Lane line detection result post-processing method used under fuzzy and shielding conditions

Abstract

The invention relates to a lane line detection result post-processing method used under fuzzy and shielding conditions, and the method comprises the steps: U1, inputting the collected image data information of a road into an image semantic segmentation model, carrying out the semantic segmentation of an image, and obtaining the data information of a semantic segmentation result of a lane line of the road; and U2, based on the data information of the semantic segmentation result of the lane line of the road, performing connected domain processing on the semantic segmentation result of each lane line by adopting an area threshold algorithm based on a sine and cosine operator to obtain data information of a connected domain of each lane line. According to the method, the problems that when the segmentation result of a certain lane line is broken due to the reasons such as fuzzy and shielding of the lane line, if no processing is added, multiple lane lines can be fitted, so that the output of the lane lines is inaccurate, and the use of planning is influenced can be solved, the fitting accuracy of the lane lines is high, and the efficiency is high. And safe driving of the automatic driving vehicle is ensured.

Description

Technical Field

[0001] The present invention relates to the technical field of lane line detection, and in particular to a post-processing method for lane line detection results in blurry and occluded situations. Background Art

[0002] Lane detection: Lane detection is a basic module in autonomous driving. Its task is to identify lane lines from images through traditional methods or deep learning methods, and derive their color, category, position, lane line fitting curve equation, etc., which are used in the planning and positioning modules of autonomous driving.

[0003] In the prior art, a Chinese patent (application number: 202010178033.0, publication number: CN111460921A) discloses a lane line detection method based on multi-task semantic segmentation, the lane line detection method comprising the steps of: acquiring a road image; constructing a feature extraction encoding network based on a convolutional neural network and a fully convolutional decoding network with multi-task branch information fusion; obtaining a lane line binary segmentation mask image through the output of the above network, and then performing point set division on the lane line pixels in the lane line binary image through a morphological processing method; finally performing polynomial fitting on each pixel point set to obtain the final lane line discrimination result. However, the lane line segmentation effect in this scheme is not good. For example, if a lane line segmentation map is disconnected due to wear, occlusion, etc., then the lane line point set division method of this scheme, the disconnected segmentation map will obtain more than one lane line, that is, the number of point sets for each lane line is reduced, and the fitting effect is definitely not good. Summary of the invention

[0004] In view of the above shortcomings of the prior art, the present invention provides a post-processing method for lane line detection results in blurred and occluded conditions. It can not only solve the problem that when the lane line segmentation result is disconnected due to reasons such as lane line blurring and occlusion, if no processing is done, multiple lane lines will be fitted, resulting in inaccurate lane line output and affecting the use of planning, but also the lane line fitting has high accuracy, ensuring the safe driving of autonomous driving vehicles.

[0005] In order to achieve the above-mentioned purpose and other related purposes, the technical solution provided by the present invention is as follows:

[0006] A post-processing method for lane line detection results under blurry and occluded conditions, the method comprising:

[0007] U1. Input the collected road image data information into the image semantic segmentation model to perform image semantic segmentation, and obtain the data information of the semantic segmentation result of the road lane line;

[0008] U2. Based on the data information of the semantic segmentation result of the lane lines of the road, the area threshold algorithm based on the sine-cosine operator is used to perform connected domain processing on the semantic segmentation result of each lane line to obtain data information of the connected domain of each lane line;

[0009] U3. Based on the data information of the connected domain of each lane line, an improved grayscale centroid algorithm integrating random inertia weight and variation difference factor is used to extract the center point of the connected domain of each lane line to obtain the data information of the center point set of each lane line;

[0010] U4. Based on the data information of the center point set of each lane line, the center point set of the lane line is optimized by using a multi-objective sparrow optimization algorithm based on Logistics mapping to obtain the data information of the center point set of each lane line after optimization;

[0011] U5. Based on the data information of the center point set of each lane line after optimization, a lane line fitting model is constructed, each lane line is fitted, and the data information of the fitted lane line is output.

[0012] Furthermore, in step U1, inputting the collected road image data information into the image semantic segmentation model to perform image semantic segmentation includes:

[0013] U11. Based on the collected image data information of the road, grayscale processing is performed to obtain the image data information of the road after grayscale processing;

[0014] U12. Input the image data information of the road after the grayscale processing into the image semantic segmentation model for training and learning, and determine the semantic segmentation function Q of the model,

[0015]

[0016] Among them, x is the image data information of the road after grayscale processing, α1, α2 and α3 are the semantic segmentation factors of the image, and the trained image semantic segmentation model is obtained;

[0017] U13. Based on the trained image semantic segmentation model, the collected image data information of the road is input to perform semantic segmentation of the image, and obtain data information of the semantic segmentation result of the lane lines of the road.

[0018] Furthermore, the semantic segmentation factors α1, α2 and α3 of the image are,

[0019]

[0020] Wherein, x is the image data information of the road after grayscale processing.

[0021] Furthermore, in step U2, the use of the area threshold algorithm based on the sine-cosine operator to perform connected domain processing on the semantic segmentation result of each lane line includes:

[0022] U21. Based on the data information of the semantic segmentation result of the lane line of the road, the pixel points of the lane line are extracted, and a pixel point matrix of the lane line is constructed to obtain the matrix data information of the pixel points of the lane line;

[0023] U22. Based on the matrix data information of the lane line pixel points, perform region division and establish the region area function W of the lane line pixel points.

[0024]

[0025] Among them, y is the matrix data information of the pixel points of the lane line, β1, β2 and β3 are the sine and cosine operators of the lane line, and the areas of different regions of the lane line are calculated to obtain the data information of the areas of different regions of the lane line;

[0026] U23. Based on the data information of the areas of different lane line regions, a preset threshold is set. If the areas of different lane line regions are smaller than the preset threshold, the pixel points of the lane lines in the region are filtered out. If the areas of different lane line regions are larger than the preset threshold, the pixel points of the lane lines in the region are retained, and the connected areas of the lane lines are constructed to obtain the data information of the connected domain of each lane line.

[0027] Furthermore, the constraints of the sine and cosine operators β1, β2 and β3 of the lane line are:

[0028] Furthermore, in step U3, the extraction of the center point of the connected domain of each lane line by using the improved grayscale centroid algorithm integrating random inertia weight and variation difference factor includes:

[0029] U31. Based on the data information of the connected domain of each lane line, a random inertia weight function R of the lane line pixel point is established.

[0030]

[0031] Among them, z is the data information of the connected domain of each lane line, δ1, δ2 and δ3 are the tangent factors of the lane line pixels, and the random inertia weight values ​​of the lane line pixels are calculated to obtain the random inertia weight value data information of the pixels in the connected domain of each lane line;

[0032] U32. Based on the data information of the connected domain of each lane line, establish the lane line pixel point variation difference function P,

[0033]

[0034] Among them, z is the data information of the connected domain of each lane line, γ1, γ2 and γ3 are the cotangent factors of the lane line pixels, and the variation differential factors of the lane line pixels are calculated to obtain the data information of the variation differential factors of the pixels in the connected domain of each lane line;

[0035] U33. Based on the data information of the variation difference factor of the pixel points in the connected domain of each lane line and the data information of the random inertia weight value of the pixel points in the connected domain of each lane line, a center point function S of the connected domain of the lane line is established.

[0036]

[0037] Among them, z is the data information of the connected domain of each lane line, η is the data information of the random inertia weight value of the pixel points in the connected domain of each lane line, λ is the data information of the variation differential factor of the pixel points in the connected domain of each lane line, and the center point of the connected domain of each lane line is extracted to obtain the data information of the center point set of each lane line.

[0038] Furthermore, the tangent factors δ1, δ2 and δ3 of the lane line pixel points are:

[0039]

[0040] The cotangent factors γ1, γ2 and γ3 of the lane line pixel points are:

[0041]

[0042] Among them, z is the data information of the connected domain of each lane line.

[0043] Furthermore, in step U4, the optimization of the center point set of the lane line by using the multi-objective sparrow optimization algorithm based on Logistics mapping includes:

[0044] U41. Based on the data information of the center point set of each lane line, a logistics mapping function F of the lane line is established.

[0045]

[0046] Among them, r is the data information of the center point set of each lane line, μ1, μ2 and μ3 are the bias constant parameters of the lane line, and the chaotic sequence of the lane line is characterized to obtain the data information of the chaotic sequence of the lane line;

[0047] U42. Based on the data information of the chaotic sequence of the lane line, the sparrow population is initialized, the population parameters and the maximum number of iterations are determined, and the data information of the initialized sparrow population is obtained;

[0048] U43. Based on the data information of the initialized sparrow population, establish the fitness function G of the population individuals,

[0049]

[0050] Among them, h is the data information of the initialized sparrow population, θ1, θ2 and θ3 are the fitness determining factors of the sparrow population individuals, and the fitness values ​​of the sparrow population individuals are calculated to obtain the data information of the fitness values ​​of the sparrow population individuals;

[0051] U44. Based on the data information of the fitness value of the sparrow population individuals, establish a target optimization function H,

[0052]

[0053] Among them, a is the data information of the fitness value of the individual sparrow population, ρ1, ρ2 and ρ3 are optimization factors, and the center point set of the lane line is optimized to obtain the data information of the center point set of each lane line after optimization.

[0054] Furthermore, the constraints of the optimization factors ρ1, ρ2 and ρ3 are:

[0055]

[0056] Furthermore, in step U5, the lane line fitting model is constructed, and fitting each lane line includes:

[0057] U51. Based on the optimized data information of the center point set of each lane line, construct a lane line center point data set;

[0058] U52. Input the lane line center point data set into the lane line fitting model for training and learning, and determine the model function V,

[0059]

[0060] Among them, c is the center point data set of the lane line, σ1, σ2 and σ3 are the fitting factors of the lane line, and the trained lane line fitting model is obtained;

[0061] U53. Based on the trained lane line fitting model, the data information of the optimized center point set of each lane line is input, each lane line is fitted, and the data information of the fitted lane line is output.

[0062] The present invention has the following positive effects:

[0063] 1. The present invention performs semantic segmentation of the image by inputting the collected image data information of the road into the image semantic segmentation model, and performs connected domain processing on the semantic segmentation result of each lane line in combination with the area threshold algorithm based on the sine-cosine operator. At the same time, an improved grayscale centroid algorithm integrating random inertia weight and variation differential factor is used to extract the center point of the connected domain of each lane line. This can not only solve the problem that when the segmentation result of a lane line is disconnected due to reasons such as lane line blurring and occlusion, if no processing is done, multiple lane lines will be fitted, resulting in inaccurate lane line output and affecting the use of planning, but also the lane line fitting has high accuracy, ensuring the safe driving of the autonomous driving vehicle.

[0064] 2. The present invention optimizes the center point set of lane lines by adopting a multi-objective sparrow optimization algorithm based on Logistics mapping, and fits each lane line by building a lane line fitting model. It can not only accurately fit each lane line, thereby improving the efficiency of the operation of the autonomous driving vehicle, but also effectively reduce the risk of collision accidents of unmanned vehicles. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 It is a schematic diagram of the method flow of the present invention;

[0066] Figure 2 It is a flow chart of the image semantic segmentation model of the present invention;

[0067] Figure 3 Schematic diagram of the flow of the area threshold algorithm based on sine-cosine operators of the present invention;

[0068] Figure 4 It is a flow chart of an improved grayscale centroid algorithm integrating random inertia weight and variation difference factor of the present invention;

[0069] Figure 5 It is a flow chart of the multi-objective sparrow optimization algorithm based on Logistics mapping of the present invention;

[0070] Figure 6 A schematic diagram of the process of constructing a lane fitting model of the present invention;

[0071] Figure 7 A schematic diagram of lane lines before optimization of the present invention;

[0072] Figure 8 Schematic diagram of the optimized lane line of the present invention. DETAILED DESCRIPTION

[0073] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0074] Example 1: Figure 1 or Figure 7 or Figure 8 As shown, a post-processing method for lane line detection results under blurry and occluded conditions, the method comprising:

[0075] U1. Input the collected road image data information into the image semantic segmentation model to perform image semantic segmentation, and obtain the data information of the semantic segmentation result of the road lane line;

[0076] U2. Based on the data information of the semantic segmentation result of the lane lines of the road, the connected domain processing is performed on the semantic segmentation result of each lane line using an area threshold algorithm based on a sine-cosine operator to obtain data information of the connected domain of each lane line;

[0077] U3. Based on the data information of the connected domain of each lane line, an improved grayscale centroid algorithm integrating random inertia weight and variation difference factor is used to extract the center point of the connected domain of each lane line to obtain the data information of the center point set of each lane line;

[0078] U4. Based on the data information of the center point set of each lane line, the center point set of the lane line is optimized by using a multi-objective sparrow optimization algorithm based on Logistics mapping to obtain the data information of the center point set of each lane line after optimization;

[0079] U5. Based on the data information of the center point set of each lane line after optimization, a lane line fitting model is constructed, each lane line is fitted, and the data information of the fitted lane line is output.

[0080] In this embodiment, if Figure 2 As shown, in step U1, the step of inputting the collected road image data information into the image semantic segmentation model to perform image semantic segmentation includes:

[0081] U11. Based on the collected image data information of the road, grayscale processing is performed to obtain the image data information of the road after grayscale processing;

[0082] U12. Input the image data information of the road after the grayscale processing into the image semantic segmentation model for training and learning, and determine the semantic segmentation function Q of the model,

[0083]

[0084] Among them, x is the image data information of the road after grayscale processing, α1, α2 and α3 are the semantic segmentation factors of the image, and the trained image semantic segmentation model is obtained;

[0085] U13. Based on the trained image semantic segmentation model, the collected image data information of the road is input to perform semantic segmentation of the image, and obtain data information of the semantic segmentation result of the lane lines of the road.

[0086] In this embodiment, the semantic segmentation factors α1, α2 and α3 of the image are,

[0087]

[0088] Wherein, x is the image data information of the road after grayscale processing.

[0089] In this embodiment, if Figure 3 As shown, in step U2, the connected domain processing of the semantic segmentation result of each lane line using the area threshold algorithm based on the sine-cosine operator includes:

[0090] U21. Based on the data information of the semantic segmentation result of the lane line of the road, the pixel points of the lane line are extracted, and a pixel point matrix of the lane line is constructed to obtain the matrix data information of the pixel points of the lane line;

[0091] U22. Based on the matrix data information of the lane line pixel points, perform region division and establish the region area function W of the lane line pixel points.

[0092]

[0093] Among them, y is the matrix data information of the pixel points of the lane line, β1, β2 and β3 are the sine and cosine operators of the lane line, and the areas of different regions of the lane line are calculated to obtain the data information of the areas of different regions of the lane line;

[0094] U23. Based on the data information of the areas of different lane line regions, a preset threshold is set. If the areas of different lane line regions are smaller than the preset threshold, the pixel points of the lane lines in the region are filtered out. If the areas of different lane line regions are larger than the preset threshold, the pixel points of the lane lines in the region are retained, and the connected areas of the lane lines are constructed to obtain the data information of the connected domain of each lane line.

[0095] In this embodiment, the constraints of the sine and cosine operators β1, β2 and β3 of the lane line are:

[0096]

[0097] In this embodiment, if Figure 4 As shown, in step U3, the use of the improved grayscale centroid algorithm integrating random inertia weight and variation difference factor to extract the center point of the connected domain of each lane line includes:

[0098] U31. Based on the data information of the connected domain of each lane line, a random inertia weight function R of the lane line pixel point is established.

[0099]

[0100] Among them, z is the data information of the connected domain of each lane line, δ1, δ2 and δ3 are the tangent factors of the lane line pixels, and the random inertia weight values ​​of the lane line pixels are calculated to obtain the random inertia weight value data information of the pixels in the connected domain of each lane line;

[0101] U32. Based on the data information of the connected domain of each lane line, establish the lane line pixel point variation difference function P,

[0102]

[0103] Among them, z is the data information of the connected domain of each lane line, γ1, γ2 and γ3 are the cotangent factors of the lane line pixels, and the variation differential factors of the lane line pixels are calculated to obtain the data information of the variation differential factors of the pixels in the connected domain of each lane line;

[0104] U33. Based on the data information of the variation difference factor of the pixel points in the connected domain of each lane line and the data information of the random inertia weight value of the pixel points in the connected domain of each lane line, a center point function S of the connected domain of the lane line is established.

[0105]

[0106] Among them, z is the data information of the connected domain of each lane line, η is the data information of the random inertia weight value of the pixel points in the connected domain of each lane line, λ is the data information of the variation differential factor of the pixel points in the connected domain of each lane line, and the center point of the connected domain of each lane line is extracted to obtain the data information of the center point set of each lane line.

[0107] In this embodiment, the tangent factors δ1, δ2 and δ3 of the lane line pixel points are:

[0108] The cotangent factors γ1, γ2 and γ3 of the lane line pixel points are:

[0109]

[0110] Among them, z is the data information of the connected domain of each lane line.

[0111] Example 2: Based on the post-processing method of lane line detection results in blurred and occluded conditions in Example 1, the present invention is further illustrated and described below.

[0112] like Figure 1 or Figure 7 or Figure 8 As shown, a post-processing method for lane line detection results under blurry and occluded conditions, the method comprising:

[0113] U1. Input the collected road image data information into the image semantic segmentation model to perform image semantic segmentation, and obtain the data information of the semantic segmentation result of the road lane line;

[0114] U2. Based on the data information of the semantic segmentation result of the lane lines of the road, the connected domain processing is performed on the semantic segmentation result of each lane line using an area threshold algorithm based on a sine-cosine operator to obtain data information of the connected domain of each lane line;

[0115] U3. Based on the data information of the connected domain of each lane line, an improved grayscale centroid algorithm integrating random inertia weight and variation difference factor is used to extract the center point of the connected domain of each lane line to obtain the data information of the center point set of each lane line;

[0116] U4. Based on the data information of the center point set of each lane line, the center point set of the lane line is optimized by using a multi-objective sparrow optimization algorithm based on Logistics mapping to obtain the data information of the center point set of each lane line after optimization;

[0117] U5. Based on the data information of the center point set of each lane line after optimization, a lane line fitting model is constructed, each lane line is fitted, and the data information of the fitted lane line is output.

[0118] In this embodiment, if Figure 5 As shown, in step U4, the optimization of the center point set of the lane line by using the multi-objective sparrow optimization algorithm based on Logistics mapping includes:

[0119] U41. Based on the data information of the center point set of each lane line, a logistics mapping function F of the lane line is established.

[0120]

[0121] Among them, r is the data information of the center point set of each lane line, μ1, μ2 and μ3 are the bias constant parameters of the lane line, and the chaotic sequence of the lane line is characterized to obtain the data information of the chaotic sequence of the lane line;

[0122] U42. Based on the data information of the chaotic sequence of the lane line, the sparrow population is initialized, the population parameters and the maximum number of iterations are determined, and the data information of the initialized sparrow population is obtained;

[0123] U43. Based on the data information of the initialized sparrow population, establish the fitness function G of the population individuals,

[0124]

[0125] Among them, h is the data information of the initialized sparrow population, θ1, θ2 and θ3 are the fitness determining factors of the sparrow population individuals, and the fitness values ​​of the sparrow population individuals are calculated to obtain the data information of the fitness values ​​of the sparrow population individuals;

[0126] U44. Based on the data information of the fitness value of the sparrow population individuals, establish a target optimization function H,

[0127]

[0128] Among them, a is the data information of the fitness value of the individual sparrow population, ρ1, ρ2 and ρ3 are optimization factors, and the center point set of the lane line is optimized to obtain the data information of the center point set of each lane line after optimization.

[0129] In this embodiment, the constraints of the optimization factors ρ1, ρ2 and ρ3 are:

[0130]

[0131] In this embodiment, if Figure 6 As shown, in step U5, the lane line fitting model is constructed and each lane line is fitted, including:

[0132] U51. Based on the optimized data information of the center point set of each lane line, construct a lane line center point data set;

[0133] U52. Input the lane line center point data set into the lane line fitting model for training and learning, and determine the model function V,

[0134]

[0135] Among them, c is the center point data set of the lane line, σ1, σ2 and σ3 are the fitting factors of the lane line, and the trained lane line fitting model is obtained;

[0136] U53. Based on the trained lane line fitting model, the data information of the optimized center point set of each lane line is input, each lane line is fitted, and the data information of the fitted lane line is output.

[0137] In this embodiment, the present invention provides a post-processing system for lane line detection results in blurred or occluded conditions, including a computer device that is programmed or configured to execute any one of the steps of the post-processing method for lane line detection results in blurred or occluded conditions.

[0138] The present invention also provides a computer-readable storage medium, on which is stored a computer program programmed or configured to execute any one of the methods for post-processing lane line detection results in blurred or occluded situations.

[0139] Any reference to memory, storage, database or other media used in the embodiments provided in the present application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM) and memory bus dynamic RAM (RDRAM).

[0140] In summary, the present invention can not only solve the problem that when the segmentation result of a lane line is disconnected due to reasons such as lane line blurring and occlusion, if no processing is done, multiple lane lines will be fitted, resulting in inaccurate lane line output and affecting the use of planning, but also the lane line fitting has high accuracy, ensuring the safe driving of autonomous driving vehicles.

[0141] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A post-processing method for lane line detection results under blurry and occluded conditions, characterized in that: The method comprises: U1. Input the collected road image data information into the image semantic segmentation model to perform image semantic segmentation, and obtain the data information of the semantic segmentation result of the road lane line; U2. Based on the data information of the semantic segmentation result of the lane lines of the road, the area threshold algorithm based on the sine-cosine operator is used to perform connected domain processing on the semantic segmentation result of each lane line to obtain data information of the connected domain of each lane line; U3. Based on the data information of the connected domain of each lane line, an improved grayscale centroid algorithm integrating random inertia weight and variation difference factor is used to extract the center point of the connected domain of each lane line to obtain the data information of the center point set of each lane line; U4. Based on the data information of the center point set of each lane line, the center point set of the lane line is optimized by using a multi-objective sparrow optimization algorithm based on Logistics mapping to obtain the data information of the center point set of each lane line after optimization; U5. Based on the data information of the center point set of each lane line after optimization, a lane line fitting model is constructed, each lane line is fitted, and the data information of the fitted lane line is output.

2. The post-processing method for lane line detection results under blur and occlusion conditions according to claim 1, characterized in that: In step U1, the step of inputting the collected road image data information into the image semantic segmentation model to perform image semantic segmentation includes: U11. Based on the collected image data information of the road, grayscale processing is performed to obtain the image data information of the road after grayscale processing; U12. Input the image data information of the road after the grayscale processing into the image semantic segmentation model for training and learning, and determine the semantic segmentation function Q of the model, Among them, x is the image data information of the road after grayscale processing, α1, α2 and α3 are the semantic segmentation factors of the image, and the trained image semantic segmentation model is obtained; U13. Based on the trained image semantic segmentation model, the collected image data information of the road is input to perform semantic segmentation of the image, and obtain data information of the semantic segmentation result of the lane lines of the road.

3. The post-processing method for lane line detection results under blur and occlusion conditions according to claim 2, characterized in that: The semantic segmentation factors α1, α2 and α3 of the image are, Wherein, x is the image data information of the road after grayscale processing.

4. The post-processing method for lane line detection results under blur and occlusion conditions according to claim 1, characterized in that: In step U2, the connected domain processing of the semantic segmentation result of each lane line using the area threshold algorithm based on the sine-cosine operator includes: U21. Based on the data information of the semantic segmentation result of the lane line of the road, the pixel points of the lane line are extracted, and a pixel point matrix of the lane line is constructed to obtain the matrix data information of the pixel points of the lane line; U22. Based on the matrix data information of the lane line pixel points, perform region division and establish the region area function W of the lane line pixel points. Among them, y is the matrix data information of the pixel points of the lane line, β1, β2 and β3 are the sine and cosine operators of the lane line, and the areas of different regions of the lane line are calculated to obtain the data information of the areas of different regions of the lane line; U23. Based on the data information of the areas of different lane line regions, a preset threshold is set. If the areas of different lane line regions are smaller than the preset threshold, the pixel points of the lane lines in the region are filtered out. If the areas of different lane line regions are larger than the preset threshold, the pixel points of the lane lines in the region are retained, and the connected areas of the lane lines are constructed to obtain the data information of the connected domain of each lane line.

5. The post-processing method for lane line detection results under blur and occlusion conditions according to claim 4, characterized in that: The constraints of the sine and cosine operators β1, β2 and β3 of the lane line are:

6. The post-processing method for lane line detection results under blur and occlusion conditions according to claim 1, characterized in that: In step U3, the use of the improved grayscale centroid algorithm integrating random inertia weight and variation difference factor to extract the center point of the connected domain of each lane line includes: U31. Based on the data information of the connected domain of each lane line, a random inertia weight function R of the lane line pixel point is established. Among them, z is the data information of the connected domain of each lane line, δ1, δ2 and δ3 are the tangent factors of the lane line pixels, and the random inertia weight values ​​of the lane line pixels are calculated to obtain the random inertia weight value data information of the pixels in the connected domain of each lane line; U32. Based on the data information of the connected domain of each lane line, establish the lane line pixel point variation difference function P, Among them, z is the data information of the connected domain of each lane line, γ1, γ2 and γ3 are the cotangent factors of the lane line pixels, and the variation differential factors of the lane line pixels are calculated to obtain the data information of the variation differential factors of the pixels in the connected domain of each lane line; U33. Based on the data information of the variation difference factor of the pixel points in the connected domain of each lane line and the data information of the random inertia weight value of the pixel points in the connected domain of each lane line, a center point function S of the connected domain of the lane line is established. Among them, z is the data information of the connected domain of each lane line, η is the data information of the random inertia weight value of the pixel points in the connected domain of each lane line, λ is the data information of the variation differential factor of the pixel points in the connected domain of each lane line, and the center point of the connected domain of each lane line is extracted to obtain the data information of the center point set of each lane line.

7. The post-processing method for lane line detection results under blur and occlusion conditions according to claim 6, characterized in that: The tangent factors δ1, δ2 and δ3 of the lane line pixel points are: The cotangent factors γ1, γ2 and γ3 of the lane line pixel points are: Among them, z is the data information of the connected domain of each lane line.

8. The post-processing method for lane line detection results under blur and occlusion conditions according to claim 1, characterized in that: In step U4, the optimization of the center point set of the lane line by using the multi-objective sparrow optimization algorithm based on Logistics mapping includes: U41. Based on the data information of the center point set of each lane line, a logistics mapping function F of the lane line is established. Among them, r is the data information of the center point set of each lane line, μ1, μ2 and μ3 are the bias constant parameters of the lane line, and the chaotic sequence of the lane line is characterized to obtain the data information of the chaotic sequence of the lane line; U42. Based on the data information of the chaotic sequence of the lane line, the sparrow population is initialized, the population parameters and the maximum number of iterations are determined, and the data information of the initialized sparrow population is obtained; U43. Based on the data information of the initialized sparrow population, establish the fitness function G of the population individuals, Among them, h is the data information of the initialized sparrow population, θ1, θ2 and θ3 are the fitness determining factors of the sparrow population individuals, and the fitness values ​​of the sparrow population individuals are calculated to obtain the data information of the fitness values ​​of the sparrow population individuals; U44. Based on the data information of the fitness value of the sparrow population individuals, establish a target optimization function H, Among them, a is the data information of the fitness value of the individual sparrow population, ρ1, ρ2 and ρ3 are optimization factors, and the center point set of the lane line is optimized to obtain the data information of the center point set of each lane line after optimization.

9. The post-processing method for lane line detection results under blur and occlusion conditions according to claim 8, characterized in that: The constraints of the optimization factors ρ1, ρ2 and ρ3 are:

10. The post-processing method for lane line detection results under blur and occlusion conditions according to claim 1, characterized in that: In step U5, the lane line fitting model is constructed, and fitting each lane line includes: U51. Based on the optimized data information of the center point set of each lane line, construct a lane line center point data set; U52. Input the lane line center point data set into the lane line fitting model for training and learning, and determine the model function V, Among them, c is the center point data set of the lane line, σ1, σ2 and σ3 are the fitting factors of the lane line, and the trained lane line fitting model is obtained; U53. Based on the trained lane line fitting model, the data information of the optimized center point set of each lane line is input, each lane line is fitted, and the data information of the fitted lane line is output.

Images