshold value I, while by the angle after optimization, Angular speed measuring and calculating value is substituted into PID controller, draws simultaneously actuating motor controlled quentity controlled variable:<mrow> <mi>P</mi> <mi>I</mi> <mi>D</mi> <mo>=</mo> <mfenced open = "{" close = ""> <mtable> <mtr> <mtd> <mrow> <mi>F</mi> <mi>u</mi> <mi>z</mi> <mi>z</mi> <mi>y</mi> <mi> </mi> <mi>P</mi> <mi>I</mi> <mi>D</mi> <mi> </mi> <mi>C</mi> <mi>o</mi> <mi>n</mi> <mi>t</mi> <mi>r</mi> <mi>o</mi> <mi>l</mi> <mi>l</mi> <mi>e</mi> <mi>r</mi> </mrow> </mtd> <mtd> <mrow> <msub> <mi>I</mi> <mrow> <mo>(</mo> <mi>t</mi> <mo>)</mo> </mrow> </msub> <mo>&le;</mo> <mi>I</mi> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <mi>P</mi> <mi>I</mi> <mi>D</mi> <mi> </mi> <mi>C</mi> <mi>a</mi> <mi>s</mi> <mi>c</mi> <mi>a</mi> <mi>d</mi> <mi>e</mi> <mi> </mi> <mi>C</mi> <mi>o</mi> <mi>n</mi> <mi>t</mi> <mi>r</mi> <mi>o</mi> <mi>l</mi> <mi>l</mi> <mi>e</mi> <mi>r</mi> </mrow> </mtd> <mtd> <mrow> <msub> <mi>I</mi> <mrow> <mo>(</mo> <mi>t</mi> <mo>)</mo> </mrow> </msub> <mo>&GreaterEqual;</mo> <mi>I</mi> </mrow> </mtd> </mtr> </mtable> </mfenced> <mo>.</mo> </mrow>
- 2. the balance car system monitoring method according to claim 1 based on camera pavement detection, it is characterised in that: In process 1, comprise the following steps that:(1) system for introducing a discrete control process, is described with a linear machine differential equation:X (k | k-1)=AX (k-1 | k-1)+BU (k) (1)According to this formula, gyroscope current time measured value is substituted into, plus having obtained working as anterior angle after the drift value of gyroscope The predicted value Angle of degree, being write as matrix form is exactly:<mrow> <mfenced open = "|" close = "|"> <mtable> <mtr> <mtd> <mi>A</mi> <mi>n</mi> <mi>g</mi> <mi>e</mi> <mi>l</mi> </mtd> </mtr> <mtr> <mtd> <mi>Q</mi> <mo>_</mo> <mi>b</mi> <mi>i</mi> <mi>a</mi> <mi>s</mi> </mtd> </mtr> </mtable> </mfenced> <mo>=</mo> <mfenced open = "|" close = "|"> <mtable> <mtr> <mtd> <mn>1</mn> </mtd> <mtd> <mrow> <mo>-</mo> <mi>d</mi> <mi>t</mi> </mrow> </mtd> </mtr> <mtr> <mtd> <mn>0</mn> </mtd> <mtd> <mn>1</mn> </mtd> </mtr> </mtable> </mfenced> <mfenced open = "|" close = "|"> <mtable> <mtr> <mtd> <mi>A</mi> <mi>n</mi> <mi>g</mi> <mi>l</mi> <mi>e</mi> </mtd> </mtr> <mtr> <mtd> <mi>Q</mi> <mo>_</mo> <mi>b</mi> <mi>i</mi> <mi>a</mi> <mi>s</mi> </mtd> </mtr> </mtable> </mfenced> <mo>+</mo> <mfenced open = "|" close = "|"> <mtable> <mtr> <mtd> <mi>d</mi> <mi>t</mi> </mtd> </mtr> <mtr> <mtd> <mn>0</mn> </mtd> </mtr> </mtable> </mfenced> <mi>G</mi> <mi>y</mi> <mi>r</mi> <mi>o</mi> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>2</mn> <mo>)</mo> </mrow> </mrow>(2) prediction of covariance matrix, need to define two input values in calculating process, be respectively:Gyro sensor The angle noise of drift noise and acceierometer sensor,P (k | k-1)=AP (k-1 | k-1) A '+Q (3)Q wherein in formula is vectorCovariance matrix, i.e.,:<mrow> <mfenced open = "|" close = "|"> <mtable> <mtr> <mtd> <mrow> <mi>cov</mi> <mrow> <mo>(</mo> <mi>A</mi> <mi>n</mi> <mi>g</mi> <mi>e</mi> <mi>l</mi> <mo>,</mo> <mi>A</mi> <mi>n</mi> <mi>g</mi> <mi>e</mi> <mi>l</mi> <mo>)</mo> </mrow> </mrow> </mtd> <mtd> <mrow> <mi>cov</mi> <mrow> <mo>(</mo> <mi>Q</mi> <mo>_</mo> <mi>b</mi> <mi>i</mi> <mi>a</mi> <mi>s</mi> <mo>,</mo> <mi>A</mi> <mi>n</mi> <mi>g</mi> <mi>e</mi> <mi>l</mi> <mo>)</mo> </mrow> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <mi>cov</mi> <mrow> <mo>(</mo> <mi>A</mi> <mi>n</mi> <mi>g</mi> <mi>e</mi> <mi>l</mi> <mo>,</mo> <mi>Q</mi> <mo>_</mo> <mi>b</mi> <mi>i</mi> <mi>a</mi> <mi>s</mi> <mo>)</mo> </mrow> </mrow> </mtd> <mtd> <mrow> <mi>cov</mi> <mrow> <mo>(</mo> <mi>Q</mi> <mo>_</mo> <mi>b</mi> <mi>i</mi> <mi>a</mi> <mi>s</mi> <mo>)</mo> </mrow> </mrow> </mtd> </mtr> </mtable> </mfenced> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>4</mn> <mo>)</mo> </mrow> </mrow>Because drift noise also have angle noise be it is separate, then;Cov (Q_bias, Angel)=0 (5)Cov (Angel, Q_bias)=0 (6)Definition procedure angle noise covariance parameter and measurement noise covariance parameter in program are respectively:Q_angle=0.001;Q_gyro=0.003;R_angle=0.5;(3) gain coefficient of Kalman filter is calculated, this is that a bivector gain coefficient is set toFor to angle with And the amendment of angular speed, expression formula are:Kg (k)=P (k | k-1) H '/(HP (k | k-1) H '+R) (7)The constant R occurred in formula refers to the noise coefficient of acceleration measurement, and coefficient original definition is in program:R_ Angle=0.5;(4) kalman gain coefficient amendment predicted value, the difference of the angle drawn using two sensors of accelerometer and gyroscope It is worthwhile to be error amount Angle_err, the amendment of value is predicted using the product of kalman gain coefficient and error, formula is such as Under:X (k | k)=X (k | k-1)+Kg (k) (Z (k)-HX (k | k-1)) (8)Update prediction error value and angle value simultaneously:Angel+=K_0*Angel_err (9)Q_Bias+=K_1*Angel_err (10)(5) this renewal is carried out to matrix covariance P matrixes, the prediction for the covariance being mainly used in during next iteration, Formula is as follows:P (k | k-1)=(I-Kg (k) H) P (k | k-1) (11)(6) difference, and every 20 angles are done into the angle measurement X at k moment (k | k-1) and k moment angle predictions It is average that an angle mean error measurement is counted in measurement,<mrow> <msub> <mover> <mi>P</mi> <mo>&OverBar;</mo> </mover> <mi>i</mi> </msub> <mo>=</mo> <munderover> <mo>&Sigma;</mo> <mrow> <mn>20</mn> <mrow> <mo>(</mo> <mi>i</mi> <mo>-</mo> <mn>1</mn> <mo>)</mo> </mrow> </mrow> <mrow> <mn>20</mn> <mi>i</mi> </mrow> </munderover> <mrow> <mo>(</mo> <mi>X</mi> <mo>(</mo> <mrow> <mi>k</mi> <mo>|</mo> <mi>k</mi> </mrow> <mo>)</mo> <mo>-</mo> <mi>X</mi> <mo>(</mo> <mrow> <mi>k</mi> <mo>|</mo> <mi>k</mi> <mo>-</mo> <mn>1</mn> </mrow> <mo>)</mo> <mo>)</mo> </mrow> <mo>/</mo> <mn>20</mn> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>12</mn> <mo>)</mo> </mrow> <mo>.</mo> </mrow>
- 3. the balance car system monitoring method according to claim 1 or 2 based on camera pavement detection, its feature exist In:In process 2, comprise the following steps that:(1) calculate Z axis acceleration information using gyroscope and accelerometer measuring and calculating X-axis angle information carries out Kalman Prediction and melted Close, obtain the posterior estimate for converging on an exact value θr,(2) by calculate in X-axis to angle Roll go to ask for its sine value sin θr, and this sine value is multiplied by car body Vehicle wheel base length D between left and right wheels, obtain h values can be expressed as the height that car body barycenter is propped up by uneven road surface,H=D*sin θr;(3) the height value h divided by the height 2d of chassis raised the car body that this is calculated by uneven road surface obtains whole The slope Q of car, this Q value is finally characterized as road surface for balance car tilting of car body degree, when value is 1, then reached The limiting safe state of maximum inclination is can bear to vehicle body, also reaches normalization effect, formula is as follows:Q=D*sin θ/2d (13).
- 4. the balance car system monitoring method according to claim 1 or 2 based on camera pavement detection, its feature exist In:In process 3, comprise the following steps that:(1) it is special using COMS cameras collection road surface picture data signal a (t), extraction LBP Sign is counted as follows step by step:1) it is converted into gray-scale map for RGB color image;<mrow> <msub> <mi>a</mi> <mrow> <msub> <mi>Gray</mi> <mrow> <mo>(</mo> <mi>t</mi> <mo>)</mo> </mrow> </msub> </mrow> </msub> <mo>=</mo> <mi>R</mi> <mo>*</mo> <mn>0.299</mn> <mo>+</mo> <mi>G</mi> <mo>*</mo> <mn>0.587</mn> <mo>+</mo> <mi>B</mi> <mo>*</mo> <mn>0.144</mn> </mrow>2) floating-point operation Structural Transformation is optimized;<mrow> <msub> <mi>a</mi> <mrow> <msub> <mi>Gray</mi> <mrow> <mo>(</mo> <mi>t</mi> <mo>)</mo> </mrow> </msub> </mrow> </msub> <mo>=</mo> <mrow> <mo>(</mo> <mi>R</mi> <mo>*</mo> <mn>299</mn> <mo>+</mo> <mi>G</mi> <mo>*</mo> <mn>587</mn> <mo>+</mo> <mi>B</mi> <mo>*</mo> <mn>114</mn> <mo>+</mo> <mn>500</mn> <mo>)</mo> </mrow> <mo>/</mo> <mn>1000</mn> </mrow>3) image digital signal is subjected to vertical divided in equal amounts, divided the image into as 1*10=10 sub-regions;So per height The pixel resolution in region is 640*48;<mfenced open = "" close = ""> <mtable> <mtr> <mtd> <mrow> <msub> <mi>S</mi> <mi>n</mi> </msub> <mo>=</mo> <munderover> <mo>&Sigma;</mo> <mrow> <mi>n</mi> <mo>=</mo> <mn>1</mn> </mrow> <mn>10</mn> </munderover> <msub> <mi>a</mi> <mrow> <mo>(</mo> <mi>t</mi> <mo>)</mo> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>n</mi> <mo>*</mo> <mi>j</mi> <mo>)</mo> </mrow> </msub> </mrow> </mtd> <mtd> <mtable> <mtr> <mtd> <mrow> <mn>0</mn> <mo><</mo> <mi>i</mi> <mo>&le;</mo> <mn>640</mn> </mrow> </mtd> <mtd> <mrow> <mi>i</mi> <mo>&Element;</mo> <mi>N</mi> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <mn>0</mn> <mo><</mo> <mi>j</mi> <mo>&le;</mo> <mn>48</mn> </mrow> </mtd> <mtd> <mrow> <mi>i</mi> <mo>&Element;</mo> <mi>N</mi> </mrow> </mtd> </mtr> </mtable> </mtd> </mtr> </mtable> </mfenced>(2) local binarization;Choose threshold method and carry out binaryzation, realize the accurate extraction of target texture;Specifically carry Take as follows step by step:1) during image traversal, 3*3 greylevel window is extracted every time as object, and selected window center gray value is made For the threshold X of this binaryzationi,j, by 8 adjacent pixels respectively with this threshold XijIt is compared, if being more than this threshold Value is just entered as 1, on the contrary then be 0;<mrow> <msub> <mi>a</mi> <msub> <mrow> <mo>(</mo> <mi>t</mi> <mo>)</mo> </mrow> <mrow> <mo>(</mo> <mi>i</mi> <mo>&PlusMinus;</mo> <mn>1</mn> <mo>,</mo> <mi>j</mi> <mo>&PlusMinus;</mo> <mn>1</mn> <mo>)</mo> </mrow> </msub> </msub> <mo>=</mo> <mfenced open = "{" close = ""> <mtable> <mtr> <mtd> <mn>1</mn> </mtd> <mtd> <mrow> <msub> <mi>a</mi> <msub> <mrow> <mo>(</mo> <mi>t</mi> <mo>)</mo> </mrow> <mrow> <mo>(</mo> <mi>i</mi> <mo>&PlusMinus;</mo> <mn>1</mn> <mo>,</mo> <mi>j</mi> <mo>&PlusMinus;</mo> <mn>1</mn> <mo>)</mo> </mrow> </msub> </msub> <mo>&GreaterEqual;</mo> <msub> <mi>a</mi> <msub> <mrow> <mo>(</mo> <mi>t</mi> <mo>)</mo> </mrow> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> </msub> </msub> </mrow> </mtd> </mtr> <mtr> <mtd> <mn>0</mn> </mtd> <mtd> <mrow> <msub> <mi>a</mi> <msub> <mrow> <mo>(</mo> <mi>t</mi> <mo>)</mo> </mrow> <mrow> <mo>(</mo> <mi>i</mi> <mo>&PlusMinus;</mo> <mn>1</mn> <mo>,</mo> <mi>j</mi> <mo>&PlusMinus;</mo> <mn>1</mn> <mo>)</mo> </mrow> </msub> </msub> <mo><</mo> <msub> <mi>a</mi> <msub> <mrow> <mo>(</mo> <mi>t</mi> <mo>)</mo> </mrow> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> </msub> </msub> </mrow> </mtd> </mtr> </mtable> </mfenced> </mrow>2) it is numbered from the upper left corner, thus produces the binary number of one 8 and (be followed successively by [bi-1, j-1 bi-1, j bi- 1, j+1bi, j+1bi+1, j+1bi+1, j bi+1, j-1bi, j-1]);This 8 binary number is just this center window The LBP values of pixel;LBP(t)(i,j)=sum [(bi-1,j-1,bi-1,j,bi-1,j+1,bi,j-1,bi,j,bi,j+1,bi+1,j-1,bi+1,jbi+1,j+1)]3) 8 binary eight kinds of rotations arrangement modes are all included, selects the LBP of minimum as center window pixel The LPB values of point characterize the texture information of this point;<mfenced open = "" close = ""> <mtable> <mtr> <mtd> <mrow> <msub> <mi>LBP</mi> <mrow> <mo>(</mo> <mi>t</mi> <mo>)</mo> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> </msub> <mo>=</mo> <mi>m</mi> <mi>i</mi> <mi>n</mi> <mo>&lsqb;</mo> <mo>(</mo> <msub> <mi>LBP</mi> <msub> <mrow> <mo>(</mo> <mi>t</mi> <mo>)</mo> </mrow> <mrow> <mi>i</mi> <mo>-</mo> <mn>1</mn> <mo>,</mo> <mi>j</mi> <mo>-</mo> <mn>1</mn> </mrow> </msub> </msub> <mo>,</mo> <msub> <mi>LBP</mi> <msub> <mrow> <mo>(</mo> <mi>t</mi> <mo>)</mo> </mrow> <mrow> <mi>i</mi> <mo>-</mo> <mn>1</mn> <mo>,</mo> <mi>j</mi> </mrow> </msub> </msub> <mo>,</mo> <msub> <mi>LBP</mi> <msub> <mrow> <mo>(</mo> <mi>t</mi> <mo>)</mo> </mrow> <mrow> <mi>i</mi> <mo>-</mo> <mn>1</mn> <mo>,</mo> <mi>j</mi> <mo>+</mo> <mn>1</mn> </mrow> </msub> </msub> <mo>,</mo> <msub> <mi>LBP</mi> <msub> <mrow> <mo>(</mo> <mi>t</mi> <mo>)</mo> </mrow> <mrow> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>-</mo> <mn>1</mn> </mrow> </msub> </msub> <mo>,</mo> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <msub> <mi>LBP</mi> <msub> <mrow> <mo>(</mo> <mi>t</mi> <mo>)</mo> </mrow> <mrow> <mi>i</mi> <mo>,</mo> <mi>j</mi> </mrow> </msub> </msub> <mo>,</mo> <msub> <mi>LBP</mi> <msub> <mrow> <mo>(</mo> <mi>t</mi> <mo>)</mo> </mrow> <mrow> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>+</mo> <mn>1</mn> </mrow> </msub> </msub> <mo>,</mo> <msub> <mi>LBP</mi> <msub> <mrow> <mo>(</mo> <mi>t</mi> <mo>)</mo> </mrow> <mrow> <mi>i</mi> <mo>+</mo> <mn>1</mn> <mo>,</mo> <mi>j</mi> <mo>-</mo> <mn>1</mn> </mrow> </msub> </msub> <mo>,</mo> <msub> <mi>LBP</mi> <mrow> <mo>(</mo> <mi>t</mi> <mo>)</mo> <msub> <mrow> <mo>(</mo> <mi>i</mi> <mo>,</mo> <mi>j</mi> <mo>)</mo> </mrow> <mrow> <mi>i</mi> <mo>+</mo> <mn>1</mn> <mo>,</mo> <mi>j</mi> </mrow> </msub> </mrow> </msub> <msub> <mi>LBP</mi> <msub> <mrow> <mo>(</mo> <mi>t</mi> <mo>)</mo> </mrow> <mrow> <mi>i</mi> <mo>+</mo> <mn>1</mn> <mo>,</mo> <mi>j</mi> <mo>+</mo> <mn>1</mn> </mrow> </msub> </msub> <mo>)</mo> <mo>&rsqb;</mo> </mrow> </mtd> </mtr> </mtable> </mfenced>(3) LBP histogram gradient distribution statisticses are counted, the normalization of pavement disease degree estimate, are comprised the following steps that;1) LBP statistics with histogram is carried out to every sub-regions figure, 0 to 255 distribution histogram is divided into 16 sections is carried out Statistics, the pixel number for counting each section account for global proportionality, obtain a set array Rnm (1≤n for including 16 elements ≦10;1≦m≦16);By this 16 element difference absolute value computings, caused one new sequence rnk (1≤n≤10;1≦k ≤ 8) LBP histograms probability density distribution gradient, is characterized with this;rN,2*K=RN,2*K-RN,2*K-12) its array rnm summation Cn (1≤n≤10) is counted;<mrow> <msub> <mi>C</mi> <mi>N</mi> </msub> <mo>=</mo> <munderover> <mo>&Sigma;</mo> <mrow> <mi>k</mi> <mo>=</mo> <mn>1</mn> </mrow> <mn>8</mn> </munderover> <msub> <mi>r</mi> <mrow> <mi>N</mi> <mo>,</mo> <mn>2</mn> <mo>*</mo> <mi>K</mi> </mrow> </msub> </mrow>CNConcentrate on fixed interval, minimum value just for 0 (now pavement texture distribution is minimum), the meaning that it is characterized be road surface most For flat state, this Cn value on the contrary nearer it is to 1 (texture contrast is maximum), then show that the degree of disease on road surface is serious.
- 5. the balance car system monitoring method according to claim 1 or 2 based on camera pavement detection, its feature exist In:In process 5, comprise the following steps that:The weight K1 of feature is determined with experience debugging method, K2, the value of K3 three will Three kinds of statistical natures are weighted summation, and three parameters of regulation are optimal until car body balance, and parameter regulation rule is as follows:<mrow> <mtable> <mtr> <mtd> <mrow> <msub> <mi>K</mi> <mn>1</mn> </msub> <mo>+</mo> <msub> <mi>K</mi> <mn>2</mn> </msub> <mo>=</mo> <mn>1</mn> </mrow> </mtd> </mtr> <mtr> <mtd> <mrow> <msub> <mi>K</mi> <mn>3</mn> </msub> <mo><</mo> <mn>1</mn> </mrow> </mtd> </mtr> </mtable> <mo>.</mo> </mrow>
- 6. the balance car system monitoring method according to claim 1 or 2 based on camera pavement detection, its feature exist In:In process 5, comprise the following steps that:According to the corresponding optimal tuning parameter I of measured Cn, Q and P value under different road surfaces To carry out under matlab least square fitting to carry out parameter tuning, the function of first order of suitable space-time is fitted, is drawn Parameter K1, K2 and K3, function are as follows:<mrow> <mo>(</mo> <msub> <mi>K</mi> <mn>1</mn> </msub> <mo>*</mo> <mi>Q</mi> <mo>+</mo> <msub> <mi>K</mi> <mn>2</mn> </msub> <mo>*</mo> <msub> <mi>C</mi> <mi>n</mi> </msub> <mo>)</mo> <mo>*</mo> <mi>&theta;</mi> <mo>+</mo> <msub> <mi>K</mi> <mn>3</mn> </msub> <mo>*</mo> <msub> <mover> <mi>P</mi> <mo>&OverBar;</mo> </mover> <mi>i</mi> </msub> <mo>=</mo> <mi>I</mi> <mo>.</mo> </mrow>
- 7. the balance car system monitoring method according to claim 1 or 2 based on camera pavement detection, its feature exist In:In process 6, comprise the following steps that:(1) Kalman filter implementation process in reference process 1, it is by the definition procedure noise covariance parameter in program:Q_angle=0.001;Q_gyro=0.003;(2) the introduced Noise Variance Estimation value I in road surface is exported as a result and gives the calculating system of balance car and carry out real-time update R_angel, realize that Kalman filter interface adaptive changes, Q_angle=0.001 and Q_gyro values are constant, and formula is as follows:R_angel=I;(3) according to complete calculating process in process 1, show that roll angle as the final angle of inclination of balance car, is sent into motor control Calculated in device processed, draw simultaneously actuating motor controlled quentity controlled variable.
- 8. the balance car system monitoring method according to claim 1 or 2 based on camera pavement detection, its feature exist In:In process 7, comprise the following steps that:(1) after calculating road bump degree estimate I, preserve and be eventually used for the decision-making foundation of motor control strategy;(2) I is worked as(t)During less than I, system thinks that road surface is relatively flat, and now sensor measurement signal is steady, and noisy amount is few; Fuzzy-adaptation PID control is selected, preferable signal follow-up capability and the faster respond of system are possessed according to its model to adapt to The operational ton that user provides, fuzzy is also accurate with controlled quentity controlled variable execution in addition, simple operation and other advantages, and specific formula is as follows:<mrow> <mi>u</mi> <mrow> <mo>(</mo> <mi>t</mi> <mo>)</mo> </mrow> <mo>=</mo> <msub> <mi>k</mi> <mi>p</mi> </msub> <mrow> <mo>(</mo> <mi>e</mi> <mo>(</mo> <mi>t</mi> <mo>)</mo> <mo>+</mo> <mfrac> <mn>1</mn> <msub> <mi>T</mi> <mi>i</mi> </msub> </mfrac> <mo>&Integral;</mo> <mi>e</mi> <mo>(</mo> <mi>t</mi> <mo>)</mo> <mi>d</mi> <mi>t</mi> <mo>+</mo> <msub> <mi>T</mi> <mi>D</mi> </msub> <mo>*</mo> <mi>d</mi> <mi>e</mi> <mo>(</mo> <mi>t</mi> <mo>)</mo> <mo>/</mo> <mi>d</mi> <mi>t</mi> <mo>)</mo> </mrow> </mrow>Parameter regulation formula is as follows:Kp=Kp0+ΔkpKi=Ki0+ΔkiKd=Kd0+Δkd(3) I is worked as(t)During less than I, system thinks that road surface is more jolted, now sensor measurement signal contain it is non-artificial shake compared with Greatly, and noisy amount is more;Cas PID control is selected, cascade PID has its simple structure relative to monocyclic PID, and control is accurate, enhancing The anti-interference of system, but because its structure diversification causes integral term excessive and caused by system control hysteresis, this is for jolting Road surface is applicable just, it is desirable to system appropriate blunt point of spike burr to caused by bumpy road, is so more favorable for user's peace Full operation;(4) after completing a controlling cycle, the arrival of waiting system clock, circulating repetition process 1 to process 7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710515182.XA CN107340298B (en) | 2017-06-29 | 2017-06-29 | Balance car system measurement and control method based on camera road surface detection |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710515182.XA CN107340298B (en) | 2017-06-29 | 2017-06-29 | Balance car system measurement and control method based on camera road surface detection |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107340298A true CN107340298A (en) | 2017-11-10 |
CN107340298B CN107340298B (en) | 2020-01-31 |
Family
ID=60218102
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710515182.XA Active CN107340298B (en) | 2017-06-29 | 2017-06-29 | Balance car system measurement and control method based on camera road surface detection |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107340298B (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108973576A (en) * | 2018-08-12 | 2018-12-11 | 苏州青科艾莉电子科技有限公司 | A kind of anti-rollover regulation method of road vehicle high-performance electronic control air suspension |
CN109507871A (en) * | 2018-12-11 | 2019-03-22 | 广东工业大学 | Pid parameter setting method and product for the control of two-wheeled balance car car body balance |
CN110554706A (en) * | 2019-09-25 | 2019-12-10 | 江苏理工学院 | visual navigation self-balancing vehicle and balancing method |
CN114495068A (en) * | 2022-04-18 | 2022-05-13 | 河北工业大学 | Road surface health detection 'element' method based on man-machine interaction and deep learning |
CN114723637A (en) * | 2022-04-27 | 2022-07-08 | 上海复瞰科技有限公司 | Color difference adjusting method and system |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2001247296A (en) * | 2000-03-08 | 2001-09-11 | Komatsu Forklift Co Ltd | Lpg cylinder bracket also used as truck |
JP2003165431A (en) * | 2001-11-30 | 2003-06-10 | Honda Motor Co Ltd | Body slip angle estimating method |
CN101950156A (en) * | 2010-09-06 | 2011-01-19 | 重庆大学 | Adaptive cascade PID control method |
CN102538781A (en) * | 2011-12-14 | 2012-07-04 | 浙江大学 | Machine vision and inertial navigation fusion-based mobile robot motion attitude estimation method |
CN104657724A (en) * | 2015-03-12 | 2015-05-27 | 福建依图网络科技有限公司 | Method for detecting pedestrians in traffic videos |
CN204423150U (en) * | 2015-02-05 | 2015-06-24 | 重庆交通大学 | Based on the two-wheeled control system of balance car of image |
CN104932505A (en) * | 2015-06-08 | 2015-09-23 | 华南理工大学 | Automatic navigation system based on camera detection, control method of system and vertical balance car |
CN106054931A (en) * | 2016-07-29 | 2016-10-26 | 北方工业大学 | Unmanned aerial vehicle fixed-point flight control system based on visual positioning |
CN106353782A (en) * | 2015-07-16 | 2017-01-25 | 深圳市华信天线技术有限公司 | Ground point coordinates measuring method and device |
-
2017
- 2017-06-29 CN CN201710515182.XA patent/CN107340298B/en active Active
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2001247296A (en) * | 2000-03-08 | 2001-09-11 | Komatsu Forklift Co Ltd | Lpg cylinder bracket also used as truck |
JP2003165431A (en) * | 2001-11-30 | 2003-06-10 | Honda Motor Co Ltd | Body slip angle estimating method |
CN101950156A (en) * | 2010-09-06 | 2011-01-19 | 重庆大学 | Adaptive cascade PID control method |
CN102538781A (en) * | 2011-12-14 | 2012-07-04 | 浙江大学 | Machine vision and inertial navigation fusion-based mobile robot motion attitude estimation method |
CN204423150U (en) * | 2015-02-05 | 2015-06-24 | 重庆交通大学 | Based on the two-wheeled control system of balance car of image |
CN104657724A (en) * | 2015-03-12 | 2015-05-27 | 福建依图网络科技有限公司 | Method for detecting pedestrians in traffic videos |
CN104932505A (en) * | 2015-06-08 | 2015-09-23 | 华南理工大学 | Automatic navigation system based on camera detection, control method of system and vertical balance car |
CN106353782A (en) * | 2015-07-16 | 2017-01-25 | 深圳市华信天线技术有限公司 | Ground point coordinates measuring method and device |
CN106054931A (en) * | 2016-07-29 | 2016-10-26 | 北方工业大学 | Unmanned aerial vehicle fixed-point flight control system based on visual positioning |
Non-Patent Citations (2)
Title |
---|
ZHANG TIEMIN,ET.AL.: ""Attitude measure system based on extended Kalman filter"", 《COMPUTERS AND ELECTRONICS IN AGRICULTURE》 * |
徐威等: ""基于图像显著性的路面裂缝检测"", 《中国图象图形学报》 * |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108973576A (en) * | 2018-08-12 | 2018-12-11 | 苏州青科艾莉电子科技有限公司 | A kind of anti-rollover regulation method of road vehicle high-performance electronic control air suspension |
CN109507871A (en) * | 2018-12-11 | 2019-03-22 | 广东工业大学 | Pid parameter setting method and product for the control of two-wheeled balance car car body balance |
CN109507871B (en) * | 2018-12-11 | 2022-03-25 | 广东工业大学 | PID parameter setting method and product for two-wheel balance vehicle body balance control |
CN110554706A (en) * | 2019-09-25 | 2019-12-10 | 江苏理工学院 | visual navigation self-balancing vehicle and balancing method |
CN114495068A (en) * | 2022-04-18 | 2022-05-13 | 河北工业大学 | Road surface health detection 'element' method based on man-machine interaction and deep learning |
CN114723637A (en) * | 2022-04-27 | 2022-07-08 | 上海复瞰科技有限公司 | Color difference adjusting method and system |
Also Published As
Publication number | Publication date |
---|---|
CN107340298B (en) | 2020-01-31 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107340298A (en) | Balance car system monitoring method based on camera pavement detection | |
CN111862157B (en) | Multi-vehicle target tracking method integrating machine vision and millimeter wave radar | |
CN109949375B (en) | Mobile robot target tracking method based on depth map region of interest | |
CN104200485B (en) | Video-monitoring-oriented human body tracking method | |
CN105405151B (en) | Anti-shelter target tracking based on particle filter and weighting Surf | |
CN105835880B (en) | Lane following system | |
CN108572648B (en) | Unmanned vehicle power supply multi-source fusion prediction method and system | |
CN110688987A (en) | Pedestrian position detection and tracking method and system | |
CN107330376A (en) | A kind of Lane detection method and system | |
CN104183127A (en) | Traffic surveillance video detection method and device | |
CN103324936B (en) | A kind of vehicle lower boundary detection method based on Multi-sensor Fusion | |
CN103903019A (en) | Automatic generating method for multi-lane vehicle track space-time diagram | |
CN107750364A (en) | Detected using the road vertically profiling of stable coordinate system | |
CN104318258A (en) | Time domain fuzzy and kalman filter-based lane detection method | |
WO2020006764A1 (en) | Path detection method, related device, and computer readable storage medium | |
CN104574439A (en) | Kalman filtering and TLD (tracking-learning-detection) algorithm integrated target tracking method | |
CN101916446A (en) | Gray level target tracking algorithm based on marginal information and mean shift | |
CN106969770A (en) | A kind of robot and its air navigation aid, computer-readable recording medium | |
CN110348332A (en) | The inhuman multiple target real-time track extracting method of machine under a kind of traffic video scene | |
CN105894540A (en) | Method and system for counting vertical reciprocating movements based on mobile terminal | |
CN111369629A (en) | Ball return trajectory prediction method based on binocular visual perception of swinging, shooting and hitting actions | |
CN109708658A (en) | A kind of visual odometry method based on convolutional neural networks | |
CN107808524A (en) | A kind of intersection vehicle checking method based on unmanned plane | |
CN111123953B (en) | Particle-based mobile robot group under artificial intelligence big data and control method thereof | |
CN111145211B (en) | Method for acquiring pixel height of head of upright pedestrian of monocular camera |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |