A lane line processing method and system based on Kalman filtering
By considering the parallel constraint relationship between left and right lane lines and dynamic noise adjustment in lane line tracing, the anomaly impact and discontinuity problems of lane line tracing in the prior art are solved, and more stable lane line tracing and lateral control are achieved.
Patent Information
- Application Number
- CN202111452500.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2021-11-30
- Publication Date
- 2024-11-22
- Estimated Expiration
- 2041-11-30
AI Technical Summary
The existing lane line tracing method based on Kalman filtering cannot effectively deal with the relationship between abnormal observation points and ignore the lane lines, resulting in abnormal jumps and discontinuities in the lane width.
By considering the parallel constraint relationship between left and right lane lines, a cubic curve model is established, a Kalman filter is used to update the state, and the observation noise is dynamically adjusted to avoid the influence of abnormal points and ensure the continuity and stability of the lane lines.
The stable tracking of lane lines is achieved, abnormal jumps and discontinuities of lane width are avoided, and the stability of lateral control is enhanced, which is especially suitable for centering control strategies.
Smart Images

Figure CN114170275B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of autonomous driving technology, and in particular to a lane line processing method and system based on Kalman filtering. Background Art
[0002] At present, there are many ways to track lane lines. For example, Chinese patent document CN102360499A discloses a multi-lane line tracking method based on a Kalman filter group. This method establishes a Kalman filter model for lane line tracking. On the basis of lane line detection in the collected road image in front of the vehicle, the Kalman filter model is used to output the optimal estimate of the lane line. When the same lane line is detected in 25 consecutive road images, the optimal estimate of the Kalman filter is considered to be valid, and the optimal estimate output by the Kalman filter is used as the detection result. This method is suitable for lane lines represented by lateral distance and azimuth. The observation noise uses a fixed value, but it cannot respond to the currently abnormal observed lane line. Each lane line is an independent Kalman filter, ignoring the relationship between lane lines. Summary of the invention
[0003] The purpose of the present invention is to provide a lane line processing method and system based on Kalman filtering, which can avoid the influence of abnormal points on the fitting results, avoid the discontinuity between the frames before and after the lane line, and suppress the abnormal jump of the lane width; consider the parallel constraint relationship between the left and right lane lines, model the center line, and make the lateral control more stable for the system adopting the centering control strategy.
[0004] To achieve the above object, the present invention provides a lane line processing method based on Kalman filtering, the steps comprising:
[0005] (S1) Obtain lane line information and determine whether the lane line information of the previous frame is valid. If so, proceed to step (S2); otherwise, reinitialize the lane line and the process ends;
[0006] (S2) Initialize the Kalman filter state vector by taking the center line of the lane in the current frame, and select the cubic curve coefficient, coefficient change rate and lane width w used to characterize the center line of the lane as the state quantity, that is, the state vector is X k =[A k , A′ k , B k , B′ k , C k , D k , w k ] T ; Among them, the state vector X at the current time k k Through the state vector X at the last moment k-1 k-1 It is inferred that Xk =F·X k-1 , the state transfer matrix F is established according to the uniform motion model, A k , B k , C k , D k and w k They represent the state quantity at time k predicted based on time k-1 respectively; where the curve equation of the center line of the lane and the curve equations of the left and right lane lines are:
[0007] y center =A t +B t x+C t x 2 +D t x 3 ;
[0008]
[0009]
[0010] Among them, A t , B t , C t and D t They represent the coefficients of the curve equation of the center line at time t respectively; and They represent the coefficients of the curve equation of the left lane line at time t respectively; and They represent the coefficients of the curve equation of the right lane line at time t respectively;
[0011] (S3) Obtain the observations, observation matrix and observation noise matrix of the left lane line and the right lane line respectively:
[0012] The center position of the front end of the vehicle is established as the origin, the vehicle's forward direction is the x-axis, and the direction perpendicular to the vehicle's forward direction is the y-axis. The left lane line and the right lane line of the current frame are both used as the observation lane lines. m discrete points (x i ,y l (x i )) and m discrete points on the right lane line (x i ,y r (x i )); wherein, i=1, 2, 3, ..., m, and m is greater than or equal to 4;
[0013] y l (x i ) and the lane width w are used as the left lane observation quantity, and the observation vector of the left lane is yr (x i ) and the lane width w are used as the left lane observation, and the observation vector of the right lane is
[0014] The m discrete points on the left lane line and the m discrete points on the right lane line are both used as observation quantities for state update. The conversion formula from state quantity to observation quantity is:
[0015]
[0016]
[0017] Therefore, the observation matrix H of the left lane is l And the observation matrix H of the right lane r They are:
[0018]
[0019] Calculate the matching degree of each discrete point of the left lane line respectively The matching degree of each discrete point of the right lane line The matching degree and Discrete points larger than the preset threshold are eliminated. For discrete points within the preset threshold range, the left lane and right lane observation noise matrices are dynamically adjusted according to the matching degree;
[0020] (S4) The state vector X is calculated in accordance with the Kalman wave formula according to the observations, observation matrix and observation noise matrix of the left lane and the observations, observation matrix and observation noise matrix of the right lane, or the observations, observation matrix and observation noise matrix of the right lane and the observations, observation matrix and observation noise matrix of the left lane. k and the covariance matrix P k Update the state and obtain the state vector X′ of the center line of the lane at time k, which is the final output of the Kalman filter k =[A k′ , A′ k′ , B k′ , B′ k′ , C k′ , D k′ , w k′ ] T , then the centerline curve of the lane at the current time k is: y k =A k′ +B k′ x+C k′ x 2 +D k′ x 3 ; Among them, the covariance matrix P kUpdated according to the Kalman filter formula, the initial value can be set to the unit matrix or determined according to the actual application;
[0021] (S5) According to the parallel relationship between the left and right lane lines and the relationship with the center line of the lane, the left lane line and the right lane line are generated. The curve equations of the left lane line and the right lane line at the current time k are:
[0022] y l =A k′ +B k′ x+C k′ x 2 +D k′ x 3 -w / 2;
[0023] y r =A k′ +B k′ x+C k′ x 2 +D k′ x 3 +w / 2.
[0024] Further, the Kalman filter state update formula is as follows:
[0025]
[0026] X′ k =X k +K k (Z k -H k X k );
[0027] P′ k =(IK k H k ) k ;
[0028] Among them, I is the identity matrix, K k is the Kalman filter gain at time k, X′ k is the state vector at time k output by the Kalman filter, P′ k is the covariance of the Kalman filter output at time k, P k For X k The covariance of k is the observation matrix at time k, H k is the transposed matrix of , and Zk is the observation vector at time k.
[0029] Further, the state transfer matrix F is:
[0030] Among them, dt is the time interval from time k-1 to time k, dl represents the distance the vehicle moves in a uniform straight line during time dt, that is, dl=v·dt, v is the speed of the vehicle.
[0031] Furthermore, the establishment of taking the center position of the front end of the vehicle as the origin is specifically: establishing taking the center position of the front bumper of the vehicle as the origin.
[0032] Furthermore, after the lane lines are reinitialized, the following steps are performed: determining whether the lane lines are initialized successfully, if so, saving the current frame; otherwise, the current frame processing ends; wherein the condition for successful initialization is that the lane lines on both sides of the lane exist at the same time and are parallel to each other.
[0033] Furthermore, the steps for determining whether the lane lines on both sides of the lane are parallel to each other are as follows:
[0034] Select several points with the same x value on the left and right lane curves respectively;
[0035] If the width deviation at each point ||y r -y l |-w|<thresh, the left and right lane lines are considered to be approximately parallel, where thresh is the preset deviation threshold.
[0036] Furthermore, the matching degree of each discrete point of the left lane line The matching degree of each discrete point of the right lane line The calculation formula is as follows:
[0037]
[0038]
[0039] The average curve of the left lane line in the past n frames is used as the left lane baseline The average curve of the right lane line in the past n frames is used as the right lane baseline n is greater than or equal to 2.
[0040] Furthermore, in step (S3), the discrete points are selected in a manner of equally spaced points within the length of the lane line. i The selection method is: range is the detectable range of the lane line.
[0041] Furthermore, the left lane observation noise matrix R l and the right lane observation noise matrix R r is a (m+1)*(m+1) dimensional matrix. The diagonal position of the left lane observation noise matrix is R l(m+1, m+1) is equal to the variance of the lane width w, and the rest are 0; the diagonal position of the right lane observation noise matrix Rr(m+1, m+1) is equal to the variance of the lane width w, and is 0 at other locations.
[0042] The present invention also provides a lane line processing system based on Kalman filtering, comprising:
[0043] An image acquisition module, used to obtain lane line information;
[0044] Data processing module, used for lane line processing based on Kalman filtering and generating left and right lane lines;
[0045] The image acquisition module is connected to the data processing module, and the lane line processing system based on Kalman filtering is configured to execute the steps of the lane line processing method based on Kalman filtering.
[0046] Compared with the prior art, the present invention has the following advantages:
[0047] The lane line processing method and system based on Kalman filtering of the present invention consider the parallel constraint relationship between the left and right lane lines, track the width of the lane, and translate the center line according to the lane width to form parallel lane lines; the center line is modeled to make the lateral control more stable for the system adopting the centering control strategy; the state quantity is the cubic curve coefficient, and the observation quantity is a discrete point, which reduces the step of finally fitting the discrete points into the cubic curve; at the same time, the observation quantity is a discrete point, which can avoid the influence of abnormal points on the fitting result, and the dynamic weight makes the fitting result more stable; it can also avoid the discontinuity between the frames before and after the lane line, and suppress the abnormal jump of the lane width. BRIEF DESCRIPTION OF THE DRAWINGS
[0048] Figure 1 This is a flow chart of a lane line processing method based on Kalman filtering of the present invention;
[0049] Figure 2 It is a schematic diagram of the vehicle coordinate system of the present invention. DETAILED DESCRIPTION
[0050] The specific implementation of the present invention will be further described below in conjunction with the accompanying drawings.
[0051] The lane line processing method based on Kalman filtering of the present invention is used for lane line processing in autonomous driving, and more specifically involves processing of target-level lane lines output by a smart camera by a fusion end in autonomous driving functions including IACC (Integrated Adapted Cruise Control) and TJP (Traffic Jam Pilot).
[0052] See also Figure 1 and Figure 2 As shown, this embodiment discloses a lane line processing method based on Kalman filtering, the steps comprising:
[0053] (S1) Acquire lane line information and determine whether the lane line information of the previous frame is valid. If so, go to step (S2); otherwise, reinitialize the lane line and the process ends. On structured roads, in most cases, the left and right lane lines of the lane are parallel to each other. In most cases, lateral control adopts a centering driving strategy. Parallel lane lines are more conducive to the stability of lateral control.
[0054] (S2) Initialize the Kalman filter state vector by taking the center line of the lane in the current frame, and select the cubic curve coefficient, coefficient change rate and lane width w used to characterize the center line of the lane as the state quantity, that is, the state vector is X k =[A k , A′ k , B k , B′ k , C k , D k , w k ] T ; Among them, the state vector X at the current time k k Through the state vector X at the last moment k-1 k-1 It is inferred that X k =F·X k-1 , the state transfer matrix F is established according to the uniform motion model, A k , B k , C k , D k and w k They represent the state quantity at time k predicted based on time k-1 respectively; where the curve equation of the center line of the lane and the curve equations of the left and right lane lines are:
[0055] y center =A t +B t x+C t x 2 +D t x 3 ;
[0056]
[0057]
[0058] Among them, A t , B t , C t and Dt They represent the coefficients of the curve equation of the center line at time t respectively; and They represent the coefficients of the curve equation of the left lane line at time t respectively; and They represent the coefficients of the curve equation of the right lane line at time t respectively;
[0059] Infer the state value at time k from the state value at time k-1: This process is mainly based on the vehicle's kinematic model. According to the vehicle speed v and yaw rate yawrate within ΔT time, the state vector at time k can be calculated.
[0060] (S3) Obtain the observations, observation matrix and observation noise matrix of the left lane line and the right lane line respectively:
[0061] The center position of the front end of the vehicle is established as the origin, the vehicle's forward direction is the x-axis, and the direction perpendicular to the vehicle's forward direction is the y-axis. The left lane line and the right lane line of the current frame are both used as the observation lane lines. m discrete points (x i ,y l (x i )) and m discrete points on the right lane line (x i ,y r (x i )); wherein, i=1, 2, 3, ..., m, and m is greater than or equal to 4;
[0062] y l (x i ) and the lane width w are used as the left lane observation quantity, and the observation vector of the left lane is y r (x i ) and the lane width w are used as the left lane observation, and the observation vector of the right lane is
[0063] The m discrete points on the left lane line and the m discrete points on the right lane line are both used as observation quantities for state update. The conversion formula from state quantity to observation quantity is:
[0064]
[0065]
[0066] Therefore, the observation matrix H of the left lane is l And the observation matrix H of the right lane r They are:
[0067]
[0068] Calculate the matching degree of each discrete point of the left lane line respectively The matching degree of each discrete point of the right lane line The matching degree and Discrete points larger than the preset threshold are eliminated. For discrete points within the preset threshold range, the left lane and right lane observation noise matrices are dynamically adjusted according to the matching degree;
[0069] Specifically, the left lane observation noise matrix R l and the right lane observation noise matrix R r is a (m+1)*(m+1) dimensional matrix. The diagonal position of the left lane observation noise matrix is R l (m+1, m+1) is equal to the variance of the lane width w, and the rest are 0; the diagonal position of the right lane observation noise matrix R r (m+1, m+1) is equal to the variance of the lane width w, and the rest of the positions are 0; the noise of this process is set to a constant matrix, and then the a priori estimated covariance matrix is updated according to the Kalman filter formula.
[0070] The lane line obtained from the camera is represented by a cubic curve. Due to the influence of external environment interference or detection fitting error, the lane line curve may have different errors at different distances. For example, the error at the near end may be small while the error at the far end may be large, or vice versa. At this time, if the entire lane line curve is discarded due to the large error of a certain section, some effective information will be lost. Therefore, this method selects several discrete points from the cubic curve as observations. These discrete points are selected from the curve according to certain rules, so there is a conversion relationship between the state vector and the observation vector, that is, this conversion relationship is represented by the observation matrix.
[0071] (S4) The state vector X is calculated in accordance with the Kalman wave formula according to the observations, observation matrix and observation noise matrix of the left lane and the observations, observation matrix and observation noise matrix of the right lane, or the observations, observation matrix and observation noise matrix of the right lane and the observations, observation matrix and observation noise matrix of the left lane. k and the covariance matrix P k Update the state and obtain the state vector X′ of the center line of the lane at time k, which is the final output of the Kalman filter k =[A k′ , A′ k′ , B k′ , B′ k′ , C k′ , D k′ , w k′ ] T, then the centerline curve of the lane at the current time k is: y k =A k′ +B k′ x+C k′ x 2 +D k′ x 3 ; Among them, the covariance matrix P k Updated according to the Kalman filter formula, the initial value can be set to the unit matrix or determined according to the actual application;
[0072] (S5) According to the parallel relationship between the left and right lane lines and the relationship with the center line of the lane, the left lane line and the right lane line are generated. The curve equations of the left lane line and the right lane line at the current time k are:
[0073] y l =A k′ +B k′ x+C k′ x 2 +D k′ x 3 -w / 2;
[0074] y r =A k′ +B k′ x+C k′ x 2 +D k′ x 3 +w / 2.
[0075] In this embodiment, the Kalman filter state update formula is as follows:
[0076]
[0077] X′ k =X k +K k (Z k -H k X k );
[0078] P′ k =(IK k H k ) k ;
[0079] Among them, I is the identity matrix, K k is the Kalman filter gain at time k, X′ k is the state vector at time k output by the Kalman filter, P′ k is the covariance of the Kalman filter output at time k, P k For X k The covariance ofk is the observation matrix at time k, H k The transposed matrix, Z k is the observation vector at time k.
[0080] In this embodiment, the state transfer matrix F is:
[0081] Among them, dt is the time interval from time k-1 to time k, dl represents the distance the vehicle moves in a uniform straight line during time dt, that is, dl=v·dt, v is the speed of the vehicle.
[0082] In this embodiment, the establishment of taking the center position of the front end of the vehicle as the origin is specifically: establishing taking the center position of the front bumper of the vehicle as the origin.
[0083] In this embodiment, after the lane lines are reinitialized, the following steps are also performed: determine whether the lane lines are initialized successfully, if so, save the current frame; otherwise, the current frame processing ends; wherein, the condition for successful initialization is that the lane lines on both sides of the lane exist at the same time and are parallel to each other.
[0084] In this embodiment, the steps for determining whether the lane lines on both sides of the lane are parallel to each other are as follows:
[0085] Select several points with the same x value on the left and right lane curves respectively;
[0086] If the width deviation at each point ||y r -y l |-w|<thresh, the left and right lane lines are considered to be approximately parallel, where thresh is a preset deviation threshold. The discrete point selection method can be to select points at equal intervals or with equal number of points within the length of the lane line.
[0087] In this embodiment, the matching degree of each discrete point of the left lane line is The matching degree of each discrete point of the right lane line The calculation formula is as follows:
[0088]
[0089]
[0090] The average curve of the left lane line in the past n frames is used as the left lane baseline The average curve of the right lane line in the past n frames is used as the right lane baseline n is greater than or equal to 2.
[0091] In this embodiment, in step (S3), the method of selecting discrete points is to select points at equal intervals within the length of the lane line. i The selection method is: range is the detectable range of the lane line.
[0092] This embodiment discloses a lane line processing system based on Kalman filtering, including:
[0093] An image acquisition module, used to obtain lane line information;
[0094] Data processing module, used for lane line processing based on Kalman filtering and generating left and right lane lines;
[0095] The image acquisition module is connected to the data processing module, and the lane line processing system based on Kalman filtering is configured to execute the steps of the above-mentioned lane line processing method based on Kalman filtering.
[0096] In this embodiment, the image acquisition module is a vehicle-mounted camera.
[0097] The lane line processing method and system based on Kalman filtering of the present invention consider the parallel constraint relationship between the left and right lane lines, track the width of the lane, and translate the center line according to the lane width to form parallel lane lines; the center line is modeled to make the lateral control more stable for the system adopting the centering control strategy; the state quantity is the cubic curve coefficient, and the observation quantity is a discrete point, which reduces the step of finally fitting the discrete points into the cubic curve; at the same time, the observation quantity is a discrete point, which can avoid the influence of abnormal points on the fitting result, and the dynamic weight makes the fitting result more stable; it can also avoid the discontinuity between the frames before and after the lane line, and suppress the abnormal jump of the lane width.
[0098] The preferred specific embodiments of the present invention are described in detail above. It should be understood that a person skilled in the art can make many modifications and changes based on the concept of the present invention without creative work. Therefore, any technical solution that can be obtained by a person skilled in the art through logical analysis, reasoning or limited experiments based on the concept of the present invention on the basis of the prior art should be within the scope of protection determined by the claims.
Claims
1. A lane line processing method based on Kalman filtering, characterized in that the steps include: (S1) Obtain lane line information and determine whether the lane line information of the previous frame is valid. If so, proceed to step (S2); otherwise, reinitialize the lane line and the process ends; (S2) Initialize the Kalman filter state vector by taking the center line of the lane in the current frame, and select the cubic curve coefficient, coefficient change rate and lane width w used to characterize the center line of the lane as the state quantity, that is, the state vector is X k =[A k , A′ k , B k , B′ k , C k , D k , w k ] T ; Among them, the state vector X at the current time k k Through the state vector X at the last moment k-1 k-1 It is inferred that X k =F·X k-1 , the state transfer matrix F is established according to the uniform motion model, A k , B k , C k , D k and w k They represent the state quantity at time k predicted based on time k-1 respectively; where the curve equation of the center line of the lane and the curve equations of the left and right lane lines are: y center =A t +B t x+C t x 2 +D t x 3 ; Among them, A t , B t , C t and D t They represent the coefficients of the curve equation of the center line at time t respectively; and They represent the coefficients of the curve equation of the left lane line at time t respectively; and They represent the coefficients of the curve equation of the right lane line at time t respectively; (S3) Obtain the observations, observation matrix and observation noise matrix of the left lane line and the right lane line respectively: The center position of the front end of the vehicle is established as the origin, the vehicle's forward direction is the x-axis, and the direction perpendicular to the vehicle's forward direction is the y-axis. The left lane line and the right lane line of the current frame are both used as the observation lane lines. m discrete points (x i ,y l (x i )) and m discrete points on the right lane line (x i ,y r (x i )); wherein, i=1, 2, 3, ..., m, and m is greater than or equal to 4; y l (x i ) and the lane width w are used as the left lane observation quantity, and the observation vector of the left lane is y r (x i ) and the lane width w are used as the left lane observation, and the observation vector of the right lane is The m discrete points on the left lane line and the m discrete points on the right lane line are both used as observation quantities for state update. The conversion formula from state quantity to observation quantity is: Therefore, the observation matrix H of the left lane is l And the observation matrix H of the right lane r They are: Calculate the matching degree of each discrete point of the left lane line respectively The matching degree of each discrete point of the right lane line The matching degree and Discrete points larger than the preset threshold are eliminated. For discrete points within the preset threshold range, the left lane and right lane observation noise matrices are dynamically adjusted according to the matching degree; (S4) The state vector X is calculated in accordance with the Kalman wave formula according to the observations, observation matrix and observation noise matrix of the left lane and the observations, observation matrix and observation noise matrix of the right lane, or the observations, observation matrix and observation noise matrix of the right lane and the observations, observation matrix and observation noise matrix of the left lane. k and the covariance matrix P k Update the state and obtain the state vector X′ of the center line of the lane at time k, which is the final output of the Kalman filter k =[A k′ , A′ k′ , B k′ , B′ k′ , C k′ , D k′ , w k′ ] T , then the centerline curve of the lane at the current time k is: y k =A k′ +B k′ x+C k′ x 2 +D k′ x 3 ; Among them, the covariance matrix P k Updated according to the Kalman filter formula, the initial value can be set to the unit matrix or determined according to the actual application; (S5) According to the parallel relationship between the left and right lane lines and the relationship with the center line of the lane, the left lane line and the right lane line are generated. The curve equations of the left lane line and the right lane line at the current time k are: y l =A k′ +B k′ x+C k′ x 2 +D k′ x 3 -w / 2; y r =A k′ +B k′ x+C k′ x 2 +D k′ x 3 +w / 2。 2. The lane line processing method based on Kalman filtering according to claim 1, characterized in that: The Kalman filter state update formula is as follows: X′ k =X k +K k (Z k -H k X k ); P′ k =(I-K k H k )P k ; Among them, I is the identity matrix, K k is the Kalman filter gain at time k, X′ k is the state vector at time k output by the Kalman filter, P′ k is the covariance of the Kalman filter output at time k, P k For X k The covariance of k is the observation matrix at time k, H k The transposed matrix, Z k is the observation vector at time k.
3. The lane line processing method based on Kalman filtering according to claim 1 or 2, characterized in that: The state transfer matrix F is: Among them, dt is the time interval from time k-1 to time k, dl represents the distance the vehicle moves in a uniform straight line during time dt, that is, dl=v·dt, v is the speed of the vehicle.
4. The lane line processing method based on Kalman filtering according to claim 3 is characterized in that: The establishment of taking the center position of the front end of the vehicle as the origin is specifically: establishing taking the center position of the front bumper of the vehicle as the origin.
5. The lane line processing method based on Kalman filtering according to claim 1, 2 or 4, characterized in that: After the lane line is reinitialized, the following steps are also performed: determine whether the lane line is initialized successfully, if so, save the current frame; otherwise, end the processing of the current frame; wherein the condition for successful initialization is that the lane lines on both sides of the lane exist at the same time and are parallel to each other.
6. The lane line processing method based on Kalman filtering according to claim 5, characterized in that: The steps for determining whether the lane lines on both sides of the lane are parallel to each other are as follows: Select several points with the same x value on the left and right lane curves respectively; If the width deviation at each point ||y r -y l |-w|<thresh, the left and right lane lines are considered to be approximately parallel, where thresh is the preset deviation threshold.
7. The lane line processing method based on Kalman filtering according to claim 1, 2, 4 or 6, characterized in that: The matching degree of each discrete point of the left lane line The matching degree of each discrete point of the right lane line The calculation formula is as follows: The average curve of the left lane line in the past n frames is used as the left lane baseline The average curve of the right lane line in the past n frames is used as the right lane baseline n is greater than or equal to 2.
8. The lane line processing method based on Kalman filtering according to claim 7, characterized in that: In step (S3), the discrete points are selected in a way that the points are equally spaced within the length of the lane line. i The selection method is: range is the detectable range of the lane line.
9. The lane line processing method based on Kalman filtering according to claim 1, 2, 4, 6 or 8, characterized in that: The left lane observation noise matrix R l and the right lane observation noise matrix R r is a (m+1)*(m+1) dimensional matrix. The diagonal position of the left lane observation noise matrix is R l (m+1, m+1) is equal to the variance of the lane width w, and the rest are 0; the diagonal position of the right lane observation noise matrix R r (m+1, m+1) is equal to the variance of the lane width w, and the rest of the positions are 0.
10. A lane line processing system based on Kalman filtering, characterized in that: include: An image acquisition module, used to obtain lane line information; Data processing module, used for lane line processing based on Kalman filtering and generating left and right lane lines; The image acquisition module is connected to the data processing module, and the lane line processing system based on Kalman filtering is configured to execute the steps of the lane line processing method based on Kalman filtering as described in any one of claims 1 to 9.
