A path planning method and system for guide vehicles in intelligent warehousing and sorting

By building a warehousing workspace map and using the optimal path selection strategy, the problem of insufficient accuracy and insufficient comprehensive consideration of factors in AGV path planning is solved, and the sorting efficiency and accuracy of AGV are improved.

CN118584954BActive Publication Date: 2024-11-19SHANGHAI DONGSHANG INFORMATION TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410625723.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-05-20
Publication Date
2024-11-19
Estimated Expiration
2044-05-20

AI Technical Summary

Technical Problem

The existing AGV path planning methods have insufficient accuracy and are unable to accurately reach the target position, which affects sorting efficiency and accuracy, and do not fully consider factors such as vehicle energy consumption, blockage conditions and collision avoidance, resulting in non-optimal solutions for the planned path.

Method used

By obtaining the storage environment image, extracting shelf placement information and path obstacle information, building a warehousing workspace map, and marking itself and target locations in this map, using the guide vehicle's optimal path selection strategy for path planning.

Benefits of technology

It improves the accuracy and efficiency of path planning, ensures that AGV can accurately reach the target position, optimizes the energy consumption and obstacle avoidance capabilities during the sorting process, and provides a more comprehensive path planning solution.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118584954B_ABST
    Figure CN118584954B_ABST
Patent Text Reader

Abstract

The present invention relates to the technical field of AGV path planning, and in particular to a path planning method and system for a guide vehicle in intelligent warehouse sorting. The present invention proposes the following scheme: first, a warehouse environment image is acquired, and shelf placement information and path obstacle information are extracted by processing the warehouse environment image. Secondly, a warehouse workspace map is constructed according to the shelf placement information and the path obstacle information. Finally, after receiving a sorting task, the guide vehicle marks its own position and the target shelf position in the warehouse workspace map, and the guide vehicle sorting is path planned by the guide vehicle optimal path selection strategy to improve the path planning efficiency.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of AGV path planning, and in particular to a path planning method and system for a guide vehicle in smart warehousing and sorting. Background Art

[0002] With the advent of the Industrial 4.0 era, the automation level of various industries is developing rapidly. The progress of science and technology has greatly promoted the prosperity of the intelligent manufacturing industry. As a part of the intelligent manufacturing industry, intelligent warehousing has received more and more attention and research from entrepreneurs and related scholars. How to improve the automation, intelligence, informatization and digitization of warehousing is one of the current research hotspots of intelligent warehousing. The widespread use of automated guided vehicles (AGVs) has enabled large-scale warehousing systems to complete cargo transportation with only a small amount of manpower. AGV path planning is a multi-objective combination optimization problem. In addition to paying attention to the rapidity of the algorithm, it is also necessary to consider the shortest path distance, the shortest time and the lowest system energy consumption.

[0003] For example, a Chinese patent with publication number CN117850415A discloses an AGV path planning method based on deep reinforcement learning, including the following process: initialization of the valuation network, target network, environment, and experience pool; path planning begins; the intelligent agent interacts with the environment and obtains at under St; obtains a reward rt, updates the state, obtains St+1, and puts St, at, rt, and St+1 into the replay pool; randomly extracts a set of St, at, rt, and St+1 from the replay pool; optimizes the loss function; updates the valuation network Q; sets the valuation network at this time as the new target network; and path planning ends. By adopting the above technical solution and the AGV path planning method based on deep reinforcement learning, the AGV system can self-optimize and find a better path for complex and unknown environments, further improving the working efficiency of the AGV system and the environmental adaptability of the AGV path planning.

[0004] The above patents all have the problems raised by this background technology: the current path planning method has the problem of insufficient accuracy when guiding vehicles, which causes the vehicle to be unable to accurately reach the target location, affecting the sorting efficiency and accuracy. When planning the path, only a single factor such as the shortest path or the fastest path may be considered, without fully considering the comprehensive influence of other factors, such as the vehicle's energy consumption, congestion, collision avoidance, etc., resulting in the planned path not being the optimal solution. In order to solve the above problems, this application designs a path planning method and system for guiding vehicles in smart warehousing sorting. Summary of the invention

[0005] The technical problem to be solved by the present invention is to address the deficiencies of the prior art and provide a path planning method and system for a guide vehicle in intelligent warehouse sorting. First, a warehouse environment image is acquired, and shelf placement information and path obstacle information are extracted by processing the warehouse environment image. Secondly, a warehouse workspace map is constructed according to the shelf placement information and path obstacle information. Finally, after receiving the sorting task, the guide vehicle marks its own position and the target shelf position in the warehouse workspace map, and the guide vehicle sorting path is planned through the guide vehicle optimal path selection strategy.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A path planning method for a guide vehicle in intelligent warehousing sorting includes the following steps:

[0008] S1: Acquire a warehouse environment image, and extract shelf placement information and path obstacle information by processing the warehouse environment image;

[0009] S2: Build a warehouse workspace map based on shelf placement information and path obstacle information;

[0010] S3: After receiving the sorting task, the guide vehicle marks its own position and the target shelf position in the warehouse workspace map, and plans the path for the guide vehicle sorting through the guide vehicle's optimal path selection strategy.

[0011] The specific steps of S1 are as follows:

[0012] S1.1: Fix a camera at the center of the warehouse ceiling, collect warehouse floor images through the camera, pre-process the warehouse floor images, and obtain foreground enhancement features, wherein the pre-processing includes transmission correction, background filtering, feature extraction, and feature enhancement;

[0013] S1.2: According to the encoder, the foreground enhancement feature is converted into an encoding feature vector. The encoding feature vector generates a hidden vector h1 through a gated recurrent unit. Another gated recurrent unit calculates in the opposite direction to generate a hidden vector h2. The two hidden vectors are concatenated and the entity feature vector is obtained through a fully connected layer.

[0014] S1.3: Obtain the attention weight of the entity feature vector through the attention network, reconstruct the entity feature vector according to the attention weight, and decode the reconstructed entity feature vector through the decoder;

[0015] S1.4: According to the decoding value and the binary cross entropy, the entity feature vector is compensated for the loss, the edge information in the entity feature vector is removed, the entity feature vector is classified, and the entity category is obtained, where the entity category includes storage shelves and path obstacles;

[0016] S1.5: Extract shelf placement information and path obstacle information based on the entity category of the entity feature vector.

[0017] The specific steps of preprocessing the warehouse ground image are as follows:

[0018] S1.1.1: The warehouse ground image is converted into a standard top-view image through the transmission transformation matrix. The calculation formula of the transmission transformation matrix is:

[0019]

[0020] Among them, A represents the transmission transformation matrix, x0, x1, y0 and y1 represent the coordinates corresponding to the four corner vertices of the warehouse ground image, and X0, X1, Y0 and Y1 represent the coordinates corresponding to the four corner vertices after the warehouse ground image matrix is ​​deformed;

[0021] S1.1.2: Perform bilateral filtering on the standard overhead image, retain the edge intensity change information of the standard overhead image, obtain the spatially enhanced image of the standard overhead image, construct a cost function, perform resolution estimation, iteratively correct the nodes, and delete the pixel nodes smaller than the background threshold according to the corrected nodes to remove the background;

[0022] S1.1.3: Use the mixed Gaussian model to differentiate the standard top-view image after removing the background, and determine whether the pixel belongs to the foreground feature according to the residence time of the color of different pixel points in the image;

[0023] S1.1.4: Aggregate the pixels belonging to the foreground feature, enhance the foreground feature pixels according to Fourier transform, and obtain the foreground enhancement feature. The calculation formula of the enhancement feature is:

[0024]

[0025] Where E represents the foreground enhancement feature, i represents a single pixel belonging to the foreground feature, J represents the total number of pixels belonging to the foreground feature, f(·) represents the feature grayscale function, g(·) represents the feature density function, and e i represents the horizontal pixel value of the i-th pixel, f i represents the vertical pixel value of the i-th pixel, j represents the Fourier transform imaginary value, u represents the frequency variable, and v represents the amplitude variable.

[0026] The specific steps of constructing a warehouse workspace map based on shelf placement information and path obstacle information are as follows:

[0027] S2.1: Proportionally reduce the size of the warehouse horizontal space, initialize the warehouse workspace map according to the reduced warehouse horizontal space, and determine the spatial range of the warehouse workspace map;

[0028] S2.2: taking the larger value of the length and width of the guide vehicle as the dividing value, setting the guide vehicle driving safety threshold, the guide vehicle driving safety threshold is determined by the guide vehicle turning radius and the guide vehicle driving deviation, and calculating the spatial cutting parameter according to the dividing value and the guide vehicle driving safety threshold;

[0029] S2.3: Determine the number of columns and rows of the storage workspace map according to the spatial range and spatial cutting parameters of the storage workspace map, and perform raster representation on the storage workspace map;

[0030] S2.4: Mark the shelf placement information and path obstacle information on the rasterized warehouse workspace map.

[0031] The specific steps of the optimal path selection strategy for the guide vehicle are as follows:

[0032] S3.1: Initialize the guide vehicle path, determine the starting point of the guide vehicle, determine the resultant force at the starting point by using an artificial potential field method, and determine the driving direction of the guide vehicle according to the resultant force, where the resultant force is determined by the gravitational force generated by the target point and the repulsive force generated by the obstacle;

[0033] S3.2: Set a random threshold p and generate a random step length q, determine the size of the random step length and the random threshold, if the random step length is less than or equal to the random threshold, based on the driving direction of the guide vehicle, set the node with a distance of q from the starting point of the guide vehicle as the random node P rand ;

[0034] S3.3: Get a random node P rand Neighborhood range, calculate the random node P rand The local cost value to the neighborhood parent node is used as the starting point of the guide vehicle. S3.1 to S3.3 are repeated until the target point is reached. The guide vehicle path is output. The calculation formula of the local cost value is:

[0035]

[0036] Among them, L v represents the local cost of guiding the car to the corresponding parent node, X d represents the distance from the leading vehicle to the corresponding parent node, V represents the speed of the leading vehicle, h s The load factor representing the linear load of the guide vehicle, h w It represents the load coefficient of the guide vehicle curve load, s represents the curvature from the guide vehicle to the corresponding parent node, and D represents the maximum curvature of the guide vehicle.

[0037] A path planning system for a guide vehicle in intelligent warehousing sorting, the system comprising a warehouse workspace map creation module and a guide vehicle path planning module;

[0038] The warehouse workspace map creation module is used to obtain a warehouse environment image, extract shelf placement information and path obstacle information by processing the warehouse environment image, and construct a warehouse workspace map based on the shelf placement information and path obstacle information;

[0039] The guide vehicle path planning module is used to plan the path for guide vehicle sorting through the guide vehicle optimal path selection strategy.

[0040] The warehouse workspace map creation module includes:

[0041] A storage environment image acquisition unit, used to acquire a storage environment image;

[0042] A storage environment image processing unit is used to process the storage environment image and extract shelf placement information and path obstacle information;

[0043] The rasterized map creation unit is used to rasterize the warehouse workspace map and mark the shelf placement information and path obstacle information on the rasterized warehouse workspace map.

[0044] The guide vehicle path planning module includes:

[0045] The guide vehicle marking unit is used for the guide vehicle to mark its own position and the target shelf position in the warehouse workspace map after receiving the sorting task;

[0046] The path planning unit is used to plan the path for the guide vehicle sorting through the guide vehicle optimal path selection strategy.

[0047] Compared with the prior art, the present invention has the following beneficial effects:

[0048] The present invention adopts a method that combines traditional path planning methods and artificial intelligence to overcome the low computational efficiency of traditional path planning methods in high-dimensional and large environments, while making up for the problem of long learning iteration time of a single artificial intelligence method, thereby improving path planning efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Other features, objects and advantages of the present invention will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0050] Figure 1 This is a flow chart of a method for path planning of a guide vehicle in intelligent warehousing sorting according to Embodiment 1 of the present invention;

[0051] Figure 2 This is a network structure diagram of entity feature extraction and recognition in Example 1 of the present invention;

[0052] Figure 3 This is a schematic diagram of rasterization of a workspace map according to Embodiment 1 of the present invention;

[0053] Figure 4 This is a module diagram of a path planning system for a guide vehicle in intelligent warehousing and sorting according to Embodiment 2 of the present invention. DETAILED DESCRIPTION

[0054] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0055] Embodiment 1:

[0056] See also Figure 1 , an embodiment provided by the present invention: a path planning method for a guide vehicle in intelligent warehousing sorting, comprising the following steps:

[0057] S1: Acquire a warehouse environment image, and extract shelf placement information and path obstacle information by processing the warehouse environment image;

[0058] S2: Build a warehouse workspace map based on shelf placement information and path obstacle information;

[0059] S3: After receiving the sorting task, the guide vehicle marks its own position and the target shelf position in the warehouse workspace map, and plans the path for the guide vehicle sorting through the guide vehicle's optimal path selection strategy.

[0060] The specific steps of S1 are as follows:

[0061] S1.1: Fix a camera at the center of the warehouse ceiling, collect warehouse floor images through the camera, pre-process the warehouse floor images, and obtain foreground enhancement features, wherein the pre-processing includes transmission correction, background filtering, feature extraction, and feature enhancement;

[0062] S1.2: According to the encoder, the foreground enhancement feature is converted into an encoding feature vector. The encoding feature vector generates a hidden vector h1 through a gated recurrent unit. Another gated recurrent unit calculates in the opposite direction to generate a hidden vector h2. The two hidden vectors are concatenated and the entity feature vector is obtained through a fully connected layer.

[0063] S1.3: Obtain the attention weight of the entity feature vector through the attention network, reconstruct the entity feature vector according to the attention weight, and decode the reconstructed entity feature vector through the decoder;

[0064] S1.4: According to the decoding value and the binary cross entropy, the entity feature vector is compensated for the loss, the edge information in the entity feature vector is removed, the entity feature vector is classified, and the entity category is obtained, where the entity category includes storage shelves and path obstacles;

[0065] S1.5: Extract shelf placement information and path obstacle information based on the entity category of the entity feature vector.

[0066] See also Figure 2 , a network structure diagram of entity feature extraction and recognition in an embodiment of the present invention, wherein the entity feature extraction and recognition network comprises an input layer, an encoding layer, a decoding layer and an output layer, wherein the input layer is used to preprocess the warehouse floor image and obtain foreground enhancement features, the encoding layer is used to obtain the entity feature vector, the decoding layer is used to obtain the attention weight of the entity feature vector through the attention network, reconstruct the entity feature vector according to the attention weight, and decode the reconstructed entity feature vector through the decoder, and the output layer is used to compensate the loss of the entity feature vector according to the decoding value and the binary cross entropy, remove the edge information in the entity feature vector, classify the entity feature vector, obtain the entity category, and extract the shelf placement information and the path obstacle information;

[0067] The specific steps of preprocessing the warehouse ground image are as follows:

[0068] S1.1.1: The warehouse ground image is converted into a standard top-view image through the transmission transformation matrix. The calculation formula of the transmission transformation matrix is:

[0069]

[0070] Among them, A represents the transmission transformation matrix, x0, x1, y0 and y1 represent the coordinates corresponding to the four corner vertices of the warehouse ground image, and X0, X1, Y0 and Y1 represent the coordinates corresponding to the four corner vertices after the warehouse ground image matrix is ​​deformed;

[0071] S1.1.2: Perform bilateral filtering on the standard overhead image, retain the edge intensity change information of the standard overhead image, obtain the spatially enhanced image of the standard overhead image, construct the cost function, perform resolution estimation, iteratively correct the nodes, and delete the pixel nodes smaller than the background threshold according to the corrected nodes to remove the background. The calculation formula of the bilateral filter of the standard overhead image is:

[0072]

[0073] Where BF[I] represents the standard top-view image bilateral filter, W prepresents the normalization factor, a represents the horizontal coordinate of the standard top-view image pixel, b represents the vertical coordinate of the standard top-view image pixel, G σs represents the standard top-view image space Gaussian, G σr represents the standard top-view image distance Gaussian, I a Indicates the horizontal pixels of the standard top-view image, I b Indicates the vertical pixels of a standard top-view image;

[0074] S1.1.3: Use the mixed Gaussian model to differentiate the standard top-view image after removing the background, and determine whether the pixel belongs to the foreground feature according to the residence time of the color of different pixel points in the image;

[0075] S1.1.4: Aggregate the pixels belonging to the foreground feature, enhance the foreground feature pixels according to Fourier transform, and obtain the foreground enhancement feature. The calculation formula of the enhancement feature is:

[0076]

[0077] Where E represents the foreground enhancement feature, i represents a single pixel belonging to the foreground feature, J represents the total number of pixels belonging to the foreground feature, f(·) represents the feature grayscale function, g(·) represents the feature density function, and e i represents the horizontal pixel value of the i-th pixel, f i represents the vertical pixel value of the i-th pixel, j represents the Fourier transform imaginary value, u represents the frequency variable, and v represents the amplitude variable.

[0078] The specific steps of constructing a warehouse workspace map based on shelf placement information and path obstacle information are as follows:

[0079] S2.1: Proportionally reduce the size of the warehouse horizontal space, initialize the warehouse workspace map according to the reduced warehouse horizontal space, and determine the spatial range of the warehouse workspace map;

[0080] S2.2: Take the larger value of the length and width of the guide vehicle as the dividing value. If the length and width are equal, select the length as the dividing value. Set the guide vehicle driving safety threshold. The guide vehicle driving safety threshold is determined by the guide vehicle turning radius and the guide vehicle driving deviation. Calculate the spatial cutting parameter according to the dividing value and the guide vehicle driving safety threshold;

[0081] S2.3: Determine the number of columns and rows of the storage workspace map according to the spatial range and spatial cutting parameters of the storage workspace map, and perform raster representation on the storage workspace map;

[0082] S2.4: Mark the shelf placement information and path obstacle information on the rasterized warehouse workspace map.

[0083] See also Figure 3 , a schematic diagram of rasterization of a workspace map according to an embodiment of the present invention, assuming that the storage space is 36m×36m in size, the guide vehicle division value is 2, and the guide vehicle driving safety threshold is 1.5, then the space cutting parameter is 3.5, and the number of columns and rows of the storage workspace map are calculated by rounding down to 10, and the map is numbered according to the number of grids and divided into 100 area numbers, and the shelf location information and the path obstacle location information are extracted from the shelf placement information and the path obstacle information, and the corresponding area of ​​the storage workspace map is marked, and the mark indicates that it is inaccessible;

[0084] The specific steps of the optimal path selection strategy for the guide vehicle are as follows:

[0085] S3.1: Initialize the guide vehicle path, determine the starting point of the guide vehicle, determine the resultant force at the starting point by using an artificial potential field method, and determine the driving direction of the guide vehicle according to the resultant force, where the resultant force is determined by the gravitational force generated by the target point and the repulsive force generated by the obstacle;

[0086] S3.2: Set a random threshold p and generate a random step length q, determine the size of the random step length and the random threshold, if the random step length is less than or equal to the random threshold, based on the driving direction of the guide vehicle, set the node with a distance of q from the starting point of the guide vehicle as the random node P rand ;

[0087] S3.3: Get a random node P rand Neighborhood range, calculate the random node P rand The local cost value to the neighborhood parent node is used as the starting point of the guide vehicle. S3.1 to S3.3 are repeated until the target point is reached. The guide vehicle path is output. The calculation formula of the local cost value is:

[0088]

[0089] Among them, L v represents the local cost of guiding the car to the corresponding parent node, X d represents the distance from the leading vehicle to the corresponding parent node, V represents the speed of the leading vehicle, h s The load factor representing the linear load of the guide vehicle, h w represents the load factor of the guide vehicle curve load, s represents the curvature from the guide vehicle to the corresponding parent node, and D represents the maximum curvature of the guide vehicle;

[0090] The principle of searching the initial direction by the artificial potential field method is to let the target point give the guide vehicle a gravitational force, and the obstacles will produce a repulsive force on the guide vehicle, so that the guide vehicle can quickly move towards the target point. The calculation formula of the gravitational force of the target point on the guide vehicle is:

[0091]

[0092] Among them, U a represents the gravitational force of the target point on the guide vehicle, μ represents the positive proportional coefficient of the gravitational potential field, and X represents the Euclidean distance from the target point to the guide vehicle;

[0093] The calculation formula of the repulsive force of the obstacle on the guide vehicle is:

[0094]

[0095] Among them, U r represents the repulsive force of the obstacle on the guide vehicle, η represents the inverse proportional coefficient of the repulsive potential field, ρ represents the Euclidean distance from the obstacle to the guide vehicle, and ρ0 represents the influence range of the obstacle.

[0096] Embodiment 2:

[0097] See also Figure 4 , the present invention provides an embodiment: a path planning system for a guide vehicle in intelligent warehousing sorting, the system comprising a warehouse workspace map creation module and a guide vehicle path planning module;

[0098] The warehouse workspace map creation module is used to obtain a warehouse environment image, extract shelf placement information and path obstacle information by processing the warehouse environment image, and construct a warehouse workspace map based on the shelf placement information and path obstacle information;

[0099] The guide vehicle path planning module is used to plan the path for guide vehicle sorting through the guide vehicle optimal path selection strategy.

[0100] The warehouse workspace map creation module includes:

[0101] A storage environment image acquisition unit, used to acquire a storage environment image;

[0102] A storage environment image processing unit is used to process the storage environment image and extract shelf placement information and path obstacle information;

[0103] The rasterized map creation unit is used to rasterize the warehouse workspace map and mark the shelf placement information and path obstacle information on the rasterized warehouse workspace map.

[0104] The guide vehicle path planning module includes:

[0105] The guide vehicle marking unit is used for the guide vehicle to mark its own position and the target shelf position in the warehouse workspace map after receiving the sorting task;

[0106] The path planning unit is used to plan the path for the guide vehicle sorting through the guide vehicle optimal path selection strategy.

[0107] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A path planning method for a guide vehicle in intelligent warehousing sorting, characterized in that: The following steps are involved: S1: Acquire a warehouse environment image, and extract shelf placement information and path obstacle information by processing the warehouse environment image; S2: Build a warehouse workspace map based on shelf placement information and path obstacle information; S3: After receiving the sorting task, the guide vehicle marks its own position and the target shelf position in the warehouse workspace map, and plans the path for the guide vehicle sorting through the guide vehicle's optimal path selection strategy; The specific steps of the optimal path selection strategy for the guide vehicle are as follows: S3.1: Initialize the guide vehicle path, determine the starting point of the guide vehicle, determine the resultant force at the starting point by using an artificial potential field method, and determine the driving direction of the guide vehicle according to the resultant force, where the resultant force is determined by the gravitational force generated by the target point and the repulsive force generated by the obstacle; S3.2: Set a random threshold p and generate a random step length q, determine the size of the random step length and the random threshold, if the random step length is less than or equal to the random threshold, based on the driving direction of the guide vehicle, set the node with a distance of q from the starting point of the guide vehicle as the random node P rand ; S3.3: Get a random node P rand Neighborhood range, calculate the random node P rand The local cost value to the neighborhood parent node is used as the starting point of the guide vehicle. S3.1 to S3.3 are repeated until the target point is reached. The guide vehicle path is output. The calculation formula of the local cost value is: Among them, L v represents the local cost of guiding the car to the corresponding parent node, X d represents the distance from the leading vehicle to the corresponding parent node, V represents the speed of the leading vehicle, h s The load factor representing the linear load of the guide vehicle, h w It represents the load coefficient of the guide vehicle curve load, s represents the curvature from the guide vehicle to the corresponding parent node, and D represents the maximum curvature of the guide vehicle.

2. According to claim 1, a path planning method for a guide vehicle in intelligent warehousing sorting, characterized in that: The specific steps of S1 are as follows: S1.1: Fix a camera at the center of the warehouse ceiling, collect warehouse floor images through the camera, pre-process the warehouse floor images, and obtain foreground enhancement features, wherein the pre-processing includes transmission correction, background filtering, feature extraction, and feature enhancement; S1.2: According to the encoder, the foreground enhancement feature is converted into an encoding feature vector. The encoding feature vector generates a hidden vector h1 through a gated recurrent unit. Another gated recurrent unit calculates in the opposite direction to generate a hidden vector h2. The two hidden vectors are concatenated and the entity feature vector is obtained through a fully connected layer. S1.3: Obtain the attention weight of the entity feature vector through the attention network, reconstruct the entity feature vector according to the attention weight, and decode the reconstructed entity feature vector through the decoder; S1.4: According to the decoding value and the binary cross entropy, the entity feature vector is compensated for the loss, the edge information in the entity feature vector is removed, the entity feature vector is classified, and the entity category is obtained, where the entity category includes storage shelves and path obstacles; S1.5: Extract shelf placement information and path obstacle information based on the entity category of the entity feature vector.

3. According to claim 2, a path planning method for a guide vehicle in intelligent warehousing sorting, characterized in that: The specific steps of preprocessing the warehouse ground image are as follows: S1.1.1: The warehouse ground image is converted into a standard top-view image through the transmission transformation matrix. The calculation formula of the transmission transformation matrix is: Among them, A represents the transmission transformation matrix, x0, x1, y0 and y1 represent the coordinates corresponding to the four corner vertices of the warehouse ground image, and X0, X1, Y0 and Y1 represent the coordinates corresponding to the four corner vertices after the warehouse ground image matrix is ​​deformed; S1.1.2: Perform bilateral filtering on the standard overhead image, retain the edge intensity change information of the standard overhead image, obtain the spatially enhanced image of the standard overhead image, construct a cost function, perform resolution estimation, iteratively correct the nodes, and delete the pixel nodes smaller than the background threshold according to the corrected nodes to remove the background; S1.1.3: Use the mixed Gaussian model to differentiate the standard top-view image after removing the background, and determine whether the pixel belongs to the foreground feature according to the residence time of the color of different pixel points in the image; S1.1.4: Aggregate the pixels belonging to the foreground feature, enhance the foreground feature pixels according to Fourier transform, and obtain the foreground enhancement feature. The calculation formula of the enhancement feature is: Where E represents the foreground enhancement feature, i represents a single pixel belonging to the foreground feature, J represents the total number of pixels belonging to the foreground feature, f(·) represents the feature grayscale function, g(·) represents the feature density function, and e i represents the horizontal pixel value of the i-th pixel, f i represents the vertical pixel value of the i-th pixel, j represents the imaginary value of Fourier transform, u represents the frequency variable, and v represents the amplitude variable.

4. According to claim 3, a path planning method for a guide vehicle in intelligent warehousing sorting is characterized in that: The specific steps of constructing a warehouse workspace map based on shelf placement information and path obstacle information are as follows: S2.1: Proportionally reduce the size of the warehouse horizontal space, initialize the warehouse workspace map according to the reduced warehouse horizontal space, and determine the spatial range of the warehouse workspace map; S2.2: taking the larger value of the length and width of the guide vehicle as the dividing value, setting the guide vehicle driving safety threshold, the guide vehicle driving safety threshold is determined by the guide vehicle turning radius and the guide vehicle driving deviation, and calculating the spatial cutting parameter according to the dividing value and the guide vehicle driving safety threshold; S2.3: Determine the number of columns and rows of the storage workspace map according to the spatial range and spatial cutting parameters of the storage workspace map, and perform rasterization on the storage workspace map; S2.4: Mark the shelf placement information and path obstacle information on the rasterized warehouse workspace map.

5. A path planning system for a guide vehicle in intelligent warehousing and sorting, implemented based on a path planning method for a guide vehicle in intelligent warehousing and sorting as claimed in any one of claims 1 to 4, characterized in that: The system includes a warehouse workspace map creation module and a guide vehicle path planning module; The warehouse workspace map creation module is used to obtain a warehouse environment image, extract shelf placement information and path obstacle information by processing the warehouse environment image, and construct a warehouse workspace map based on the shelf placement information and path obstacle information; The guide vehicle path planning module is used to plan the path for guide vehicle sorting through the guide vehicle optimal path selection strategy.

6. According to claim 5, a path planning system for a guide vehicle in intelligent warehousing and sorting, characterized in that: The warehouse workspace map creation module includes: A storage environment image acquisition unit, used to acquire a storage environment image; A storage environment image processing unit is used to process the storage environment image and extract shelf placement information and path obstacle information; The rasterized map creation unit is used to rasterize the warehouse workspace map and mark the shelf placement information and path obstacle information on the rasterized warehouse workspace map.

7. According to claim 5, a path planning system for a guide vehicle in intelligent warehousing and sorting, characterized in that: The guide vehicle path planning module includes: The guide vehicle marking unit is used for the guide vehicle to mark its own position and the target shelf position in the warehouse workspace map after receiving the sorting task; The path planning unit is used to plan the path for the guide vehicle sorting through the guide vehicle optimal path selection strategy.