A visual intelligent navigation and perception method and system for automated line marking vehicles

Abstract

This invention discloses a visual intelligent navigation perception method and system for automated lane marking vehicles, relating to the field of lane marking vehicle navigation technology. The method includes the following steps: installing a binocular camera at the front end of the automated lane marking vehicle to acquire a traveling visual image and a dense depth map; performing image preprocessing on the traveling visual image to obtain a preprocessed traveling image; extracting lateral features of the road based on the preprocessed traveling image, and identifying the road boundaries on both sides of the road through these lateral features; calculating the vehicle's position based on the road boundaries and the dense depth map, and planning a driving path for the automated lane marking vehicle based on the positional relationship between the vehicle's position and the road boundaries. This invention addresses the problem that existing lane marking vehicle navigation technologies still rely on waterlines or maintain a relative distance from guardrails to plan paths, leading to the vehicle easily losing its target or becoming dependent on guardrails and unable to adaptively adjust to changes in road width.

Description

Technical Field

[0001] This invention relates to the field of navigation technology for lane marking vehicles, specifically to a visual intelligent navigation perception method and system for automated lane marking vehicles. Background Technology

[0002] Lane marking vehicle navigation technology refers to the technology that comprehensively utilizes technologies such as perception, positioning, planning and control to accurately navigate the driving path of an automated lane marking vehicle in the absence of lane lines as a reference, ensuring that the lane lines painted by the automated lane marking vehicle meet the standards.

[0003] Existing road marking vehicle navigation technologies typically plan routes by following water lines. However, water lines are prone to wear, shadows, and dirt, causing vehicles to easily lose their target. Furthermore, existing technologies often rely on maintaining a relative distance from roadside guardrails, which is unusable in areas without guardrails and cannot adapt to changes in road width. For example, patent application CN114808649A discloses a "road marking method based on vision system control," which uses roadside guardrails as references to control the relative position of the marking vehicle to the highway's concrete guardrails. However, highway guardrails are diverse, and there are no guardrails in tunnels, making them unrecognizable. Additionally, it cannot adapt to changes in highway width. Existing road marking vehicle navigation technologies suffer from problems such as relying on water lines or maintaining a relative distance from guardrails for route planning, leading to vehicle loss of target, dependence on guardrails, and an inability to adapt to changes in road width. Summary of the Invention

[0004] This invention aims to at least partially solve one of the technical problems in the prior art. By installing a binocular camera at the front end of an automated road marking vehicle, it acquires a travel visual image and a dense depth image of the vehicle. Then, it preprocesses the travel visual image to obtain a travel preprocessed image. Based on the grayscale values ​​of each row of pixels in the travel preprocessed image, it extracts the lateral features of the road. Then, it identifies the road boundaries on both sides of the road by the differences in the lateral features of the road boundaries. Then, it calculates the vehicle position of the automated road marking vehicle based on the depth values ​​of the left and right boundary points in the dense depth image. Finally, it plans a driving path for the automated road marking vehicle based on the positional relationship between the vehicle position and the road boundaries. This solves the problem that existing road marking vehicle navigation technologies still use water lines or maintain a relative distance from railings to plan paths, which makes the road marking vehicle prone to losing its target or relying on railings and unable to adaptively adjust to changes in road width.

[0005] To achieve the above objectives, in a first aspect, this application provides a visual intelligent navigation perception method for automated line marking vehicles, comprising the following steps:

[0006] A binocular camera is installed at the front of the automated line marking vehicle to capture the vehicle's visual image and density depth map as it travels.

[0007] The moving visual image is preprocessed to obtain a moving preprocessed image;

[0008] Based on the preprocessed images of the road, the lateral features of the road are extracted, and the road boundaries on both sides of the road are identified through the lateral features.

[0009] The position of the automated lane marking vehicle is calculated based on the road boundary and dense depth map, and the driving path of the automated lane marking vehicle is planned based on the positional relationship between the vehicle position and the road boundary.

[0010] Furthermore, installing a binocular camera at the front of the automated line marking vehicle to acquire its traveling visual image and density depth map includes the following sub-steps:

[0011] A binocular camera is installed at the front end of the automatic line marking vehicle. The binocular camera needs to be installed on the central axis of the automatic line marking vehicle, and the height above the ground should be maintained at the first height.

[0012] During the movement of the automated line marking vehicle, a moving visual image and a dense depth image are collected. The moving visual image is an RGB image obtained by direct shooting.

[0013] Furthermore, the image preprocessing of the moving visual image specifically involves denoising, image enhancement, and grayscale conversion of the moving visual image to obtain the final moving preprocessed image.

[0014] Furthermore, based on the preprocessed images of the road, the lateral features of the road are extracted, and the road boundaries on both sides are identified through these lateral features, including the following sub-steps:

[0015] The lateral features of the road are extracted based on the gray values ​​of each row of pixels in the preprocessed image.

[0016] Identify the road boundaries on both sides of the road by recognizing the differences in the lateral features of the road boundaries.

[0017] Furthermore, extracting the lateral features of the road based on the grayscale values ​​of each row of pixels in the preprocessed image includes the following sub-steps:

[0018] The pixels in the preprocessed image are numbered from bottom to top, denoted by the symbol n, where n is the row number. The pixels in the preprocessed image are numbered from left to right, denoted by the symbol m, where m is the column number. The pixel in the nth row and mth column of the preprocessed image is marked as PL(n,m), where n and m are both non-zero natural numbers and (n,m) is the index of PL. The gray value of PL(n,m) is marked as PG(n,m).

[0019] Analyze each row independently and calculate [PG(n+1,m)+PG(n,m)+PG(n-1,m)] / 3 to obtain the sliding grayscale mean of all PL(n,m) in the nth row, represented by the symbol SG(n,m). Obtain the maximum value of m, which is max(m), where max() is the maximum value operator. Name max(m) / 2 as the center column value, represented by the symbol CC.

[0020] Sort and number SG(n,m) whose m is less than CC in descending order, using the symbol SA. i This indicates that SG(n,m) with m greater than CC are sorted and numbered in ascending order of m, using the symbol SB. i In this representation, i is a non-zero natural number and i is the index of SA and SB, where SA i and SB i This refers to the horizontal feature.

[0021] Furthermore, identifying the road boundaries on both sides of the road by the differences in lateral features includes the following sub-steps:

[0022] Cluster analysis is performed on the horizontal features, which are then grouped into different gray-level clusters, starting with i=1, to determine the SA. i with SA i+1 Whether they belong to the same gray-level cluster group; if so, then SA i with SA i+1 They are grouped into a left homogeneous set, and SB is determined. i With SB i+1 Whether they belong to the same gray-level cluster group; if so, then SB i With SB i+1 Summarize them into a right homogeneous group, add one to i and judge again until max(i) is reached. Analyze and obtain different left homogeneous groups and right homogeneous groups, which are collectively called feature homogeneous groups. Merge feature homogeneous groups that have intersection.

[0023] Sort and number the left and right homogeneous groups in ascending order of i, and use the symbol LT respectively. h and RT gThis indicates that h and g are both non-zero natural numbers, with h being the index of LT and g being the index of RT;

[0024] Name LT1 the left baseline group, and name the gray-level cluster group to which the left baseline group belongs the left baseline cluster, starting with h=2. If LT h LT h+1 and LT h+2 If none of them are in the left-benchmark cluster, then LT will be... h The smallest SA in i i Mark it as the left boundary point; otherwise, increment h and re-evaluate.

[0025] Name RT1 the right baseline group, and name the gray-level cluster group to which the right baseline group belongs the right baseline cluster, starting with g=2. If RT g RT g+1 and RT g+2 If none of them are in the right-benchmark cluster, then RT g The smallest SB in i i Mark it as the right boundary point; otherwise, increment g and re-evaluate.

[0026] Starting with n=2, extract the left and right boundary points of each row. If no left or right boundary point is found, increment n and extract again until the left and right boundary points are obtained. The left and right boundary points are the road boundaries.

[0027] Furthermore, the calculation of the vehicle's position based on the road boundary and dense depth map, and the planning of the driving path for the automated lane marking vehicle based on the positional relationship between the vehicle's position and the road boundary, includes the following sub-steps:

[0028] The position of the vehicle body of the automatic line marking vehicle is calculated based on the depth values ​​of the left and right boundary points in the dense depth map.

[0029] The system plans the driving path for the automated lane marking vehicle by analyzing the positional relationship between the vehicle body and the road boundary.

[0030] Furthermore, calculating the vehicle body position of the automated line-marking vehicle based on the depth values ​​of the left and right boundary points in the dense depth map includes the following sub-steps:

[0031] The depth value corresponding to the left boundary point in the dense depth map is named the left front boundary depth, denoted by the symbol LL, and the depth value corresponding to the right boundary point in the dense depth map is named the right front boundary depth, denoted by the symbol RL. The depth value represents the distance between the pixel and the stereo camera.

[0032] Name the PL(n,CC) in the same row as the left and right boundary points as the simulation points, where PL(n,CC) represents PL(n,m) for m=CC. Obtain the depth value of the simulation points in the dense depth map and name it as the simulation depth, represented by the symbol ML.

[0033] Connect PL(1,CC) to the simulation point, the left boundary point, and the right boundary point respectively to obtain the center line, the left line, and the right line. Name the acute angle between the center line and the left line as the left auxiliary angle, and represent it with the symbol KL. Name the acute angle between the center line and the right line as the right auxiliary angle, and represent it with the symbol KR.

[0034] Solve the system of equations Obtain the left spacing of the automatic line marking vehicle, where LD is the left spacing, and solve the system of equations. The right spacing of the automatic line marking vehicle is obtained, where RD is the right spacing, and the left spacing and right spacing together constitute the vehicle body position.

[0035] Furthermore, planning the driving path for the automated lane-marking vehicle based on the positional relationship between the vehicle body and the road boundary includes the following sub-steps:

[0036] When put into use, control the automatic line marking vehicle to reach the line marking starting point, and ensure that the road width of the first starting distance in front of the line marking starting point remains unchanged and that it is a straight road;

[0037] The LD and RD of the automatic lane marking vehicle are detected at this time and named as the left starting distance and right starting distance, respectively, and represented by the symbols LSD and RSD. The starting road width is obtained by calculating LSD+RSD.

[0038] The row number of the left and right boundary points obtained from the analysis is marked as E. Starting from n=E+1, the left and right boundary points of the nth row are analyzed. At the same time, PL(n,CC) is named the predicted travel point. The LD and RD of the predicted travel point are calculated and named as the left prediction distance and the right prediction distance, respectively, and represented by the symbols LYD and RYD.

[0039] Select any PL(n,m) in the nth row and name it as the navigation step point. Calculate the LD and RD of the navigation step point and name them as the left navigation distance and right navigation distance, respectively, and represent them by the symbols LFD and RFD.

[0040] Calculate LYD+RYD to obtain the predicted road width. Determine whether the predicted road width is equal to the initial road width. If yes, output a signal indicating equal road widths; otherwise, output a signal indicating unequal road widths.

[0041] If the output path widths are equal, the selected navigation travel point must satisfy LFD=LSD and RFD=RSD;

[0042] If the output path widths are unequal, calculate |LYD-LSD| and |RYD-RSD|, and name the calculation results as left channel offset and right channel offset, respectively, and represent them by the symbols LU and RU.

[0043] If LU≤RU, the selected navigation step point must satisfy LFD=LSD; if LU>RU, the selected navigation step point must satisfy RFD=RSD.

[0044] Starting from n=E+1, the navigation travel points are analyzed. Then, n is incremented and the analysis of navigation travel points is repeated. All navigation travel points are connected by a smooth curve to obtain the driving path. The path is updated in real time and the automatic lane marking vehicle is controlled to travel according to the path.

[0045] Secondly, this application provides a visual intelligent navigation and perception system for automatic lane marking vehicles, including a visual image acquisition module, an image preprocessing module, a road boundary recognition module, and a driving path planning module; the visual image acquisition module, the image preprocessing module, and the driving path planning module are respectively connected to the road boundary recognition module for data transmission.

[0046] The visual image acquisition module is used to install a binocular camera at the front end of the automatic line marking vehicle to acquire the moving visual image and dense depth image of the automatic line marking vehicle.

[0047] The image preprocessing module is used to preprocess the moving visual image to obtain a moving preprocessed image.

[0048] The road boundary recognition module is used to extract the lateral features of the road based on the preprocessed image and to identify the road boundaries on both sides of the road through the lateral features.

[0049] The driving path planning module is used to calculate the vehicle position of the automatic lane marking vehicle based on the road boundary and dense depth map, and to plan the driving path of the automatic lane marking vehicle based on the positional relationship between the vehicle position and the road boundary.

[0050] The beneficial effects of this invention are as follows: By installing a binocular camera at the front end of the automatic lane marking vehicle, the vehicle's travel visual image and dense depth image are acquired. Then, the travel visual image is preprocessed to obtain a travel preprocessed image. Based on the grayscale value of each row of pixels in the travel preprocessed image, the lateral features of the road are extracted. Then, the differences in the lateral features of the road boundary are used to identify the road boundaries on both sides. The advantage is that, regardless of the type of road, there are boundaries on both sides. Within the road, the lateral features have the same commonality. However, once the road boundary is reached, this commonality is broken, thereby identifying the road boundary without relying on fixed reference objects such as railings, which improves the accuracy and effectiveness of the lane marking vehicle navigation.

[0051] This invention calculates the position of an automated lane marking vehicle based on the depth values ​​of the left and right boundary points in a dense depth map. Finally, it plans a driving path for the automated lane marking vehicle based on the positional relationship between the vehicle and the road boundary. The advantage is that when planning the driving path, the reference road boundary or the relative distance to the road boundary is changed in real time based on changes in road width, thereby providing the most accurate and intelligent navigation for the automated lane marking vehicle, further improving the accuracy and effectiveness of the lane marking vehicle navigation. Attached Figure Description

[0052] Figure 1 This is a schematic diagram of the system of the present invention;

[0053] Figure 2 This is a schematic diagram of the center distance line, left distance line, and right distance line of the present invention.

[0054] Figure 3 This is a schematic diagram showing the positional relationship between the center line, left line, and right line of the present invention and the automatic line marking vehicle;

[0055] Figure 4 This is a flowchart of the steps of the method of the present invention. Detailed Implementation

[0056] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0057] Example 1, please refer to Figure 1 As shown, this application provides a visual intelligent navigation and perception system for automatic lane marking vehicles, including a visual image acquisition module, an image preprocessing module, a road boundary recognition module, and a driving path planning module; the visual image acquisition module, the image preprocessing module, and the driving path planning module are respectively connected to the road boundary recognition module for data transmission.

[0058] The visual image acquisition module is used to install a binocular camera at the front end of the automatic line marking vehicle to acquire the moving visual image and dense depth image of the automatic line marking vehicle.

[0059] The visual image acquisition module is configured with a visual image acquisition strategy, which includes:

[0060] A binocular camera is installed at the front of the automatic line marking vehicle. The binocular camera must be installed on the center line of the automatic line marking vehicle and kept at the first height above the ground.

[0061] During the movement of the automated line marking vehicle, a moving visual image and a dense depth image are collected. The moving visual image is an RGB image obtained by direct shooting.

[0062] In practical applications, the centerline is the midpoint where the two front wheels are connected, thus ensuring that the distance between the binocular camera and both sides of the automatic line marking vehicle is equal. The first height can be adjusted arbitrarily. In this embodiment, the first height is set to 1m, that is, the binocular camera is installed at a height of 1m above the ground at the center of the front end of the automatic line marking vehicle. The traveling vision map is a normal RGB image, and the dense depth map is an image in which a depth value is assigned to each pixel. It is usually used for tasks such as 3D reconstruction, object recognition, and scene understanding. Both types of images can be captured by the binocular camera and do not require special processing in this embodiment, so they are not described in this embodiment.

[0063] The image preprocessing module is used to preprocess the moving visual image to obtain a moving preprocessed image; specifically, it performs noise reduction, image enhancement and grayscale processing on the moving visual image to finally obtain the moving preprocessed image.

[0064] In practical applications, denoising, image enhancement, and grayscale conversion are all commonly used preprocessing techniques in the field of image processing, so they will not be described in detail in this embodiment.

[0065] The road boundary recognition module is used to extract the lateral features of the road based on the preprocessed image and to identify the road boundaries on both sides of the road through the lateral features. The road boundary recognition module includes a lateral feature extraction unit and a road boundary recognition unit.

[0066] The lateral feature extraction unit is used to extract the lateral features of the road based on the gray values ​​of each row of pixels in the preprocessed image.

[0067] The lateral feature extraction unit is configured with a lateral feature extraction strategy, which includes:

[0068] The pixels in the preprocessed image are numbered from bottom to top, denoted by the symbol n, where n is the row number. The pixels in the preprocessed image are numbered from left to right, denoted by the symbol m, where m is the column number. The pixel in the nth row and mth column of the preprocessed image is labeled as PL(n,m), where n and m are both non-zero natural numbers and (n,m) is the index of PL. The gray value of PL(n,m) is labeled as PG(n,m).

[0069] Analyze each row independently and calculate [PG(n+1,m)+PG(n,m)+PG(n-1,m)] / 3 to obtain the sliding grayscale mean of all PL(n,m) in the nth row, represented by the symbol SG(n,m). Obtain the maximum value of m, which is max(m), where max() is the maximum value operator. Name max(m) / 2 as the center column value, represented by the symbol CC.

[0070] Sort and number SG(n,m) whose m is less than CC in descending order, using the symbol SA. i This indicates that SG(n,m) with m greater than CC are sorted and numbered in ascending order of m, using the symbol SB. i Let $\frac{ ... i and SB i This refers to the horizontal feature;

[0071] In practical applications, since the automatic lane marking vehicle is used on newly constructed roads, which are not yet subject to wear and tear from daily use or environmental erosion, their surface colors are very similar, making it possible to extract lateral features. In the captured images, the bottom row is closest to the automatic lane marking vehicle, while the top row is furthest away. Therefore, the rows are numbered from bottom to top for easier subsequent analysis. Different PL(n,m) and PG(n,m) are obtained through numbering, where 1≤n≤760 and 1≤m≤980. Each row is analyzed independently; for example, the pixel in the 10th row and 13th column is analyzed. Analyzing PL(10,13), we obtain PG(10,13) as 89. PG(10+1,13) is PG(11,13), specifically 91, and PG(10-1,13) is PG(9,13), specifically 88. Calculating [PG(11,13)+PG(10,13)+PG(9,13)] / 3 yields SG(10,13) as 89. The result is rounded to the nearest integer. This is to perform a moving average on PG(n,m) to reduce the difference between similar pixels. max(m) is 980. The calculated center column value CC is 490, and the numbering gives SA. i and SB i 1≤i≤490, to obtain the horizontal features. It should be noted that if the calculated CC is an integer, then SG(n,m) with m less than or equal to CC need to be sorted and numbered in descending order of m, rather than less than.

[0072] The road boundary recognition unit is used to identify the road boundaries on both sides of the road by the differences in the lateral features of the road boundaries.

[0073] The road boundary recognition unit is configured with a road boundary recognition strategy, which includes:

[0074] Cluster analysis is performed on the horizontal features, which are then grouped into different gray-level clusters, starting with i=1, to determine the SA. i with SA i+1 Whether they belong to the same gray-level cluster group; if so, then SA i with SA i+1 They are grouped into a left homogeneous set, and SB is determined. i With SB i+1 Whether they belong to the same gray-level cluster group; if so, then SB i With SB i+1 Summarize them into a right homogeneous group, add one to i and judge again until max(i) is reached. Analyze and obtain different left homogeneous groups and right homogeneous groups, which are collectively called feature homogeneous groups. Merge feature homogeneous groups that have intersection.

[0075] Sort and number the left and right homogeneous groups in ascending order of i, and use the symbol LT respectively. h and RT g This indicates that h and g are both non-zero natural numbers, with h being the index of LT and g being the index of RT;

[0076] In practical applications, since automatic lane marking vehicles are used on newly constructed roads that have not yet experienced wear and tear or environmental erosion, the grayscale values ​​of their surfaces are very similar. Cluster analysis can group similar grayscale values ​​into the same grayscale cluster. Pixels within the same grayscale cluster are generally considered to be of the same material. For example, pixels corresponding to asphalt pavement belong to the same grayscale cluster. Lane lines are white, showing a significant color difference from the pavement; therefore, white lane lines will be grouped into one grayscale cluster. Road edges often have different types of railings or greenery, resulting in multiple grayscale clusters for these pixels. A feature homogeneous group is essentially a set of consecutive, adjacent pixels all belonging to the same grayscale cluster. Since there are no lane lines directly in front of the automatic lane marking vehicle, and if lane lines exist, they must be on either side of the road, the left and right homogeneous groups are sorted and numbered in ascending order of i to obtain LT. h and RT g .

[0077] Name LT1 the left baseline group, and name the gray-level cluster group to which the left baseline group belongs the left baseline cluster, starting with h=2. If LT h LT h+1 and LT h+2 If none of them are in the left-benchmark cluster, then LT will be... hThe smallest SA in i i Mark it as the left boundary point; otherwise, increment h and re-evaluate.

[0078] Name RT1 the right baseline group, and name the gray-level cluster group to which the right baseline group belongs the right baseline cluster, starting with g=2. If RT g RT g+1 and RT g+2 If none of them are in the right-benchmark cluster, then RT g The smallest SB in i i Mark it as the right boundary point; otherwise, increment g and re-evaluate.

[0079] Starting with n=2, extract the left and right boundary points of each row. If no left or right boundary point is found, add one to n and extract again until the left and right boundary points are found. The left and right boundary points are the road boundaries.

[0080] In practical applications, the smaller the value of i, the closer it is to the lane where the automated lane marking vehicle is located. That is, the smaller the values ​​of h and g, the closer it is to the lane where the automated lane marking vehicle is located. Therefore, LT1 and RT1 must be feature homogeneous groups composed of pixels on the road surface. In other words, the gray-level clustering groups to which LT1 and RT1 belong must be the gray-level clustering groups corresponding to the road surface. This yields the left reference group and the right reference group. h For example, the analysis yielded LT1 to LT 14 Starting with h=2, among LT2, LT3, and LT4, LT3 is in the left reference group. h is incremented and the judgment is recalculated. The principle is as follows: Since LT1 belongs to the left reference group, LT2 must not belong to the left reference group. If LT2 is a lane line, then LT3 or LT4 must contain a feature homogeneous group belonging to the left reference group. Taking three consecutive feature homogeneous groups is to prevent accidental factors from affecting the judgment result. If LT2 is not a lane line, then LT3 and LT4 must not belong to the left reference group, because the color values ​​of the road boundary are very rich due to factors such as railings and greenery, making it difficult to be the same as the left reference group. The left boundary point is identified using this principle. The identification of the right boundary point is similar, therefore, it will not be described in detail in this embodiment. Since the shooting range of the image presents a funnel-shaped diffusion, the lower the number of rows in the image, the closer it is to the automatic lane marking vehicle, and the smaller the shooting range, making it difficult to capture the complete road and road boundary. Therefore, when the road boundary cannot be analyzed, the next row is analyzed.

[0081] The driving path planning module is used to calculate the vehicle position of the automatic lane marking vehicle based on the road boundary and dense depth map, and to plan the driving path of the automatic lane marking vehicle based on the positional relationship between the vehicle position and the road boundary; the driving path planning module includes a vehicle position calculation unit and a driving path planning unit;

[0082] The vehicle body position calculation unit is used to calculate the vehicle body position of the automatically marked vehicle based on the depth values ​​of the left and right boundary points in the dense depth map;

[0083] The vehicle position calculation unit is equipped with a vehicle position calculation strategy, which includes:

[0084] The depth value corresponding to the left boundary point in the dense depth map is named the left front boundary depth, denoted by the symbol LL. The depth value corresponding to the right boundary point in the dense depth map is named the right front boundary depth, denoted by the symbol RL. The depth value represents the distance between the pixel and the stereo camera.

[0085] Name the PL(n,CC) in the same row as the left and right boundary points as the simulation points, where PL(n,CC) represents PL(n,m) for m=CC. Obtain the depth value of the simulation points in the dense depth map and name it as the simulation depth, represented by the symbol ML.

[0086] Please see Figures 2 to 3 As shown, PL(1,CC) is connected to the simulation point, the left boundary point and the right boundary point respectively to obtain the center distance line, the left distance line and the right distance line. The acute angle between the center distance line and the left distance line is named the left auxiliary angle and represented by the symbol KL. The acute angle between the center distance line and the right distance line is named the right auxiliary angle and represented by the symbol KR.

[0087] Solve the system of equations Obtain the left spacing of the automatic line marking vehicle, where LD is the left spacing, and solve the system of equations. The right spacing of the automatically marked line vehicle is obtained, where RD is the right spacing, and the left spacing and right spacing together constitute the vehicle body position;

[0088] In practical applications, LL is obtained as 12m and RL as 15m. In this embodiment, the left and right boundary points are located in row 84, so PL(84,490) is named the simulation point. The simulation depth ML is obtained as 9.2m. PL(1,CC) is closest to the stereo camera, so the stereo camera can be approximated as being located at PL(1,CC). Connecting these lines yields the mid-distance line, left-distance line, and right-distance line as follows: Figure 2 As shown, Figure 2 To illustrate the center, left, and right distance lines in a moving view diagram, Figure 3 The diagram illustrates the positional relationship between the center line, left line, and right line and the automatic line marking vehicle. Since the automatic line marking vehicle is 12m away from the left boundary point, this means the length of the left line is 12m. The same applies to the center line and right line. Further calculations yield a left distance LD of 7.7m and a right distance RD of 11.8m. All calculation results are rounded to one decimal place.

[0089] The driving path planning unit is used to plan the driving path for the automated lane marking vehicle based on the positional relationship between the vehicle body and the road boundary.

[0090] The driving route planning unit is configured with a driving route planning strategy, which includes:

[0091] When put into use, control the automatic line marking vehicle to reach the line marking starting point, and ensure that the road width of the first starting distance in front of the line marking starting point remains unchanged and that it is a straight road;

[0092] The LD and RD of the automatic lane marking vehicle are detected at this time and named as the left starting distance and right starting distance, respectively, and represented by the symbols LSD and RSD. The starting road width is obtained by calculating LSD+RSD.

[0093] In practical applications, since the automatic lane marking vehicle can usually identify the road boundary 10m ahead, the initial starting distance is set to 20m. This ensures that the automatic lane marking vehicle can distinguish the positional relationship between the lane line to be marked and the road boundary in the initial state. The starting point of the marking is set manually. For example, on an unmarked highway, it is necessary to draw a lane line on the right side of the second lane from the left. At this time, the automatic lane marking vehicle is controlled to reach a straight section of road at least 20m long, and the binocular camera is 7.6m away from the left guardrail. The navigation system then identifies that the left distance LD after the automatic lane marking vehicle has traveled straight for 9m is 7.7m and the right distance RD is 11.8m. Since the road is straight and the automatic lane marking vehicle is traveling straight, the left starting distance LSD in the initial state of the automatic lane marking vehicle is also 7.7m and the right starting distance RSD is also 11.8m. At this time, the initial positional relationship between the automatic lane marking vehicle and the road boundary is determined, and the initial road width is calculated to be 19.5m.

[0094] The row number of the left and right boundary points obtained from the analysis is marked as E. Starting from n=E+1, the left and right boundary points of the nth row are analyzed. At the same time, PL(n,CC) is named the predicted travel point. The LD and RD of the predicted travel point are calculated and named as the left prediction distance and the right prediction distance, respectively, and represented by the symbols LYD and RYD.

[0095] Select any PL(n,m) in the nth row and name it as the navigation step point. Calculate the LD and RD of the navigation step point and name them as the left navigation distance and right navigation distance, respectively, and represent them by the symbols LFD and RFD.

[0096] Calculate LYD+RYD to obtain the predicted road width. Determine whether the predicted road width is equal to the initial road width. If yes, output a signal indicating equal road widths; otherwise, output a signal indicating unequal road widths.

[0097] If the output path widths are equal, the selected navigation travel point must satisfy LFD=LSD and RFD=RSD;

[0098] If the output path widths are unequal, calculate |LYD-LSD| and |RYD-RSD|, and name the calculation results as left channel offset and right channel offset, respectively, and represent them by the symbols LU and RU.

[0099] If LU≤RU, the selected navigation step point must satisfy LFD=LSD; if LU>RU, the selected navigation step point must satisfy RFD=RSD.

[0100] Starting from n=E+1, the navigation travel points are analyzed. Then, n is incremented and the analysis of navigation travel points is repeated. All navigation travel points are connected by a smooth curve to obtain the driving path. The path is updated in real time and the automatic lane marking vehicle is controlled to travel according to the path.

[0101] In practical applications, the left and right boundary points are analyzed in line 84, so E is 84. Starting with n=85, the left and right boundary points in line n are analyzed. The predicted travel point is actually the point reached when the vehicle travels straight. The left predicted distance LYD is calculated to be 7.7m and the right predicted distance RYD is 11.1m, calculated using the same method. The calculations for LFD and RFD are similar and will not be explained further in this embodiment. The calculated predicted road width is 18.8m, which is not equal to the initial road width. Therefore, a road width inequality signal is output. The calculated left lane offset LU is 0m and the right lane offset RU is 0.7m, where LU < RU. Generally, if LU is 0 and RU is not 0, it indicates that the right side of the road is narrowing or widening. In this case, the left road boundary should be used as a reference for path planning; otherwise, the opposite applies. If both LU and RU are not 0, it indicates that there is a curve in the road. In this case, the road boundary with the smaller change is used as a reference for path planning. It is usually difficult for LU=RU to occur. If it does occur, it means that the automatic lane marking vehicle has deviated from the driving trajectory. It can be readjusted according to the road boundary on either side. Since LU<RU in this embodiment, the selected navigation point must meet the requirement of LFD=7.7m, that is, navigation is based on the left boundary of the road and is not affected by the change in the width of the right lane. The same applies when LU>RU. Analyze the navigation points of each row and connect them to obtain the driving path. It should be noted that the driving path is updated in real time. That is, the navigation point at a distance will be updated when it reaches a nearby location. The latest updated navigation point shall prevail. In addition, this embodiment is only used for navigation. The vehicle driving and obstacle avoidance functions can use existing technologies.

[0102] Example 2, please refer to Figure 4 As shown, this application provides a visual intelligent navigation perception method for automated line marking vehicles, including the following steps:

[0103] Step S1: Install a binocular camera at the front end of the automatic line marking vehicle to capture the vehicle's visual image and density depth map during its journey; Step S1 includes the following sub-steps:

[0104] Step S101: Install a binocular camera at the front end of the automatic line marking vehicle. The binocular camera must be installed on the centerline of the automatic line marking vehicle and kept at the first height above the ground.

[0105] Step S102: During the movement of the automatic line marking vehicle, a moving visual image and a dense depth image are collected. The moving visual image is an RGB image obtained by direct shooting.

[0106] Step S2: Perform image preprocessing on the moving visual image to obtain a moving preprocessed image; specifically, perform denoising, image enhancement, and grayscale processing on the moving visual image to finally obtain the moving preprocessed image.

[0107] Step S3 involves extracting the lateral features of the road based on the preprocessed image and identifying the road boundaries on both sides using these features. Step S3 includes the following sub-steps:

[0108] Step S301: Extract the lateral features of the road based on the gray values ​​of each row of pixels in the preprocessed image.

[0109] Step S301 includes the following sub-steps:

[0110] Step S3011: Number the rows of pixels in the preprocessed image from bottom to top, denoted by the symbol n, where n is the row number. Number the columns of pixels in the preprocessed image from left to right, denoted by the symbol m, where m is the column number. Mark the pixel in the nth row and mth column of the preprocessed image as PL(n,m), where n and m are both non-zero natural numbers and (n,m) is the index of PL. Mark the gray value of PL(n,m) as PG(n,m).

[0111] Step S3012: Analyze each row independently and calculate [PG(n+1,m)+PG(n,m)+PG(n-1,m)] / 3 to obtain the sliding grayscale mean at all PL(n,m) in the nth row, represented by the symbol SG(n,m). Obtain the maximum value of m, which is max(m), where max() is the maximum value operator. Name max(m) / 2 as the center column value, represented by the symbol CC.

[0112] Step S3013: Sort and number SG(n,m) whose m is less than CC in descending order, and use the symbol SA. iThis indicates that SG(n,m) with m greater than CC are sorted and numbered in ascending order of m, using the symbol SB. i Let $\frac{ ... i and SB i This refers to the horizontal feature;

[0113] Step S302: Identify the road boundaries on both sides of the road by the differences in the lateral features of the road boundaries;

[0114] Step S302 includes the following sub-steps:

[0115] Step S3021: Perform cluster analysis on the horizontal features, and classify the horizontal features into different gray-level cluster groups, starting with i=1, and determine SA. i with SA i+1 Whether they belong to the same gray-level cluster group; if so, then SA i with SA i+1 They are grouped into a left homogeneous set, and SB is determined. i With SB i+1 Whether they belong to the same gray-level cluster group; if so, then SB i With SB i+1 Summarize them into a right homogeneous group, add one to i and judge again until max(i) is reached. Analyze and obtain different left homogeneous groups and right homogeneous groups, which are collectively called feature homogeneous groups. Merge feature homogeneous groups that have intersection.

[0116] Step S3022: Sort and number the left homogeneous group and the right homogeneous group in ascending order of i, and use the symbol LT respectively. h and RT g This indicates that h and g are both non-zero natural numbers, with h being the index of LT and g being the index of RT;

[0117] Step S3023: Name LT1 as the left reference group, and name the gray-level cluster group to which the left reference group belongs as the left reference cluster, starting with h=2. If LT h LT h+1 and LT h+2 If none of them are in the left-benchmark cluster, then LT will be... h The smallest SA in i i Mark it as the left boundary point; otherwise, increment h and re-evaluate.

[0118] Step S3024: Name RT1 as the right reference group, and name the gray-level cluster group to which the right reference group belongs as the right reference cluster, starting with g=2. If RT g RT g+1 and RT g+2 If none of them are in the right-benchmark cluster, then RTg The smallest SB in i i Mark it as the right boundary point; otherwise, increment g and re-evaluate.

[0119] Step S3025: Starting with n=2, extract the left and right boundary points of each row. If no left or right boundary point is found, add one to n and extract again until the left and right boundary points are found. The left and right boundary points are the road boundaries.

[0120] Step S4: Calculate the vehicle's position based on the road boundary and dense depth map, and plan a driving path for the automatic lane marking vehicle based on the positional relationship between the vehicle's position and the road boundary; Step S4 includes the following sub-steps:

[0121] Step S401: Calculate the body position of the automatic line marking vehicle based on the depth values ​​of the left and right boundary points in the dense depth map;

[0122] Step S401 includes the following sub-steps:

[0123] Step S4011: Name the depth value corresponding to the left boundary point in the dense depth map as the left front boundary depth, represented by the symbol LL; name the depth value corresponding to the right boundary point in the dense depth map as the right front boundary depth, represented by the symbol RL; the depth value represents the distance between the pixel and the stereo camera.

[0124] Step S4012: Name the PL(n,CC) in the same row as the left boundary point and the right boundary point as the simulation point, where PL(n,CC) represents PL(n,m) with m=CC. Obtain the depth value of the simulation point in the dense depth map and name it as the simulation depth, represented by the symbol ML.

[0125] Step S4013: Connect PL(1,CC) to the simulation point, the left boundary point and the right boundary point respectively to obtain the mid-distance line, the left distance line and the right distance line. Name the acute angle between the mid-distance line and the left distance line as the left auxiliary angle, and represent it with the symbol KL. Name the acute angle between the mid-distance line and the right distance line as the right auxiliary angle, and represent it with the symbol KR.

[0126] Step S4014: Solve the system of equations Obtain the left spacing of the automatic line marking vehicle, where LD is the left spacing, and solve the system of equations. The right spacing of the automatically marked line vehicle is obtained, where RD is the right spacing, and the left spacing and right spacing together constitute the vehicle body position;

[0127] Step S402: Plan the driving path for the automatic lane marking vehicle based on the positional relationship between the vehicle body and the road boundary;

[0128] Step S402 includes the following sub-steps:

[0129] Step S4021: When put into use, control the automatic line marking vehicle to reach the line marking starting point, and ensure that the road width of the first starting distance in front of the line marking starting point remains unchanged and that it is a straight road.

[0130] Step S4022: Detect the LD and RD of the automatic lane marking vehicle at this time, and name them as the left starting distance and the right starting distance, respectively, and represent them by the symbols LSD and RSD. Calculate LSD+RSD to obtain the starting road width.

[0131] Step S4023: Mark the row number of the left and right boundary points obtained from the analysis as E. Starting from n=E+1, analyze the left and right boundary points of the nth row. At the same time, name PL(n,CC) as the predicted travel point. Calculate the LD and RD of the predicted travel point, and name them as the left prediction distance and the right prediction distance, respectively, and represent them by the symbols LYD and RYD.

[0132] Step S4024: Select any PL(n,m) in the nth row and name it as the navigation travel point. Calculate the LD and RD of the navigation travel point and name them as the left navigation distance and the right navigation distance, respectively, and represent them by the symbols LFD and RFD.

[0133] Step S4025: Calculate LYD+RYD to obtain the predicted road width. Determine whether the predicted road width is equal to the initial road width. If yes, output a road width equal signal; otherwise, output a road width unequal signal.

[0134] Step S4026: If the output road width is equal, the selected navigation travel point must satisfy LFD=LSD and RFD=RSD.

[0135] Step S4027: If the output path width is unequal, calculate |LYD-LSD| and |RYD-RSD|, and name the calculation results as left track offset and right track offset respectively, represented by the symbols LU and RU respectively;

[0136] Step S4028: If LU≤RU, the selected navigation point must satisfy LFD=LSD; if LU>RU, the selected navigation point must satisfy RFD=RSD.

[0137] Step S4029: Starting from n=E+1, analyze the navigation travel points, add one to n and repeat the analysis of navigation travel points, connect all navigation travel points through a smooth curve to obtain the driving path, update the formal path in real time and control the automatic line marking vehicle to travel according to the formal path.

[0138] Example 3: This application provides an electronic device, which may include a processor, a communication interface, a memory, and a communication bus. The processor, communication interface, and memory communicate with each other via the communication bus. The memory stores computer-readable instructions, and the processor can call the instructions in the memory. When the computer-readable instructions are executed by the processor, steps such as those in a visual intelligent navigation perception method for an automated lane marking vehicle are performed to achieve the following functions: installing a binocular camera at the front end of the automated lane marking vehicle to acquire a traveling visual image and a dense depth map of the vehicle; performing image preprocessing on the traveling visual image to obtain a preprocessed traveling image; extracting lateral features of the road based on the preprocessed traveling image, and identifying the road boundaries on both sides of the road through the lateral features; calculating the vehicle's position based on the road boundaries and the dense depth map, and planning a driving path for the automated lane marking vehicle based on the positional relationship between the vehicle's position and the road boundaries.

[0139] Furthermore, when the logical instructions in the aforementioned memory can be implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0140] Example 4: This application also provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it performs the steps of the above-described visual intelligent navigation perception method for an automated lane marking vehicle to achieve the following functions: installing a binocular camera at the front end of the automated lane marking vehicle to acquire a traveling visual image and a dense depth image of the automated lane marking vehicle; performing image preprocessing on the traveling visual image to obtain a traveling preprocessed image; extracting the lateral features of the road based on the traveling preprocessed image, and identifying the road boundaries on both sides of the road through the lateral features; calculating the vehicle position of the automated lane marking vehicle based on the road boundaries and the dense depth image, and planning a driving path for the automated lane marking vehicle based on the positional relationship between the vehicle position and the road boundaries.

[0141] Based on the above description of the embodiments, the embodiments of the present invention can be provided as methods, systems, or computer program products. Based on this understanding, the above technical solutions, in essence or in terms of their contribution to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or certain parts of the embodiments.

[0142] In the embodiments provided in this application, it should be understood that the disclosed system or method can be implemented in other ways. The embodiments described above are merely illustrative. For example, the division of modules or units is only a logical functional division, and there may be other division methods in actual implementation. Furthermore, multiple modules or units may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the coupling or direct coupling or communication connection shown or discussed may be through some communication interfaces. The indirect coupling or communication connection between systems, modules, and units may be electrical, mechanical, or other forms.

[0143] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A visual intelligent navigation perception method for automated line marking vehicles, characterized in that, Includes the following steps: A binocular camera is installed at the front of the automated line marking vehicle to capture the vehicle's visual image and density depth map as it travels. The moving visual image is preprocessed to obtain a moving preprocessed image; Based on the preprocessed images of the road, the lateral features of the road are extracted, and the road boundaries on both sides of the road are identified through the lateral features. The position of the automated lane marking vehicle is calculated based on the road boundary and dense depth map, and the driving path of the automated lane marking vehicle is planned based on the positional relationship between the vehicle position and the road boundary.

2. The visual intelligent navigation and perception method for automated line marking vehicles according to claim 1, characterized in that, Installing a binocular camera at the front of the automated line marking vehicle to acquire its traveling visual image and dense depth map includes the following sub-steps: A binocular camera is installed at the front end of the automatic line marking vehicle. The binocular camera needs to be installed on the central axis of the automatic line marking vehicle, and the height above the ground should be maintained at the first height. During the movement of the automated line marking vehicle, a moving visual image and a dense depth image are collected. The moving visual image is an RGB image obtained by direct shooting.

3. The visual intelligent navigation perception method for automated line marking vehicles according to claim 2, characterized in that, Image preprocessing of the moving visual image specifically involves denoising, image enhancement, and grayscale conversion to obtain the preprocessed moving visual image.

4. The visual intelligent navigation perception method for automated line marking vehicles according to claim 3, characterized in that, Extracting lateral features of the road from the preprocessed image and identifying the road boundaries on both sides using these features includes the following sub-steps: The lateral features of the road are extracted based on the gray values ​​of each row of pixels in the preprocessed image. Identify the road boundaries on both sides of the road by recognizing the differences in the lateral features of the road boundaries.

5. A visual intelligent navigation perception method for automated line marking vehicles according to claim 4, characterized in that, Extracting the lateral features of the road based on the grayscale values ​​of each row of pixels in the preprocessed image includes the following sub-steps: The pixels in the preprocessed image are numbered from bottom to top, denoted by the symbol n, where n is the row number. The pixels in the preprocessed image are numbered from left to right, denoted by the symbol m, where m is the column number. The pixel in the nth row and mth column of the preprocessed image is marked as PL(n,m), where n and m are both non-zero natural numbers and (n,m) is the index of PL. The gray value of PL(n,m) is marked as PG(n,m). Analyze each row independently and calculate [PG(n+1,m)+PG(n,m)+PG(n-1,m)] / 3 to obtain the sliding grayscale mean of all PL(n,m) in the nth row, represented by the symbol SG(n,m). Obtain the maximum value of m, which is max(m), where max() is the maximum value operator. Name max(m) / 2 as the center column value, represented by the symbol CC. Sort and number SG(n,m) whose m is less than CC in descending order, using the symbol SA. i This indicates that SG(n,m) with m greater than CC are sorted and numbered in ascending order of m, using the symbol SB. i In this representation, i is a non-zero natural number and i is the index of SA and SB, where SA i and SB i This refers to the horizontal feature.

6. The visual intelligent navigation perception method for automated line marking vehicles according to claim 5, characterized in that, Identifying road boundaries on both sides of a road by recognizing differences in lateral features includes the following sub-steps: Cluster analysis is performed on the horizontal features, which are then grouped into different gray-level clusters, starting with i=1, to determine the SA. i with SA i+1 Whether they belong to the same gray-level cluster group; if so, then SA i with SA i+1 They are grouped into a left homogeneous set, and SB is determined. i With SB i+1 Whether they belong to the same gray-level cluster group; if so, then SB i With SB i+1 Summarize them into a right homogeneous group, add one to i and judge again until max(i) is reached. Analyze and obtain different left homogeneous groups and right homogeneous groups, which are collectively called feature homogeneous groups. Merge feature homogeneous groups that have intersection. Sort and number the left and right homogeneous groups in ascending order of i, and use the symbol LT respectively. h and RT g This indicates that h and g are both non-zero natural numbers, with h being the index of LT and g being the index of RT; Name LT1 the left baseline group, and name the gray-level cluster group to which the left baseline group belongs the left baseline cluster, starting with h=2. If LT h LT h+1 and LT h+2 If none of them are in the left-benchmark cluster, then LT will be... h The smallest SA in i i Mark it as the left boundary point; otherwise, increment h and re-evaluate. Name RT1 the right baseline group, and name the gray-level cluster group to which the right baseline group belongs the right baseline cluster, starting with g=2. If RT g RT g+1 and RT g+2 If none of them are in the right-benchmark cluster, then RT g The smallest SB in i i Mark it as the right boundary point; otherwise, increment g and re-evaluate. Starting with n=2, extract the left and right boundary points of each row. If no left or right boundary point is found, increment n and extract again until the left and right boundary points are obtained. The left and right boundary points are the road boundaries.

7. A visual intelligent navigation perception method for automated line marking vehicles according to claim 6, characterized in that, The calculation of the vehicle's position based on the road boundary and dense depth map, and the planning of the vehicle's driving path based on the positional relationship between the vehicle's position and the road boundary, includes the following sub-steps: The position of the vehicle body of the automatic line marking vehicle is calculated based on the depth values ​​of the left and right boundary points in the dense depth map. The system plans the driving path for the automated lane marking vehicle by analyzing the positional relationship between the vehicle body and the road boundary.

8. A visual intelligent navigation perception method for automated line marking vehicles according to claim 7, characterized in that, Calculating the vehicle body position of the automated line-marking vehicle based on the depth values ​​of the left and right boundary points in the dense depth map includes the following sub-steps: The depth value corresponding to the left boundary point in the dense depth map is named the left front boundary depth, denoted by the symbol LL, and the depth value corresponding to the right boundary point in the dense depth map is named the right front boundary depth, denoted by the symbol RL. The depth value represents the distance between the pixel and the stereo camera. Name the PL(n,CC) in the same row as the left and right boundary points as the simulation points, where PL(n,CC) represents PL(n,m) for m=CC. Obtain the depth value of the simulation points in the dense depth map and name it as the simulation depth, represented by the symbol ML. Connect PL(1,CC) to the simulation point, the left boundary point, and the right boundary point respectively to obtain the center line, the left line, and the right line. Name the acute angle between the center line and the left line as the left auxiliary angle, and represent it with the symbol KL. Name the acute angle between the center line and the right line as the right auxiliary angle, and represent it with the symbol KR. Solve the system of equations Obtain the left spacing of the automatic line marking vehicle, where LD is the left spacing, and solve the system of equations. The right spacing of the automatic line marking vehicle is obtained, where RD is the right spacing, and the left spacing and right spacing together constitute the vehicle body position.

9. A visual intelligent navigation perception method for automated line marking vehicles according to claim 8, characterized in that, Planning a driving path for an automated lane-marking vehicle based on the positional relationship between the vehicle body and the road boundary includes the following sub-steps: When put into use, control the automatic line marking vehicle to reach the line marking starting point, and ensure that the road width of the first starting distance in front of the line marking starting point remains unchanged and that it is a straight road; The LD and RD of the automatic lane marking vehicle are detected at this time and named as the left starting distance and right starting distance, respectively, and represented by the symbols LSD and RSD. The starting road width is obtained by calculating LSD+RSD. The row number of the left and right boundary points obtained from the analysis is marked as E. Starting from n=E+1, the left and right boundary points of the nth row are analyzed. At the same time, PL(n,CC) is named the predicted travel point. The LD and RD of the predicted travel point are calculated and named as the left prediction distance and the right prediction distance, respectively, and represented by the symbols LYD and RYD. Select any PL(n,m) in the nth row and name it as the navigation step point. Calculate the LD and RD of the navigation step point and name them as the left navigation distance and right navigation distance, respectively, and represent them by the symbols LFD and RFD. Calculate LYD+RYD to obtain the predicted road width. Determine whether the predicted road width is equal to the initial road width. If yes, output a signal indicating equal road widths; otherwise, output a signal indicating unequal road widths. If the output path widths are equal, the selected navigation travel point must satisfy LFD=LSD and RFD=RSD; If the output path widths are unequal, calculate |LYD-LSD| and |RYD-RSD|, and name the calculation results as left channel offset and right channel offset, respectively, and represent them by the symbols LU and RU. If LU≤RU, the selected navigation step point must satisfy LFD=LSD; if LU>RU, the selected navigation step point must satisfy RFD=RSD. Starting from n=E+1, the navigation travel points are analyzed. Then, n is incremented and the analysis of navigation travel points is repeated. All navigation travel points are connected by a smooth curve to obtain the driving path. The path is updated in real time and the automatic lane marking vehicle is controlled to travel according to the path.

10. A visual intelligent navigation and perception system for automated line marking vehicles, used to implement the visual intelligent navigation and perception method for automated line marking vehicles as described in any one of claims 1-9, characterized in that, It includes a visual image acquisition module, an image preprocessing module, a road boundary recognition module, and a driving path planning module; the visual image acquisition module, the image preprocessing module, and the driving path planning module are respectively connected to the road boundary recognition module for data transmission. The visual image acquisition module is used to install a binocular camera at the front end of the automatic line marking vehicle to acquire the moving visual image and dense depth image of the automatic line marking vehicle. The image preprocessing module is used to preprocess the moving visual image to obtain a moving preprocessed image. The road boundary recognition module is used to extract the lateral features of the road based on the preprocessed image and to identify the road boundaries on both sides of the road through the lateral features. The driving path planning module is used to calculate the vehicle position of the automatic lane marking vehicle based on the road boundary and dense depth map, and to plan the driving path of the automatic lane marking vehicle based on the positional relationship between the vehicle position and the road boundary.

Images