A lane keeping method for autonomous sanitation vehicles under interference from large vehicle side lights

Abstract

The present invention belongs to the field of autonomous driving technology and specifically discloses a method for autonomous sanitation vehicles to maintain lanes under interference from large vehicle sidelights. The method first obtains environmental data such as road images and obstacle targets in front of the vehicle. Highlight features are then extracted from the road images, and their speed relative to the vehicle is calculated using the frame difference method. This is then compared with the speed of the obstacle targets to distinguish lane line features from interference features. When a lane line feature is determined, lane keeping is achieved by averaging the weighted distances of the lane lines on both sides relative to the vehicle's centerline. When interference is determined, a dual-threshold method is used, combined with prior knowledge of lane width, to generate virtual lane lines, and the lane center is determined using comprehensive weighting. The present invention enables autonomous sanitation vehicles to achieve reliable lane keeping at night or in low-light environments at low cost, effectively avoiding interference from large vehicle sidelights, significantly improving driving safety and stability, and possessing the advantages of high precision, strong real-time performance, and adaptability to complex lighting conditions.

Description

Technical Field

[0001] The present invention belongs to the field of autonomous driving technology, and specifically relates to a lane keeping method for an autonomous sanitation vehicle under interference from the side lights of a large vehicle, which is particularly suitable for nighttime or low-light scenes. Background Art

[0002] In modern urban management, the intelligent upgrade of sanitation vehicles is accelerating, and low-speed autonomous driving has now been achieved on urban roads. These autonomous sanitation vehicles typically travel at low speeds in the rightmost lane, relying on cameras and radar mounted on the roof or front of the vehicle to perceive the road environment and plan their own routes.

[0003] However, the actual operation of autonomous sanitation vehicles, especially at night or in low-light environments, poses a significant challenge to their perception systems. Urban road lighting conditions are extremely complex, with the road surface reflecting light from a variety of sources, including streetlights, traffic sign lights, vehicle headlights, and sidelights on large vehicles. Large vehicles, in particular, often use their side and side lights at night to identify their position and illuminate blind spots. Side lights are typically mounted on the sides of large vehicles, illuminating the ground vertically. When turned on, they project a large bright spot on the ground. Due to the high brightness of the side lights and the heavy traffic on urban roads, lane markings are severely worn by constant traffic. Once the side lights shine on the lane markings, the strong reflection from the road surface blurs the lane features, making the lane boundary features difficult to accurately distinguish.

[0004] For the reasons mentioned above, when a large vehicle passes in the left lane of a sanitation vehicle, the vision-based autonomous driving solution is easily disturbed, resulting in the inability to obtain accurate lane features. In this case, the system may mistakenly identify the entire bright spot area, or even other slender features on the large vehicle's body, as a lane. In the image, these highlighted features often appear as irregularly shaped light spots, and there may also be slender strips of higher brightness within the light spots, or multiple slender strips arranged in parallel. Since the operating mechanism of the lane keeping system is to keep the vehicle near the predetermined lane centerline, once these erroneous features are identified as lane lines, the lane keeping system will steer the vehicle closer to the large vehicle in an attempt to reach the preset lane centerline position, which undoubtedly poses a major safety hazard to the driving of sanitation vehicles.

[0005] While LiDAR-based autonomous driving perception methods can accurately detect the specific locations of other vehicles and lane markings on the road and are insensitive to changes in lighting, LiDAR deployment and maintenance costs are extremely high. This makes it a pressing challenge in practical applications to cost-effectively obtain reliable lane centering results and effectively avoid interference from sidelights on the lane-keeping systems of autonomous sanitation vehicles.

[0006] In view of this, this invention is proposed. Summary of the Invention

[0007] The present invention aims to overcome the shortcomings of the aforementioned prior art by providing a lane-keeping method for autonomous sanitation vehicles under interference from large vehicle sidelights. This method is primarily intended to address the difficulties in lane feature recognition, misidentification, and insufficient lane-keeping system stability caused by interference from large vehicle sidelights at night or in low-light environments. By extracting highlight features, combining the frame difference method to calculate movement speed, a dual-threshold method, and the generation of virtual lane lines, the present invention can achieve reliable lane-keeping functionality at a relatively low cost, significantly improving the driving safety and stability of autonomous sanitation vehicles under complex lighting conditions.

[0008] The purpose of the present invention is to solve the problem through the following technical solutions:

[0009] The present invention provides a lane keeping method for an autonomous sanitation vehicle under interference from large vehicle sidelights, the lane keeping method comprising the following steps:

[0010] Step 1: Acquire environmental data in front of the vehicle (in this invention, the vehicle refers to an autonomous sanitation vehicle), wherein the environmental data includes a road image and obstacle targets;

[0011] Step 2: Processing the road image to extract highlight features;

[0012] Step 3: Calculate the moving speed v0 of the highlighted feature relative to the vehicle using the frame difference method;

[0013] Step 4: If there is an obstacle in front of the vehicle, the obstacle's speed v1 relative to the vehicle is automatically calculated. Based on the relationship between the obstacle's speed v1 and the highlighted feature's speed v0 relative to the vehicle, the highlighted feature is determined to be a lane feature.

[0014] If there is no obstacle in front of the vehicle, the highlighted feature is directly determined to be the lane line feature;

[0015] Step 5: When the highlighted feature is determined to be a lane line feature, the distances of the lane lines on both sides relative to the center line of the vehicle are averaged to keep the vehicle in the center of the lane;

[0016] Step 6: When it is determined that the highlighted feature is not a lane feature, a dual-threshold method is used to obtain slender strip features (first, a low-threshold segmentation is used to obtain a mask for the entire highlighted feature area, and then a high-threshold segmentation is used to separate the slender lane features within the highlighted feature area). If the result obtained using the dual-threshold method contains multiple slender strip features arranged in parallel, only the slender strip feature on the right is selected as the output. At the same time, a virtual lane line is formed based on the prior knowledge of the lane line and lane width on the right side of the ego vehicle. The distance between the virtual lane line and the result obtained by the dual-threshold method relative to the ego vehicle centerline is comprehensively weighted to ultimately determine the lane center, which is used to keep the ego vehicle driving in the lane center.

[0017] Furthermore, in step 1, the road image is captured by a vehicle-mounted camera installed on the vehicle, and the camera is a high-definition camera.

[0018] Furthermore, in step 1, the obstacle target is scanned by a radar installed on the vehicle, and the radar is a millimeter wave radar with a scanning range of 0 to 100 meters in front of the vehicle.

[0019] Furthermore, in step 2, the specific process of processing the road image is as follows:

[0020] Step 2.1: Select four points based on the lane lines in the road image (one point symmetrically located in front and behind the left and right lanes). Transform the lane lines that intersect at one point into parallel lane lines using a perspective transformation.

[0021] Step 2.2: Use the Otsu algorithm to automatically calculate the distribution pattern of each pixel in the road image and extract the highlight features in the road image.

[0022] Furthermore, in step 3, the specific calculation process of the moving speed v0 of the highlighted feature relative to the vehicle is:

[0023] Step 3.1, projecting the highlighted features at the positions of two consecutive road image frames into the same road image;

[0024] Step 3.2: Calculate the vertical displacement x of the highlighted feature in pixels. Divide the displacement x by the sampling time to obtain the moving speed v0 of the highlighted feature relative to the vehicle in pixels / second.

[0025] Step 3.3: Convert the unit of the highlighted feature's speed v0 relative to the vehicle to km / h based on the calibration relationship.

[0026] Furthermore, in step 4, when v1-v0>v1*A, and the vehicle speed is greater than the set speed threshold v, it is determined that the highlighted feature in the road image is a lane line feature relative to the vehicle's movement and has nothing to do with the obstacle target;

[0027] When v1-v0≤v1*A, the highlight feature in the road image is determined to be a light spot formed by the side lights of a large vehicle or other slender features on the vehicle body;

[0028] Among them, A is set to 8% to 12%.

[0029] Furthermore, in step 4, the speed threshold v ranges from 20 km / h to 30 km / h.

[0030] Furthermore, the specific process of step 5 is as follows:

[0031] Step 5.1: Consider the centerline of the road image as the centerline of the vehicle, and calculate the distances Xl and Xr of the left and right lane lines relative to the centerline of the vehicle, respectively.

[0032] Step 5.2: Add the distances Xl and Xr of the left and right lane lines relative to the vehicle's centerline and divide by 2. The resulting distance is considered the ideal lane center distance.

[0033] Step 5.3: The lane keeping system controls the vehicle to move along the ideal lane center distance.

[0034] Furthermore, the specific process of step 6 is as follows:

[0035] Step 6.1: Use the Otsu algorithm to automatically calculate the distribution pattern of each pixel in the road image, thereby extracting the highlight features in the road image and obtaining a highlight feature mask;

[0036] Step 6.2: Extract only the highlighted feature portion of the original road image based on the highlighted feature mask, and again use the Otsu algorithm to automatically calculate the distribution pattern of each pixel in the original road image to extract the elongated lane features in the road image;

[0037] Step 6.3: When multiple long and thin strip features are arranged in parallel in the result obtained using the dual threshold method, only the long and thin strip feature on the right is selected as the lane feature output by horizontally comparing the pixel coordinates;

[0038] Step 6.4: Consider the centerline of the road image as the centerline of the vehicle, and calculate the distances Xl and Xr of the left lane line and the right slender strip feature relative to the centerline of the vehicle, respectively.

[0039] Step 6.5: Based on the prior knowledge of lane width, draw a virtual lane line 3.75m away from the left lane line. The calculated distance of the virtual lane line from the centerline of the vehicle is 3.75-X1.

[0040] Step 6.6. Calculate the ideal lane center distance using the following formula:

[0041]

[0042] Where: w1 and w2 are weight coefficients, representing the influence of the right lane line and the virtual lane line on the lane center distance, w1+w2=1, and w1 <w2;

[0043] Step 6.7: The lane keeping system controls the vehicle to move along the ideal lane center distance in step 6.6.

[0044] Furthermore, the lane keeping method is applicable to self-driving sanitation vehicles operating on urban roads at night or in low-light environments, and is used to avoid interference of large vehicle side lights on the lane keeping system.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] The lane keeping method provided by the present invention uses an on-board camera and millimeter-wave radar to acquire environmental data. It then uses the Otsu algorithm to extract highlight features from road images and, combined with the frame difference method, calculates the speed of the highlight features relative to the vehicle. By comparing this speed with the speed of obstacles, it accurately distinguishes lane line features from the light spots produced by large vehicle sidelights or other slender features on the vehicle body. When a highlight feature is identified as a lane line, lane keeping is achieved by averaging the weighted distances of the lane lines on both sides relative to the vehicle's centerline. When a highlight feature is identified as interference, a dual-threshold method is used to extract slender strip features. A virtual lane line is generated based on prior knowledge of lane width. The weighted virtual lane line is then combined with the results obtained by the left dual-threshold method to ultimately determine the lane center position. The lane keeping method provided by the present invention enables autonomous sanitation vehicles to achieve reliable lane keeping at night or in low-light environments at a low cost, effectively avoiding interference from large vehicle sidelights and significantly improving the safety and stability of autonomous sanitation vehicles. It offers the advantages of high precision, strong real-time performance, and adaptability to complex lighting conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The accompanying drawings are incorporated in and constitute a part of this specification and, together with the description, serve to explain the principles of the invention.

[0048] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0049] Figure 1 This is a flow chart of a method for lane keeping for an autonomous sanitation vehicle under interference from large vehicle sidelights according to the present invention;

[0050] Figure 2 This is a schematic diagram of reference point selection for perspective transformation provided by the present invention;

[0051] Figure 3 Schematic diagram of the calculation of the moving speed of the highlight feature relative to the vehicle after perspective transformation of the road image according to the present invention;

[0052] Figure 4 It is a characteristic schematic diagram of the combination feature in the highlight feature of the present invention;

[0053] Figure 5 Schematic diagram of the present invention for calculating the distance between the weighted virtual lane line and the highlighted feature relative to the vehicle centerline. DETAILED DESCRIPTION

[0054] Here, exemplary embodiments are described in detail. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Instead, they are merely examples consistent with some aspects of the present invention as detailed in the appended claims.

[0055] In order to enable those skilled in the art to better understand the technical solutions of the present invention, the present invention is described in further detail below with reference to the accompanying drawings and embodiments.

[0056] See also Figures 1 to 5 The method for keeping lanes for an autonomous sanitation vehicle under interference from large vehicle sidelights provided by the present invention specifically comprises the following steps:

[0057] Step 1: Obtain environmental data in front of the vehicle, including road images and obstacle targets;

[0058] Specifically, the road image is captured by a vehicle-mounted camera, preferably a high-definition camera, mounted on the upper portion of the vehicle's windshield, facing forward. Obstacles are scanned by a radar, preferably a millimeter-wave radar, with a scanning range of 0 to 100 meters in front of the vehicle to determine whether there are any obstacles ahead. The millimeter-wave radar automatically calculates the obstacle's velocity v1 relative to the vehicle and transmits this velocity v1 to the lane keeping system in real time.

[0059] Step 2: Process the road image obtained in step 1 to extract highlight features;

[0060] The specific process is as follows:

[0061] 2.1) Select 4 points based on the lane lines in the road image (such as Figure 2 As shown in the figure, one point is taken symmetrically in front and behind the left and right lanes respectively. The lane lines that intersect at one point are converted into parallel lane lines through perspective transformation (this is used to eliminate the perspective effect, that is, the originally parallel lane lines will intersect at a point in the distance in the road image);

[0062] 2.2) Using the Otsu algorithm, the distribution pattern of each pixel in the road image is automatically calculated to extract the highlight features in the road image.

[0063] The specific calculation process of Otsu algorithm is as follows:

[0064] For a road image, its size is M×N. The grayscale level i of each pixel in the road image belongs to the range of [0, 255]. The number of pixels belonging to grayscale level i is Ni. Assuming that there is an optimal threshold T that can divide the image into two categories: background and target, the ratio of background pixels to the entire image ω0, the ratio of target pixels to the entire image ω1, the background pixel grayscale mean μ0 and the target pixel grayscale mean μ1 can be calculated by the following formula:

[0065]

[0066] Then the inter-class variance σ between background pixels and target pixels can be calculated 2 ,

[0067] σ 2 =ω0*ω1*(μ0-μ1) 2

[0068] When σ 2 When it is the maximum, the segmentation threshold i is considered to be the optimal threshold T.

[0069] Step 3: Calculate the moving speed v0 of the highlighted feature relative to the vehicle using the frame difference method;

[0070] The specific process is as follows:

[0071] 3.1) Projecting the highlighted features at the positions of two consecutive road image frames into the same road image;

[0072] 3.2) Calculate the vertical displacement x of the highlighted feature in pixels. Divide the displacement x by the sampling time to obtain the moving speed v0 of the highlighted feature relative to the vehicle in pixels / second.

[0073] 3.3) According to the calibration relationship (the corresponding relationship between unit pixels and the length in the actual environment), convert the unit of the moving speed v0 of the highlighted feature relative to the host vehicle to km / h.

[0074] Step 4: If there is an obstacle target in front of the host vehicle, calculate the moving speed v1 of the obstacle target relative to the host vehicle, and based on the relationship between the moving speed v1 of the obstacle target relative to the host vehicle and the moving speed v0 of the highlighted feature relative to the host vehicle, determine whether the highlighted feature is a lane line feature;

[0075] Specifically, when v1 - v0 > v1 * A (A is generally set to 8% - 12%), and at this time the speed of the host vehicle is greater than the set speed threshold v (this speed threshold is the normal operating speed of the autonomous sanitation vehicle. Only when the speed of the host vehicle is greater than or equal to v, it is determined that the autonomous sanitation vehicle is operating normally. When it is less than v, it is determined that the autonomous sanitation vehicle is parked by the roadside or decelerating, and at this time, deviation does not need to be considered. Generally, the speed threshold v is set to 20 km / h - 30 km / h), it is determined that the highlighted feature in the road image is a lane line feature moving relative to the host vehicle and has nothing to do with the obstacle target;

[0076] When v1 - v0 < v1 * A, it is determined that the highlighted feature in the road image is a light spot formed by the side lights of a large vehicle or other slender shape features on the vehicle body. At this time, the highlighted feature should have an uneven shape or a pattern of multiple slender strip features arranged in parallel;

[0077] If there is no obstacle target in front of the host vehicle, directly determine that the highlighted feature is a lane line feature.

[0078] Step 5: When it is determined that the highlighted feature is a lane line feature, calculate the average weighted distance of the lane lines on both sides relative to the center line of the host vehicle to keep the host vehicle driving in the center of the lane;

[0079] The specific process is as follows:

[0080] 5.1) Regard the center line of the road image as the center line of the host vehicle, and calculate the distances Xl and Xr of the lane lines on the left and right sides relative to the center line of the host vehicle respectively;

[0081] 5.2) Add the distances Xl and Xr of the lane lines on the left and right sides relative to the center line of the host vehicle and divide by 2, and regard the obtained distance as the ideal lane center distance;

[0082] 5.3) The lane keeping system controls the vehicle to move along the ideal lane center distance

[0083] Step 6: When it is determined that the highlighted feature is not a lane line feature, the double-threshold method is used to obtain slender strip features (first, use a low threshold to segment to obtain the entire highlighted feature area mask, and then use a high threshold to segment the slender lane features in the highlighted feature area). When there are multiple slender strip features arranged in parallel in the obtained result, only the slender strip feature on the right is selected as the output. At the same time, according to the prior knowledge of the lane line on the right side of the vehicle and the lane width, a virtual lane line is formed, and the distances of the virtual lane line and the result obtained by the double-threshold method on the left relative to the center line of the vehicle are comprehensively weighted to finally determine the lane center for keeping the vehicle driving in the center of the lane.

[0084] The specific process is as follows:

[0085] 6.1) Use the Otsu algorithm to automatically calculate the distribution law of each pixel in the road image, so as to extract the highlighted features in the road image and obtain the highlighted feature mask;

[0086] 6.2) Only extract the original road image of the highlighted feature part according to the highlighted feature mask, and use the Otsu algorithm again for the original road image to automatically calculate the distribution law of each pixel therein, so as to extract the slender lane features in the road image;

[0087] 6.3) When there are multiple slender strip features arranged in parallel in the result obtained by using the double-threshold method, only select the slender strip feature on the right as the lane feature output by horizontally comparing the pixel coordinates;

[0088] 6.4) Regard the center line of the road image as the center line of the vehicle, and calculate the distances Xl and Xr of the left lane line and the right slender strip feature relative to the center line of the vehicle respectively;

[0089] 6.5) According to the prior knowledge of the lane width (the lane width of urban roads is usually 3.5 meters to 3.75 meters. In this embodiment, 3.75 meters is taken as an example), draw a virtual lane line at a position 3.75 m away from the left lane line, and calculate that the distance of the virtual lane line relative to the center line of the vehicle is 3.75 - Xl;

[0090] 6.6) Calculate the ideal lane center distance according to the following formula:

[0091]

[0092] In the formula: w1 and w2 are weight coefficients, which respectively represent the influence degrees of the right lane line and the virtual lane line on the lane center distance, w1 + w2 = 1, and w1 < w2 (for example, w1 takes a value of 0.1 to 0.4, and w2 takes a corresponding value of 0.6 to 0.9);

[0093] 6.7) The lane keeping system controls the vehicle to move along the ideal lane center distance in Step 6.6.

[0094] The foregoing description is intended only to provide specific embodiments of the present invention, which will enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention.

[0095] It should be understood that the present invention is not limited to the above description and that various modifications and changes may be made without departing from the scope thereof. The scope of the present invention is limited only by the appended claims.

Claims

1. A method for lane keeping of an autonomous sanitation vehicle under interference from large vehicle side lights, characterized in that: The lane keeping method comprises the following steps: Step 1: Acquire environmental data in front of the vehicle, including road images and obstacle targets; Step 2: Processing the road image to extract highlight features; Step 3: Calculate the moving speed v0 of the highlighted feature relative to the vehicle using the frame difference method; Step 4: If there is an obstacle in front of the vehicle, the obstacle's speed v1 relative to the vehicle is automatically calculated. Based on the relationship between the obstacle's speed v1 and the highlighted feature's speed v0 relative to the vehicle, the highlighted feature is determined to be a lane feature. If there is no obstacle in front of the vehicle, the highlighted feature is directly determined to be the lane line feature; In step 4, when , and at this time the vehicle speed is greater than the set speed threshold v, then it is determined that the highlighted feature in the road image is the lane line feature relative to the vehicle's movement and has nothing to do with the obstacle target; when When , it is determined that the highlight feature in the road image is a light spot formed by the side lights of a large vehicle or other thin and long strip features on the vehicle body; Among them, A is set to 8% to 12%; Step 5: If the highlighted feature is determined to be a lane line feature, the lane center is finally determined by averaging the weighted distances of the lane lines on both sides relative to the vehicle centerline to keep the vehicle in the lane center. Step 6: When it is determined that the highlighted feature is not a lane feature, a dual-threshold method is used to obtain a slender strip feature. Specifically, a low-threshold segmentation method is first used to obtain a mask for the entire highlighted feature area. Then, a high-threshold segmentation method is used to segment the slender strip features within the highlighted feature area. When the result obtained by the dual-threshold method contains multiple slender strip features arranged in parallel, only the slender strip feature on the right is selected as the output. At the same time, a virtual lane line is formed based on the prior knowledge of the left lane line and lane width of the ego vehicle. The distances of the left lane line, the virtual lane line, and the right lane line obtained by the dual-threshold method relative to the ego vehicle centerline are comprehensively weighted to ultimately determine the lane center, which is used to keep the ego vehicle driving in the lane center. The specific process of step 6 is as follows: Step 6.1: Use the Otsu algorithm to automatically calculate the distribution pattern of each pixel in the road image, thereby extracting the highlight features in the road image and obtaining a highlight feature mask; Step 6.2: extracting only the highlighted feature portion of the original road image based on the highlighted feature mask, and again using the Otsu algorithm to automatically calculate the distribution pattern of each pixel in the original road image to extract the elongated strip features in the road image; Step 6.3: When multiple long and thin strip features are arranged in parallel in the result obtained using the dual threshold method, only the long and thin strip feature on the right is selected as the lane feature output by horizontally comparing the pixel coordinates; Step 6.4: Consider the centerline of the road image as the centerline of the vehicle, and calculate the distances Xl and Xr of the left lane line and the right elongated strip feature relative to the centerline of the vehicle, respectively. Step 6.5: Based on the prior knowledge of lane width, draw a virtual lane line 3.75m away from the left lane line. The calculated distance of the virtual lane line from the centerline of the vehicle is 3.75-X1. Step 6.

6. Calculate the ideal lane center distance using the following formula: Where: w1 and w2 are weight coefficients, representing the influence of the slender strip feature on the right and the virtual lane line on the lane center distance, w1+w2=1, and w1 <w2; Step 6.7: The lane keeping system controls the vehicle to move along the ideal lane center distance described in step 6.

6.

2. The lane keeping method for an autonomous sanitation vehicle under interference from large vehicle side lights according to claim 1 is characterized in that: In step 1, the road image is captured by a vehicle-mounted camera installed on the vehicle, and the camera is a high-definition camera.

3. The lane keeping method for an autonomous sanitation vehicle under interference from large vehicle side lights according to claim 1 is characterized in that: In step 1, the obstacle target is scanned by a radar installed on the vehicle. The radar is a millimeter wave radar with a scanning range of 0 to 100 meters in front of the vehicle.

4. The lane keeping method for an autonomous sanitation vehicle under interference from large vehicle side lights according to claim 1 is characterized in that: In step 2, the specific process of processing the road image is as follows: Step 2.1: Select four points based on the lane lines in the road image, and transform the lane lines that intersect at one point into parallel lane lines through perspective transformation; Step 2.2: Use the Otsu algorithm to automatically calculate the distribution pattern of each pixel in the road image and extract the highlight features in the road image.

5. The lane keeping method for an autonomous sanitation vehicle under interference from large vehicle side lights according to claim 1 is characterized in that: In step 3, the specific calculation process of the moving speed v0 of the highlighted feature relative to the vehicle is: Step 3.1, projecting the highlighted features at the positions of two consecutive road image frames into the same road image; Step 3.2: Calculate the vertical displacement x of the highlighted feature in pixels. Divide the displacement x by the sampling time to obtain the moving speed v0 of the highlighted feature relative to the vehicle in pixels / second. Step 3.3: Convert the unit of the highlighted feature's speed v0 relative to the vehicle to km / h based on the calibration relationship.

6. The lane keeping method for an autonomous sanitation vehicle under interference from large vehicle side lights according to claim 1 is characterized in that: In step 4, the speed threshold v ranges from 20 km / h to 30 km / h.

7. The lane keeping method for an autonomous sanitation vehicle under interference from large vehicle side lights according to claim 1 is characterized in that: The specific process of step 5 is as follows: Step 5.1: Consider the centerline of the road image as the centerline of the vehicle, and calculate the distances Xl and Xr of the left and right lane lines relative to the centerline of the vehicle, respectively. Step 5.2: Add the distances Xl and Xr of the left and right lane lines relative to the vehicle's centerline and divide by 2. The resulting distance is considered the ideal lane center distance. Step 5.3: The lane keeping system controls the vehicle to move along the ideal lane center distance.

8. The method for lane keeping of an autonomous sanitation vehicle under interference from large vehicle side lights according to any one of claims 1 to 7, characterized in that: The lane keeping method is applicable to autonomous sanitation vehicles operating on urban roads at night or in dim light environments, and is used to prevent large vehicle side lights from interfering with the lane keeping system of the autonomous sanitation vehicle.

Images