A laser cleaning path planning method, system and electronic device

Abstract

This invention provides a laser cleaning path planning method, system, and electronic device, relating to the field of laser surface treatment technology. The method includes: acquiring an image of the workpiece surface to be cleaned, performing coordinate registration and binarization processing to obtain a binarized image of the workpiece surface; dividing the workpiece surface into regular partitions based on the binarized image to obtain multiple sub-partitions, and performing path coverage for each sub-partition to obtain a local path for each sub-partition; mapping the starting points of the local paths in each sub-partition to obtain a global representative point set for all local paths, and generating a traveling salesman problem (TSP) model based on this; solving the TSP model using a native heterogeneous competition-cooperative ant colony optimization algorithm to obtain the optimal global access sequence for all local paths; and finally generating a laser cleaning scanning path for the workpiece surface. This invention improves the sensitivity of path planning and ultimately generates a high-quality laser cleaning scanning path.

Description

Technical Field

[0001] This invention relates to the field of laser surface treatment technology, and more specifically, to a laser cleaning path planning method, system, and electronic device. Background Technology

[0002] As modern industry transforms towards high-end, intelligent, and green development, the requirements for surface cleanliness, structural integrity, and operational reliability of key components and core equipment are becoming increasingly stringent. Cleaning technology, as an advanced surface treatment method, with its flexible control characteristics and combined with machine vision and intelligent optimization algorithms, has been applied to multi-level precision selective cleaning. Currently, traditional ant colony algorithms are commonly used for laser cleaning path planning.

[0003] In related technologies, traditional ant colony algorithms rely excessively on hyperparameters such as pheromone heuristic factors, requiring technicians to manually configure and repeatedly debug them based on experience. This is not only highly subjective and time-consuming, but also has significant limitations in adapting hyperparameters to different scenarios. When the distribution of contaminants on the workpiece surface changes or the required cleaning precision is adjusted during laser cleaning, path planning struggles to respond quickly to dynamic conditions, resulting in low cleaning efficiency and severely hindering the large-scale application of laser intelligent cleaning technology. Summary of the Invention

[0004] The problem addressed by this invention is how to improve the sensitivity of path planning under the ant colony algorithm.

[0005] To address the above problems, this invention provides a laser cleaning path planning method, system, and electronic device.

[0006] In a first aspect, the present invention provides a laser cleaning path planning method, comprising:

[0007] The surface of the workpiece to be cleaned is imaged, and coordinate registration and binarization are performed to obtain a binarized image of the surface of the workpiece to be cleaned.

[0008] The surface of the workpiece to be cleaned is regularly partitioned according to the binarized image to obtain multiple sub-partitions, and path coverage is performed for each sub-partition to obtain the local path of each sub-partition;

[0009] Map the starting point of the local path in each sub-partition to obtain a global representative point set for all local paths, and generate a traveling salesman problem model based on the global representative point set.

[0010] The Traveling Salesman Problem model is solved using the native heterogeneous competition-cooperative ant colony optimization algorithm to obtain the optimal global access sequence for all local paths;

[0011] Based on the optimal global access sequence, a laser cleaning scanning path is generated for the surface of the workpiece to be cleaned.

[0012] Optionally, the step of imaging and acquiring the surface of the workpiece to be cleaned, performing coordinate registration and binarization processing to obtain a binarized image of the surface of the workpiece to be cleaned includes:

[0013] Digital imaging is performed on the surface of the workpiece to be cleaned to obtain the original grayscale image of the surface of the workpiece to be cleaned;

[0014] The original grayscale image is registered with the actual spatial position of the surface of the workpiece to be cleaned by using a coordinate transformation algorithm to obtain a registered grayscale image.

[0015] Adaptive thresholding or maximum inter-class variance segmentation is performed on the registered grayscale image to obtain the binarized image of the surface of the workpiece to be cleaned.

[0016] Optionally, the step of regularly partitioning the surface of the workpiece to be cleaned based on the binarized image to obtain multiple sub-partitions includes:

[0017] Connectivity analysis is performed on the foreground pixels in the binarized image to obtain multiple connected components to be cleaned;

[0018] For each of the connected components to be cleaned, extract the minimum bounding contour region of the connected component to be cleaned;

[0019] The threshold for partition size is determined based on the scanning range of the field lens of the laser cleaning device, the cleaning accuracy requirements, and the path planning efficiency requirements.

[0020] Based on the partition size threshold, the connected component to be cleaned is adaptively divided to obtain multiple sub-partitions.

[0021] Optionally, the step of performing path overlay for each of the sub-partitions to obtain the local path of each sub-partition includes:

[0022] The scanning reference direction of the local path is determined based on the shape of the sub-partition and the adaptability of the scanning direction of the laser cleaning device.

[0023] Based on the scanning reference direction and the path coverage algorithm corresponding to the shape of the sub-partition, the starting point, ending point, and path node sequence in the sub-partition are obtained;

[0024] An initial local path is generated for each sub-partition based on the starting point, the ending point, and the path node sequence in the sub-partition.

[0025] The initial local path is optimized by using a Bézier curve to obtain the local path.

[0026] Optionally, mapping the starting point of the local path in each of the sub-partitions to obtain a global representative point set for all the local paths includes:

[0027] Obtain the local coordinates of the starting point of each sub-partition in the local coordinate system of the sub-partition;

[0028] Based on the local coordinates of the starting point, determine the coordinate registration relationship between the image pixels of the sub-partition and their actual spatial positions;

[0029] Based on the coordinate registration relationship, the local coordinates of the starting point are converted into global coordinates in the global coordinate system;

[0030] The global coordinates of the starting points of all sub-partitions are summarized to form the global representative point set.

[0031] Optionally, generating a traveling salesman problem model based on the global representative point set includes:

[0032] Each of the starting points in the global representative point set is taken as a node in the traveling salesman problem model, and the Euclidean distance between any two nodes is calculated.

[0033] Using the Euclidean distance as the path cost between the two nodes, traverse all the nodes to determine the total path cost;

[0034] With the goal of minimizing the total path cost, a model of the Traveling Salesman Problem is constructed, which includes a set of nodes, a path cost matrix, and traversal constraints.

[0035] Optionally, the step of solving the Traveling Salesman Problem model using the native heterogeneous competition-cooperative ant colony optimization algorithm to obtain the optimal global access sequence for all local paths includes:

[0036] Based on the ant population size, initial pheromone concentration, pheromone volatile factor, and preset regulatory threshold and preset grouping step size coefficient of the original heterogeneous competition stage, a parameter mapping table of the expected heuristic factor is constructed.

[0037] In native heterogeneous competition, keeping the initial pheromone concentration unchanged, the expected heuristic factors are injected into the ant population in the order of the parameter mapping table, driving the ants to independently complete path construction, and obtaining the total path cost, the first-order difference and the second-order difference of the total path cost;

[0038] When the number of iterations of the ant population reaches a preset control threshold, the second-order difference is classified by support vector machine to obtain the fluctuation convergence segment, and the expected heuristic factor corresponding to the smallest first-order difference in the fluctuation convergence segment is taken as the optimal expected heuristic factor.

[0039] Based on the optimal expectation heuristic factor, and combined with the probability transition algorithm, ant colony iteration is performed to obtain the optimal global access sequence.

[0040] Optionally, generating the laser cleaning scanning path for the surface of the workpiece to be cleaned based on the optimal global access sequence includes:

[0041] Map the node order in the optimal global access sequence to the execution order of the corresponding sub-partitions;

[0042] The local paths of each sub-partition are extracted sequentially according to the execution order of the sub-partitions, and the local paths are concatenated to obtain the concatenated overall path.

[0043] Speed ​​planning is performed on the spliced ​​overall path to generate the laser cleaning scanning path.

[0044] In a second aspect, the present invention provides a laser cleaning path planning system, comprising:

[0045] The imaging acquisition and processing unit is used to acquire images of the surface of the workpiece to be cleaned, perform coordinate registration and binarization processing, and obtain a binarized image of the surface of the workpiece to be cleaned.

[0046] The local path determination unit is used to regularly partition the surface of the workpiece to be cleaned according to the binarized image to obtain multiple sub-partitions, and to perform path coverage for each sub-partition to obtain the local path of each sub-partition.

[0047] A modeling unit is used to map the starting point of the local path in each sub-partition to obtain a global representative point set of all local paths, and to generate a traveling salesman problem model based on the global representative point set.

[0048] The solving unit is used to solve the traveling salesman problem model using the native heterogeneous competition-cooperative ant colony optimization algorithm to obtain the optimal global access sequence for all the local paths;

[0049] The path generation unit is used to generate a laser cleaning scanning path for the surface of the workpiece to be cleaned based on the optimal global access sequence.

[0050] Thirdly, an electronic device according to the present invention includes a memory and a processor;

[0051] The memory is used to store computer programs;

[0052] The processor is configured to implement the laser cleaning path planning method as described above when executing the computer program.

[0053] The laser cleaning path planning method, system, and electronic device of this invention, through imaging the surface of the workpiece to be cleaned, performing coordinate registration and binarization processing, can accurately obtain the distribution information of contaminants on the workpiece surface. Coordinate registration ensures accurate correspondence between the image and the actual workpiece surface position, and binarization processing simplifies the image into contaminant and non-contaminant areas, highlighting the key areas for cleaning and providing a clear basis for subsequent zoning and path planning. Based on the binarized image, the workpiece surface is regularly partitioned to obtain multiple sub-regions, and path coverage is performed for each sub-region to obtain local paths. Dividing the workpiece surface into multiple sub-regions helps to decompose the complex cleaning task into several relatively simple sub-tasks. This decomposition method reduces the complexity of the problem, while generating local paths for each sub-region ensures that contaminants in each area can be effectively cleaned. For example, for workpiece surfaces with irregular shapes and uneven contaminant distribution, regular partitioning can divide them into multiple easily manageable small areas. The local paths within each small area can be flexibly designed according to the contamination situation of that area, improving the targeting and efficiency of cleaning. By mapping the starting point of the local path to a global representative point, the problem is transformed into a traveling salesman problem. The optimal global access sequence is found using the traveling salesman problem solution method. This invention makes full use of the optimization theory and algorithm of the traveling salesman problem to provide an effective framework for solving the laser cleaning path planning problem.

[0054] The Traveling Salesman Problem (TSP) model is solved using a native heterogeneous competition-cooperative ant colony optimization algorithm to obtain the optimal global access sequence for all local paths. Compared to traditional ant colony optimization, the native heterogeneous competition-cooperative ant colony optimization algorithm reduces excessive dependence on hyperparameters by introducing a competition-cooperation mechanism. During the algorithm's operation, different ant colonies can compete and cooperate with each other, sharing information, thus converging to the optimal solution more quickly. This algorithm can quickly replan the path under dynamic conditions such as changes in the distribution of contaminants on the workpiece surface or adjustments to the cleaning accuracy requirements, improving the sensitivity of laser cleaning path planning. For example, when new contaminated areas appear on the workpiece surface or the contamination level in certain areas increases, the algorithm can adjust the global access sequence in a timely manner to ensure that the cleaning path can adapt to the new situation, thereby improving cleaning efficiency. Based on the optimal global access sequence, a laser cleaning scanning path for the workpiece surface to be cleaned is generated. The final generated laser cleaning scanning path comprehensively considers the local paths of each sub-region and the globally optimal access order, enabling efficient and accurate laser cleaning. This path planning method ensures that the laser cleaning equipment visits each contaminated area in the optimal order during the cleaning process, reducing unnecessary movement of the equipment, improving cleaning efficiency, and solving the problems of low sensitivity and difficulty in adapting to dynamic working conditions in traditional ant colony algorithm path planning. This has powerfully promoted the large-scale application of laser intelligent cleaning technology.

[0055] In summary, this invention effectively addresses the shortcomings of traditional ant colony algorithms in dynamic path planning, significantly improves the sensitivity of path planning, and enables the entire laser cleaning path planning method to efficiently adapt to different working conditions, ultimately generating high-quality laser cleaning scanning paths. Attached Figure Description

[0056] Figure 1 This is a flowchart illustrating the laser cleaning path planning method in an embodiment of the present invention.

[0057] Figure 2 This is a schematic diagram of the target feature binarization image in an embodiment of the present invention;

[0058] Figure 3 This is a schematic diagram of the target partition and local path coverage in an embodiment of the present invention;

[0059] Figure 4 This is a schematic diagram of parameter competition curves under native heterogeneous regulation of ant colonies in an embodiment of the present invention;

[0060] Figure 5 This is a schematic diagram of the characteristic curve of the differential change in an embodiment of the present invention;

[0061] Figure 6 This is a schematic diagram of the hyperparameter extraction results in an embodiment of the present invention;

[0062] Figure 7 This is a schematic diagram of the dynamic adaptive optimization curve based on the native heterogeneous competition-cooperation mechanism of ant colonies in an embodiment of the present invention;

[0063] Figure 8 This is a schematic diagram illustrating the global adaptive planning effect in an embodiment of the present invention;

[0064] Figure 9 This is a schematic diagram of the laser cleaning path planning system in an embodiment of the present invention;

[0065] Figure 10 This is a schematic diagram of the structure of an electronic device in an embodiment of the present invention. Detailed Implementation

[0066] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Although some embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the present invention. It should be understood that the accompanying drawings and embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of protection of the present invention.

[0067] It should be understood that the various steps described in the method embodiments of the present invention may be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present invention is not limited in this respect.

[0068] The term "comprising" and its variations as used herein are open-ended, meaning "including but not limited to"; the term "based on" means "at least partially based on"; the term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments"; and the term "optionally" means "optional embodiments". Definitions of other terms will be given in the following description. It should be noted that the concepts of "first," "second," etc., mentioned in this invention are used only to distinguish different devices, modules, or units, and are not intended to limit the order of functions performed by these devices, modules, or units or their interdependencies.

[0069] It should be noted that the terms "a" and "a plurality of" used in this invention are illustrative rather than restrictive. Those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0070] The names of the messages or information exchanged between the multiple devices in the embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of these messages or information.

[0071] Combination Figure 1 As shown in the figure, an embodiment of the present invention provides a laser cleaning path planning method, comprising:

[0072] The surface of the workpiece to be cleaned is imaged, and coordinate registration and binarization are performed to obtain a binarized image of the surface of the workpiece to be cleaned.

[0073] Specifically, the workpiece to be cleaned, such as an aircraft workpiece with irregular paint residue, is first fixed in a fixed state using tooling fixtures to ensure no displacement deviation on the workpiece surface, providing a stable reference for subsequent operations. Then, the image acquisition module, coaxially mounted with the paint removal laser, is activated. This module performs a digital image acquisition on the substrate area covered by the laser scanning area on the workpiece surface. During the acquisition process, the image coordinates are simultaneously matched with the actual spatial coordinates of the workpiece, clarifying the correspondence between image pixels and the physical positions on the workpiece surface. In an optional embodiment of this invention, the scanning unit size is set to be consistent with the pixel unit size, meaning the discrete coordinate system of the path points is equivalent to the image pixel coordinate system, achieving coordinate registration.

[0074] After coordinate registration, the acquired raw images are converted to grayscale and filtered for noise reduction. Then, an adaptive threshold segmentation technique is used to generate a binarized image. Areas requiring cleaning (e.g., residual paint on aircraft surfaces) are marked in white, while areas of the workpiece substrate that do not require cleaning (e.g., exposed substrate with a damaged surface) are marked in black, clearly distinguishing between areas to be cleaned and those not. Finally, the edge vertices of the binarized image are calibrated, and the spatial coordinate information of the image matrix is ​​integrated. The processed binarized image data is then uploaded to the local path heuristic module, providing a location basis for subsequent partitioning and path planning.

[0075] The surface of the workpiece to be cleaned is regularly partitioned according to the binarized image to obtain multiple sub-partitions, and path coverage is performed for each sub-partition to obtain the local path of each sub-partition.

[0076] Specifically, after receiving the binarized image data, the local path heuristic module, based on geometric principles and the contour features of the area to be cleaned, uses connected component analysis to regularly partition the white area to be cleaned in the binarized image: dividing the continuous white connected components into multiple independent sub-partitions with regular contours. Each sub-partition can be planned independently, ensuring that subsequent path coverage is complete and without overlap. Simultaneously, for each sub-partition, an inner spiral heuristic is used for local path filling. Specifically, starting from the geometric center of the sub-partition, the path extends outwards along a spiral trajectory until it covers the entire area to be cleaned in the sub-partition, thus ensuring the smoothness and cleaning efficiency of the local path and avoiding resource waste caused by path duplication. After completing the local path planning for all sub-partitions, the local path heuristic module integrates the local path information of each sub-partition into an indexable data packet and uploads it to the global ant colony adaptive path optimization module. In this embodiment of the invention, the local path information includes the partition index, path start / end coordinates, and path trajectory data.

[0077] The starting point of the local path in each sub-partition is mapped to obtain a global representative point set for all local paths, and a traveling salesman problem model is generated based on the global representative point set.

[0078] Specifically, after receiving the local path data packet, the global ant colony adaptive path optimization module first analyzes the local paths of each sub-partition, defining the starting point of the local path within each sub-partition as a global planning relay node. This node represents the core position of the corresponding sub-partition in the workpiece's global coordinate system. Then, through coordinate mapping technology, all relay nodes of all sub-partitions are transformed into global representative points, and all global representative points are integrated to form a global representative point set, where each global representative point uniquely corresponds to one sub-partition.

[0079] Based on the spatial distribution characteristics of the global representative point set, this embodiment transforms the sub-partition global access order optimization problem into a Traveling Salesman Problem (TSP) model. This involves finding a path that traverses all global representative points, minimizing the total path length to reduce cleaning time and resource consumption, and ensuring each global representative point is traversed only once to avoid redundant cleaning. This transformation solves the global path ordering problem between sub-partitions while avoiding the path cost degradation problem caused by parameter sensitivity in path ordering in traditional ant colony algorithms.

[0080] The Traveling Salesman Problem model is solved using the native heterogeneous competition-cooperative ant colony optimization algorithm to obtain the optimal global access sequence for all local paths.

[0081] Specifically, the native heterogeneous competition-cooperative ant colony optimization algorithm is used to solve the TSP model, and the process is divided into three stages.

[0082] Specifically, the first stage, the native heterogeneous competition stage, involves setting a control threshold (e.g., 60 observations) and a grouping step size coefficient (e.g., 20 groups of 3 observations each). An incremental parameter mapping table is constructed within a preset parameter range (e.g., 0.01-20), and different parameters are assigned to each of the 20 groups (this parameter affects the ant colony's search strategy). In this stage, the pheromone positive feedback update and iterative coverage mechanism in the ant colony algorithm are temporarily suppressed, and the path search performance of the ant colony is observed only by adjusting the parameters: when the parameters are too large, the ant colony tends to choose paths based on fixed experience, easily getting trapped in local optima; when the parameters are too small, the randomness of the ant colony search increases, the path length increases, and the results fluctuate significantly. Simultaneously, the changing trends of the ant colony's path search results are analyzed, dividing the path search state into three intervals: an oscillation zone, a rapid convergence zone, and a stable zone. In the oscillation zone, the path length fluctuates dramatically; in the rapid convergence zone, the path length shortens rapidly; and in the stable zone, the path length tends to stabilize. The first part of the stable zone corresponds to the parameter range where the ant colony's search effect is better, providing a basis for subsequent hyperparameter extraction.

[0083] The second stage, the hyperparameter extraction stage, employs a Support Vector Machine (FFD-SVM) model with integrated feedforward discrete passband diagnostics to analyze the path search data and parameter features obtained in the previous competition stage. The FFD-SVM model classifies and predicts the parameter features, filtering out effective parameters. Combined with feedforward discrete passband diagnostics, the prediction results are further optimized, classification errors are reduced, and the hyperparameters that maximize the ant colony search performance are accurately located, such as the final determined optimal parameter values. This stage enables high-precision parameter extraction under small sample conditions, ensuring the efficiency of subsequent ant colony algorithm optimization.

[0084] The third stage, the ant colony cooperative pathfinding stage, activates the pheromone positive feedback update mechanism in the ant colony algorithm. The extracted optimal hyperparameters regulate the global search behavior of the ants. When traversing global representative points, the ants release pheromones along the paths they pass. Subsequent ants tend to choose paths with higher pheromone concentrations. Simultaneously, the pheromones gradually evaporate during the iteration process, preventing the path from getting trapped in local optima. A total number of iterations is set (e.g., 300, including observations during the competition phase and iterations during the cooperation phase). After multiple iterations, the ant colony converges to the optimal path solution, obtaining the optimal traversal order of global representative points for all sub-partitions, i.e., the optimal global access sequence of local paths in each sub-partition.

[0085] Based on the optimal global access sequence, a laser cleaning scanning path is generated for the surface of the workpiece to be cleaned.

[0086] Specifically, the global ant colony adaptive path optimization module merges the obtained optimal global access sequence with the local paths of each sub-partition generated by the local path heuristic module, and sequentially concatenates each sub-partition according to the order of the optimal global access sequence. At the junctions of adjacent sub-partitions, a smooth transition technique (such as curved transition) is used to eliminate path inflection points, ensuring the continuity of the laser motion trajectory and avoiding the decrease in cleaning accuracy caused by frequent laser head starts and stops. After fusion, the module integrates the final laser cleaning scanning path data into a data package and uploads it to the subsequent laser execution control unit. The laser cleaning scanning path data includes the sub-partition access order, local path trajectory coordinates, and sub-partition transition path coordinates. Based on this path data, the laser execution control unit drives the laser head to perform cleaning operations on the surface of the workpiece to be cleaned according to the planned path, achieving efficient and precise intelligent laser cleaning.

[0087] The laser cleaning path planning method in this embodiment accurately acquires the distribution information of contaminants on the workpiece surface by imaging the surface, performing coordinate registration, and binarization. Coordinate registration ensures accurate correspondence between the image and the actual workpiece surface position, while binarization simplifies the image into contaminant and non-contaminant areas, highlighting key cleaning areas and providing a clear basis for subsequent partitioning and path planning. The workpiece surface is regularly partitioned based on the binarized image, resulting in multiple sub-partitions. Path coverage is then applied to each sub-partition to obtain local paths. Dividing the workpiece surface into multiple sub-partitions helps decompose the complex cleaning task into several relatively simple sub-tasks. This decomposition reduces the complexity of the problem, while generating local paths for each sub-partition ensures that contaminants in each area are effectively cleaned. For example, for workpiece surfaces with irregular shapes and uneven contaminant distribution, regular partitioning can divide them into multiple easily manageable small areas. The local paths within each small area can be flexibly designed according to the contamination situation of that area, improving the targeting and efficiency of the cleaning process. By mapping the starting point of the local path to a global representative point, the problem is transformed into a traveling salesman problem. The optimal global access sequence is found using the traveling salesman problem solution method. This invention makes full use of the optimization theory and algorithm of the traveling salesman problem to provide an effective framework for solving the laser cleaning path planning problem.

[0088] The Traveling Salesman Problem (TSP) model is solved using a native heterogeneous competition-cooperative ant colony optimization algorithm to obtain the optimal global access sequence for all local paths. Compared to traditional ant colony optimization, the native heterogeneous competition-cooperative ant colony optimization algorithm reduces excessive dependence on hyperparameters by introducing a competition-cooperation mechanism. During the algorithm's operation, different ant colonies can compete and cooperate with each other, sharing information, thus converging to the optimal solution more quickly. This algorithm can quickly replan the path under dynamic conditions such as changes in the distribution of contaminants on the workpiece surface or adjustments to the cleaning accuracy requirements, improving the sensitivity of laser cleaning path planning. For example, when new contaminated areas appear on the workpiece surface or the contamination level in certain areas increases, the algorithm can adjust the global access sequence in a timely manner to ensure that the cleaning path can adapt to the new situation, thereby improving cleaning efficiency. Based on the optimal global access sequence, a laser cleaning scanning path for the workpiece surface to be cleaned is generated. The final generated laser cleaning scanning path comprehensively considers the local paths of each sub-region and the globally optimal access order, enabling efficient and accurate laser cleaning. This path planning method ensures that the laser cleaning equipment visits each contaminated area in the optimal order during the cleaning process, reducing unnecessary movement of the equipment, improving cleaning efficiency, and solving the problems of low sensitivity and difficulty in adapting to dynamic working conditions in traditional ant colony algorithm path planning. This has powerfully promoted the large-scale application of laser intelligent cleaning technology.

[0089] In summary, this embodiment effectively addresses the shortcomings of traditional ant colony algorithms in dynamic path planning, greatly improves the sensitivity of path planning, and enables the entire laser cleaning path planning method to efficiently adapt to different working conditions, ultimately generating high-quality laser cleaning scanning paths.

[0090] Optionally, the step of imaging and acquiring the surface of the workpiece to be cleaned, performing coordinate registration and binarization processing to obtain a binarized image of the surface of the workpiece to be cleaned includes:

[0091] Digital imaging is performed on the surface of the workpiece to be cleaned to obtain the original grayscale image of the surface of the workpiece to be cleaned;

[0092] The original grayscale image is registered with the actual spatial position of the surface of the workpiece to be cleaned by using a coordinate transformation algorithm to obtain a registered grayscale image.

[0093] Adaptive thresholding or maximum inter-class variance segmentation is performed on the registered grayscale image to obtain the binarized image of the surface of the workpiece to be cleaned.

[0094] Specifically, workpieces with randomly irregularly shaped contaminants, such as residual paint layers on aircraft surfaces or areas covered by industrial oil stains, are selected for maintenance. These workpieces are then adjusted to a horizontal and fixed state using high-precision tooling fixtures to ensure that the workpiece surface remains stable and free from displacement or shaking during subsequent imaging, preventing image acquisition from deviating from the actual position due to workpiece offset. In a preferred embodiment of the invention, the imaging parameters of the graphics acquisition module, coaxially mounted with the paint removal laser, are calibrated. The imaging resolution is set to match the laser scanning field, ensuring that the pixel unit size of the acquired image is equivalent to the laser scanning unit size. This means that in subsequent path planning, the discrete coordinate system of the path points is completely consistent with the image pixel coordinate system. Simultaneously, the exposure time and focal length of the module are calibrated to ensure a clear image that accurately distinguishes between contaminant and substrate areas on the workpiece surface. By controlling the calibrated graphics acquisition module, the focus is placed on the substrate area corresponding to the laser scanning field on the workpiece surface to be cleaned—the area actually covered by the subsequent laser cleaning operation—and a complete digital imaging acquisition is performed to obtain an original color image containing the distribution of contaminants on the workpiece surface and details of the substrate contour. The acquired original color images are preprocessed to grayscale. A weighted average method (e.g., calculating grayscale values ​​according to RGB channel weights of 0.299, 0.587, and 0.114) is used to convert the color images into original grayscale images. In this process, the brightness difference characteristics between dirt and substrate in the image are preserved, color interference is removed, and the data dimension is simplified for subsequent coordinate registration and threshold segmentation. At the same time, the original grayscale images are denoised using a Gaussian filtering algorithm to eliminate noise caused by ambient light fluctuations and sensor noise, ensuring that the edge contours of dirt and substrate in the image are clearly distinguishable.

[0095] Next, on the fixed surface of the workpiece to be cleaned, several distinctive edge vertices are selected as coordinate reference points, such as corner vertices of the workpiece substrate and boundary feature vertices of the contaminant area. High-precision displacement sensors are used to measure the three-dimensional coordinates of these reference points in the actual spatial coordinate system of the workpiece, i.e., the physical position coordinates. Simultaneously, the pixel coordinates corresponding to these reference points are located and recorded in the original grayscale image, i.e., the two-dimensional coordinates in the image coordinate system. A perspective transformation coordinate conversion algorithm is used, based on the calibrated reference point actual spatial coordinates-pixel coordinate correspondence group, to construct a one-to-one mapping relationship between the pixel coordinate system of the original grayscale image and the actual spatial coordinate system of the workpiece surface to be cleaned. Through this mapping relationship, the coordinates of any pixel in the original grayscale image are directly converted into the actual physical position coordinates of the workpiece surface, achieving accurate registration between the original grayscale image and the actual spatial position of the workpiece, resulting in a registered grayscale image. This registration ensures that the subsequent image-based path planning coordinates can be directly applied to the laser cleaning execution unit, avoiding path offset.

[0096] The registered grayscale image is then smoothed again (e.g., by median filtering) to further eliminate the impact of residual noise on the segmentation results. This ensures that the grayscale difference between the areas of dirt to be cleaned and the areas of the substrate to be cleaned is more significant, laying the foundation for accurate segmentation. If there is uneven local illumination on the surface of the workpiece to be cleaned, such as uneven grayscale distribution caused by partial occlusion, an adaptive threshold segmentation algorithm is further adopted. The algorithm dynamically adjusts the segmentation threshold according to the average grayscale value of different local areas of the image, ensuring that dirt and substrate can be accurately distinguished in each local area. For example, the segmentation threshold is lowered in darker areas and raised in brighter areas to avoid some dirt areas being missed or misclassified due to differences in illumination. If the surface of the workpiece to be cleaned is uniformly illuminated and the grayscale difference between the contaminant area and the substrate area is generally stable (e.g., the grayscale of the contaminant is generally higher or lower than that of the substrate), then the Otsu algorithm (maximum inter-class variance segmentation algorithm) is used. The algorithm iterates through all possible grayscale thresholds, calculates the inter-class variance between the contaminant pixel class and the substrate pixel class at each threshold, and selects the threshold with the largest inter-class variance as the optimal segmentation threshold, thereby achieving accurate segmentation in the global range.

[0097] After segmentation, areas in the image identified as dirt to be cleaned are marked in white (if the pixel value is set to 255), and areas identified as substrates that do not need cleaning are marked in black (if the pixel value is set to 0). This generates a binarized image of the workpiece surface to be cleaned. Subsequently, the edge contours of the binarized image are recalibrated, and the spatial coordinate information corresponding to the image matrix is ​​integrated. The binarized image data is then uploaded to the local path heuristic module of the laser paint removal intelligent planning system in the form of an indexable data packet, providing a clear regional location basis for the subsequent regular partitioning of the area to be cleaned and local path planning.

[0098] In this optional embodiment, by directional digital imaging of the workpiece surface and converting it into an original grayscale image, the brightness difference characteristics between the contaminated area and the substrate area can be effectively preserved. At the same time, environmental interference is eliminated through noise reduction processing, laying a data foundation for subsequent accurate differentiation of the area to be cleaned. Secondly, by using a coordinate transformation algorithm to register the original grayscale image with the actual spatial position of the workpiece, and by setting the pixel unit size to be equivalent to the laser scanning unit size, a one-to-one mapping relationship between the image pixel coordinates and the physical coordinates of the workpiece is established, eliminating the deviation between the image and the actual position, and ensuring that the path coordinates planned based on the image can be directly and accurately mapped to the workpiece surface, avoiding missed cleaning or incorrect cleaning caused by coordinate misalignment during the laser cleaning process. Finally, by flexibly selecting adaptive threshold segmentation based on the workpiece surface illumination conditions to handle local uneven illumination scenarios or maximum inter-class variance segmentation to handle global grayscale difference stable scenarios, it can achieve efficient and accurate differentiation between the dirt-to-be-cleaned area and the non-cleaned substrate area. The generated binarized image can clearly and intuitively present the outline and position of the area to be cleaned, providing accurate and reliable regional location basis for the subsequent regular partitioning and path filling of the local path heuristic module and the global planning of the global ant colony adaptive path optimization module. This effectively ensures the accuracy and effectiveness of laser cleaning path planning, and provides key technical support for reducing the time cost and resource consumption of the cleaning process and improving the overall efficiency of laser intelligent cleaning.

[0099] Optionally, the step of regularly partitioning the surface of the workpiece to be cleaned based on the binarized image to obtain multiple sub-partitions includes:

[0100] Connectivity analysis is performed on the foreground pixels in the binarized image to obtain multiple connected components to be cleaned;

[0101] For each of the connected components to be cleaned, extract the minimum bounding contour region of the connected component to be cleaned;

[0102] The threshold for partition size is determined based on the scanning range of the field lens of the laser cleaning device, the cleaning accuracy requirements, and the path planning efficiency requirements.

[0103] Based on the partition size threshold, the connected component to be cleaned is adaptively divided to obtain multiple sub-partitions.

[0104] Specifically, after receiving the binarized image data, the local path heuristic module first performs connected component analysis on the white foreground pixels in the image, i.e. the areas of dirt to be cleaned. Using an eight-neighbor connectivity judgment algorithm, it traverses all pixels in the binarized image and groups adjacent (vertical, horizontal, and diagonal) white foreground pixels into the same connected unit, thereby separating all independent connected components to be cleaned in the image. For example, in the aircraft paint cleaning scenario, multiple independent residual paint areas scattered on the workpiece surface each form a connected component to be cleaned. At the same time, it records the pixel coordinate range and number of pixels of each connected component to be cleaned, clarifying the spatial distribution and size of each connected component. For each separated connected component to be cleaned, the local path heuristic module extracts the edge features of the connected component using a contour detection algorithm (such as Canny edge detection combined with contour fitting). Then, based on the principle of geometric fitting, it generates the minimum bounding contour region of the connected component. If the connected component is irregular in shape (such as an arc or a polygonal residual paint layer), it fits and generates a minimum bounding shape that can completely enclose the connected component and has a regular contour. In a preferred embodiment of the present invention, it is preferably a rectangle, and in special scenarios, it can be a polygon. This ensures that the bounding contour region only contains the connected component to be cleaned and a very small number of matrix edge pixels, which avoids missing the cleaned area and reduces the inclusion of invalid matrix areas. At the same time, the vertex coordinates of each minimum bounding contour region are recorded. In this embodiment, the vertex coordinates mainly refer to the nine-point calibration in the camera coordinate calibration process, including the vertices and midpoints of the cross, to achieve alignment between the pixel coordinate system and the scanning coordinate system, thereby providing a substantial matching coordinate basis for the laser focal field scanning and forming regularized region boundary data.

[0105] By calling the preset parameter configuration interface through the local path heuristic module, and combining the core performance parameters of the laser cleaning device with actual operational requirements, the partition size threshold is determined. Specifically, the scanning range of the laser cleaning device's field lens is used as a basic constraint, such as the maximum horizontal / vertical dimension of a single effective scan of the field lens, to ensure that the maximum size of a single sub-partition does not exceed the coverage range of a single scan of the field lens, avoiding path connection errors caused by crossing the field lens range. Combined with cleaning accuracy requirements, such as micron-level cleaning accuracy scenarios, a lower limit for the partition size is set, such as the minimum side length of the partition not being less than 5 times the diameter of the laser spot, to prevent the partition from being too small, resulting in overly fragmented local paths that affect cleaning continuity and accuracy. At the same time, the path planning efficiency requirements are taken into account. Through historical data calibration, such as cleaning efficiency tests of the same workpiece type, the threshold is adjusted to a balance range between the time taken to fill the path of a single partition and the time taken to connect partitions, avoiding excessive partitions that increase the complexity of global path sorting. Finally, the final determined partition size threshold is stored in the form of maximum side length × minimum side length as the judgment standard for subsequent adaptive partitioning.

[0106] Finally, the minimum bounding contour region size of each connected domain to be cleaned is compared with the determined partition size threshold using the local path heuristic module, and adaptive partitioning is performed. Specifically, if the minimum bounding contour region size of the connected domain to be cleaned is less than or equal to the partition size threshold, such as the side length of the bounding rectangle of a small independent residual paint layer being less than the field lens scanning range, then the minimum bounding contour region is directly treated as an independent sub-partition without additional partitioning. If the minimum bounding contour region size of the connected domain to be cleaned is greater than the partition size threshold, such as the side length of the bounding rectangle of a large continuous residual paint layer exceeding the field lens scanning range, then the bounding contour region is partitioned using the partition size threshold as a benchmark, and the method of grid equal division + edge adaptation is adopted. First, a virtual grid is constructed according to the threshold size, and the bounding contour region is divided into several standard sub-partitions with a size equal to the threshold. Then, the non-standard size regions remaining at the edge, which are smaller than the threshold but larger than the lower limit of the size, are separately divided into one adaptive sub-partition, ensuring that the size of all sub-partitions is within the partition size threshold range.

[0107] After partitioning, a unique index number is assigned to each sub-partition, and the vertex coordinates and original connected component identifiers of each sub-partition are recorded to form a regularized sub-partition dataset, which provides structured regional units for subsequent sub-partition local path filling and global ant colony adaptive path optimization.

[0108] In this optional embodiment, by performing connected component analysis on the foreground pixels in the binarized image, multiple independent areas to be cleaned scattered on the workpiece surface can be accurately separated, such as residual paint layers and oil stains in different locations, avoiding mutual interference between different areas to be cleaned in subsequent partitioning and path planning; at the same time, the minimum bounding contour region is extracted for each connected component to be cleaned, transforming irregularly shaped dirt areas into region units with regular boundaries, eliminating the influence of chaotic shapes on the partitioning standardization, and providing a unified regional benchmark for subsequent standardized division; the partitioning size threshold is determined by combining the scanning range of the laser cleaning device's field lens, cleaning accuracy requirements, and path planning efficiency requirements, so that the threshold has both design and implementation requirements. It ensures adaptability and operational practicality; based on this threshold, the connected components are adaptively divided to ensure that all sub-partitions are reasonably sized. This avoids the increased difficulty of path planning caused by overly large partitions, and also prevents frequent path connection problems caused by overly small partitions. The resulting regular sub-partitions not only provide ordered and controllable regional units for subsequent local path filling in sub-partitions, but also lay a structured foundation for the sub-partition indexing and global path planning of the global ant colony adaptive path optimization module. This effectively improves the standardization, accuracy, and efficiency of laser cleaning path planning, while ensuring the continuity and accuracy of subsequent cleaning operations, and reducing the planning complexity and time cost of the intelligent cleaning process.

[0109] Optionally, the step of performing path overlay for each of the sub-partitions to obtain the local path of each sub-partition includes:

[0110] The scanning reference direction of the local path is determined based on the shape of the sub-partition and the adaptability of the scanning direction of the laser cleaning device.

[0111] Based on the scanning reference direction and the path coverage algorithm corresponding to the shape of the sub-partition, the starting point, ending point, and path node sequence in the sub-partition are obtained;

[0112] An initial local path is generated for each sub-partition based on the starting point, the ending point, and the path node sequence in the sub-partition.

[0113] The initial local path is optimized by using a Bézier curve to obtain the local path.

[0114] Specifically, the shape data of each sub-partition is read through the local path heuristic module. The shape data, such as regular contours like rectangles and polygons, along with the coordinate range, is combined with the scanning direction characteristics of the laser cleaning device, such as the smoothness of the laser head's lateral / vertical scanning movement and the difference in turning energy consumption, to determine the scanning reference direction of the local path. If the sub-partition is rectangular, the direction parallel to the long side of the sub-partition is preferentially selected as the reference direction, thereby reducing the number of laser head turns and improving scanning efficiency. If the sub-partition is an irregular polygon, the longest axis direction of the sub-partition is extracted through geometric analysis and set as the reference direction, thereby ensuring the continuity of the laser movement trajectory during path coverage. At the same time, reference directions that conflict with the laser head's movement limit direction are avoided, such as avoiding frequent triggering of the laser head's movement stroke boundary, ensuring that the reference direction is both adapted to the shape of the sub-partition and conforms to the movement characteristics of the laser device.

[0115] Based on the determined scanning reference direction, the local path heuristic module calls the preset path coverage algorithm. In this optional embodiment, the inner spiral heuristic is used to perform path coverage calculation for each sub-partition. The geometric center of the sub-partition is used as the starting point of the path to ensure coverage from the center of the sub-partition to the edge, reducing the risk of edge omission. The path extends outward spirally along the scanning reference direction at a preset step size, where the preset step size is equal to the laser spot diameter, to ensure gapless coverage. During the extension process, the boundary intersections between the path and the sub-partition contour are detected in real time, and the intersection point that finally reaches the edge of the sub-partition is set as the termination point. At the same time, the coordinates of all turning points and boundary intersections during the path extension process are recorded to form a path node sequence containing the starting point, termination point, and intermediate turning points. If the sub-partition has a special shape, such as a concave contour, the non-cleaned part of the concave area is automatically avoided when the path node sequence is generated to ensure that the node sequence only covers the area to be cleaned in the sub-partition.

[0116] The local path heuristic module connects all nodes in the path node sequence with straight lines in the order of starting point, intermediate turning point, and then ending point to form the initial local path for each sub-partition. At the same time, during the connection process, it verifies in real time whether the path is completely within the sub-partition. If the path exceeds the sub-partition outline, such as due to calculation deviation of edge nodes, the coordinates of the corresponding nodes are automatically adjusted to the sub-partition boundary to ensure that the initial local path completely covers the area to be cleaned in the sub-partition and that there are no redundant paths, thus providing a basic path framework for subsequent path smoothing optimization.

[0117] Furthermore, in this embodiment, to eliminate inflection points formed by straight-line connections between nodes in the initial local path and to avoid decreased cleaning accuracy and increased energy consumption due to frequent sharp turns of the laser head, the local path heuristic module uses Bézier curves to perform transition optimization on the initial local path. Specifically, for two adjacent nodes in the initial path, especially the intersection of turning points and boundaries, the tangent direction between the nodes is calculated, and a third-order Bézier curve is constructed as a transition segment. By adjusting the control point parameters of the Bézier curve, the transition segment is smoothly connected to the straight-line segments of adjacent nodes, ensuring that the curvature of the optimized path is continuous and without obvious inflection points. After optimization, the final local path of each sub-partition is generated, and the starting point coordinates, ending point coordinates, complete path trajectory data, and sub-partition index of the local path are associated and stored, and uploaded to the global ant colony adaptive path optimization module in the form of an indexable data packet, providing data support for the subsequent concatenation of global paths and the precise execution of laser cleaning actions.

[0118] In this optional embodiment, by determining the scanning reference direction based on the compatibility between the sub-partition shape and the scanning direction of the laser cleaning device, the local path planning can be precisely matched with the laser head motion characteristics, reducing the number of invalid laser head turns and laying the foundation for improving scanning efficiency. By combining the path coverage algorithm corresponding to the scanning reference direction and the sub-partition shape to obtain the starting point, ending point, and path node sequence, it can be ensured that the path nodes completely cover the area to be cleaned in the sub-partition, avoiding missed cleaning or redundant paths due to missing nodes. When generating the initial local path based on the starting point, ending point, and path node sequence, the integrity and accuracy of the initial path can be further guaranteed by verifying whether the path is within the sub-partition range and adjusting the deviation nodes in real time, eliminating invalid paths that exceed the sub-partition. Finally, the initial local path is transitionally optimized using Bézier curves to eliminate the inflection points formed by straight connections between nodes, making the curvature of the optimized local path continuous. This effectively avoids the decrease in cleaning accuracy and increase in energy consumption caused by frequent sharp turns of the laser head. It provides a smooth local path basis for the subsequent sub-partition path concatenation of the global ant colony adaptive path optimization module, and also improves the continuity, accuracy, and energy efficiency of laser cleaning operations, reducing path planning errors and resource consumption in the intelligent cleaning process.

[0119] Optionally, mapping the starting point of the local path in each of the sub-partitions to obtain a global representative point set for all the local paths includes:

[0120] Obtain the local coordinates of the starting point of each sub-partition in the local coordinate system of the sub-partition;

[0121] Based on the local coordinates of the starting point, determine the coordinate registration relationship between the image pixels of the sub-partition and their actual spatial positions;

[0122] Based on the coordinate registration relationship, the local coordinates of the starting point are converted into global coordinates in the global coordinate system;

[0123] The global coordinates of the starting points of all sub-partitions are summarized to form the global representative point set.

[0124] Specifically, the global ant colony adaptive path optimization module calls the sub-partition data packets uploaded by the local path heuristic module to extract the starting point information of the local path for each sub-partition. This starting point is the geometric center of the sub-partition or the optimal starting position that fits the shape of the sub-partition, as determined during the local path planning stage. It also carries the local coordinates of the starting point in the local coordinate system of the sub-partition. The local coordinate system of the sub-partition takes the lower left corner vertex of its smallest bounding contour as the origin and the axis parallel to the scanning reference direction of the sub-partition as the horizontal and vertical axes, with the unit consistent with the image pixel coordinate system. The module reads and stores the local coordinates of the starting point one by one according to the sub-partition index order, such as the unique number assigned during partitioning, to form a corresponding list of sub-partition index and starting point local coordinates, ensuring that the starting point of each sub-partition has a clear local location identifier.

[0125] The global ant colony adaptive path optimization module retrieves two types of coordinate relationships from the system's coordinate database, established by the graphics acquisition module during the imaging acquisition stage of the workpiece surface to be cleaned. The first relationship is the registration relationship between the image pixel coordinate system and the actual spatial coordinate system of the workpiece. Specifically, this is constructed based on a perspective transformation algorithm, including the size mapping ratio between pixel units and actual spatial units. The pixel unit size is pre-defined as equivalent to the laser scanning unit size, with a ratio of 1:1, meaning one pixel unit corresponds to one actual scanning unit. The second relationship is the association between the sub-region local coordinate system and the image pixel coordinate system. This includes the pixel coordinates of the origin of the sub-region local coordinate system in the image pixel coordinate system, and the angle between the local coordinate system axis and the pixel coordinate system axis. Since the sub-region is a regular sub-region with a 0° angle, only the origin pixel coordinates need to be retrieved. The module integrates these two relationships into a complete mapping link: sub-region local coordinates - image pixel coordinates - actual workpiece spatial coordinates. This serves as the basis for subsequent coordinate transformation calculations, ensuring a deviation-free transformation process.

[0126] Based on the coordinate registration relationship, the global ant colony adaptive path optimization module performs a two-step transformation on the local coordinates of the starting point of each sub-partition. Specifically, the first step is to convert the local coordinates of the starting point into image pixel coordinates. The pixel coordinates are obtained by superimposing the local coordinates of the starting point with the pixel coordinates of the origin of the sub-partition's local coordinate system. It should be noted that since the directional angle is 0°, no additional rotation calculation is required. The second step is to convert the image pixel coordinates into global coordinates in the global coordinate system, i.e., the actual spatial coordinate system of the workpiece to be cleaned. Based on the 1:1 size mapping ratio between the pixel coordinate system and the actual spatial coordinate system, the pixel coordinates are directly used as the global coordinates of the starting point. In one embodiment of the present invention, if there is a slight calibration deviation in the global coordinates, it can be fine-tuned by a preset correction coefficient to ensure complete consistency with the actual position of the workpiece. After the transformation is completed, the module records the global coordinates of the starting point corresponding to each sub-partition index, forming the associated data of sub-partition index-starting point global coordinates. The global ant colony adaptive path optimization module aggregates and integrates the global coordinates of the starting points of all sub-partitions according to the sub-partition index order. Each starting point's global coordinates is treated as a global representative point, uniquely corresponding to a sub-partition. This point represents the core position of the sub-partition in the actual workpiece space. Each global representative point is assigned an index identifier consistent with the sub-partition, ultimately forming a structured dataset containing global representative point coordinates and sub-partition indices—the global representative point set. In one embodiment of this invention, the representative point is based on each local sub-partition, treating each independent sub-partition as a sequential child node under global planning. Spatially, the point closest to the center of the entire figure can be taken to represent the spatial coordinates of that sub-partition. The spatial domain distribution characteristics are characterized by spatial coordinates and spatial Euclidean distance, representing the distribution features of the sub-partitions and the adaptive (computational reference) characteristics during the global path optimization process. This dataset serves as input data for subsequently constructing the Traveling Salesman Problem (TSP) model, providing basic location information for solving the global access sequence of sub-partitions.

[0127] In this optional embodiment, by obtaining the local coordinates of the starting point of each sub-partition's local path in the sub-partition's local coordinate system, the relative position reference of the starting point within the sub-partition can be clearly defined, avoiding confusion in subsequent coordinate transformations due to ambiguous starting point position markings. This provides a clear and unified local position basis for subsequent global coordinate mapping. Simultaneously, by combining the starting point's local coordinates to determine the coordinate registration relationship between the sub-partition's image pixels and their actual spatial positions, the association link between the sub-partition's local coordinate system and the workpiece's actual spatial coordinate system can be accurately established, eliminating mapping gaps between local and global coordinates. This ensures that the underlying logic of coordinate transformation is consistent with the technical parameters of the previous imaging and partitioning stages, avoiding additional errors introduced by registration relationship deviations. Converting the starting point's local coordinates to global coordinates in the global coordinate system based on this coordinate registration relationship allows the starting point, which originally only represented the relative position within the sub-partition, to be accurately... Mapping the physical location to the actual workpiece space completely solves the problem of the disconnect between local path coordinates and the actual cleaning location, ensuring that subsequent path planning based on global coordinates can directly match the actual cleaning needs of the workpiece and prevent path deviation, missed cleaning, or incorrect cleaning during laser cleaning. Finally, the global coordinates of the starting points of all sub-partitions are summarized to form a global representative point set. This set, in the form of global representative points-sub-partition correspondence, transforms the scattered sub-partitions into a unified unit that can be identified by global path optimization. This provides input data for the subsequent transformation of the sub-partition global access sequence optimization problem into a Traveling Salesman Problem (TSP) model, giving the global ant colony adaptive optimization algorithm a clear optimization object, significantly improving the orderliness and efficiency of global path planning, and laying the positional foundation for subsequent global access sequence solving and final laser cleaning path generation, further ensuring the overall accuracy and efficiency of intelligent laser cleaning.

[0128] Optionally, generating a traveling salesman problem model based on the global representative point set includes:

[0129] Each of the starting points in the global representative point set is taken as a node in the traveling salesman problem model, and the Euclidean distance between any two nodes is calculated.

[0130] Using the Euclidean distance as the path cost between the two nodes, traverse all the nodes to determine the total path cost;

[0131] With the goal of minimizing the total path cost, a model of the Traveling Salesman Problem is constructed, which includes a set of nodes, a path cost matrix, and traversal constraints.

[0132] Specifically, the global ant colony adaptive path optimization module reads the global representative point set data and defines the starting point of the local path of each sub-partition in the set as the node to be traversed in the Traveling Salesman Problem model. Each node is uniquely associated with a sub-partition, and its corresponding sub-partition number and local path information can be traced through the node index, ensuring that the model nodes match the sub-partitions to be covered by laser cleaning one by one, with no omissions or redundancies. For any two nodes to be traversed in the model, the spatial straight-line distance calculation logic is used to traverse all node combinations and calculate the Euclidean distance between them. The calculated Euclidean distance is directly mapped to the path cost between the two nodes. According to the rule of traversing all nodes and each node being visited only once, for any set of node traversal order (e.g., node 1 → node 2 → … → node n → node 1), the path costs between adjacent nodes during the traversal are accumulated sequentially to obtain the total path cost corresponding to the traversal order, clarifying the core quantitative indicators that the model needs to optimize. Simultaneously, with minimizing the total path cost as the core optimization objective, all nodes to be traversed are aggregated into a node set. Each node in the set is accompanied by global coordinates and sub-partition indexes, ensuring that the model can associate the correspondence between nodes and sub-partitions. The path costs of all nodes, such as node i to node j, are organized into an n×n two-dimensional matrix according to the node index order, where n is the number of nodes, and the matrix element C(i,j) represents the path cost from node i to node j. This embodiment uses matrix form to facilitate the rapid invocation and calculation of the subsequent native heterogeneous competition-cooperative ant colony optimization algorithm, improving the algorithm's solution efficiency. Furthermore, each node is set to be visited at least once to ensure that all sub-partitions are covered, with no missed or duplicate cleaning, and the path is a continuous traversal sequence. This avoids meaningless empty movement of the laser head between non-adjacent nodes, reducing resource waste and ensuring that the model output meets the requirements of actual laser cleaning operations for path integrity and continuity.

[0133] In this optional embodiment, by defining the starting point of the corresponding sub-partition in the global representative point set as the Traveling Salesman Problem model node, a one-to-one precise correspondence between the model node and the actual cleaning sub-partition can be achieved, avoiding the problem of subsequent path planning being out of sync with the actual position of the workpiece caused by node position deviation, and ensuring that the cleaning operation is free from missed or incorrect cleaning. Using the Euclidean distance between any two nodes as the path cost, the abstract cost can be directly related to the actual path length, time and energy consumption of the laser head movement, so that the optimization objective of minimizing the total path cost effectively responds to the need to reduce cleaning resource consumption. The model is constructed by combining the constraints that each node is visited at least once and the path is traversed continuously, which not only ensures that all sub-partitions are fully covered and there are no repeated visits, avoiding invalid movements, but also adapts to the rapid solution of the subsequent native heterogeneous competition-cooperative ant colony optimization algorithm through the standardized node set and path cost matrix, reducing the time consumption of algorithm parameter conversion and improving the efficiency and accuracy of path planning.

[0134] Optionally, the step of solving the Traveling Salesman Problem model using the native heterogeneous competition-cooperative ant colony optimization algorithm to obtain the optimal global access sequence for all local paths includes:

[0135] Based on the ant population size, initial pheromone concentration, pheromone volatile factor, and preset regulatory threshold and preset grouping step size coefficient of the original heterogeneous competition stage, a parameter mapping table of the expected heuristic factor is constructed.

[0136] In native heterogeneous competition, keeping the initial pheromone concentration unchanged, the expected heuristic factors are injected into the ant population in the order of the parameter mapping table, driving the ants to independently complete path construction, and obtaining the total path cost, the first-order difference and the second-order difference of the total path cost;

[0137] When the number of iterations of the ant population reaches a preset control threshold, the second-order difference is classified by support vector machine to obtain the fluctuation convergence segment, and the expected heuristic factor corresponding to the smallest first-order difference in the fluctuation convergence segment is taken as the optimal expected heuristic factor.

[0138] Based on the optimal expectation heuristic factor, and combined with the probability transition algorithm, ant colony iteration is performed to obtain the optimal global access sequence.

[0139] Specifically, a parameter mapping table of desired heuristic factors is constructed based on the ant population size, initial pheromone concentration, pheromone evaporation factor, and preset control thresholds and preset grouping step size coefficients for the primary heterogeneous competition stage. Then, during the primary heterogeneous competition stage, keeping the initial pheromone concentration constant, the desired heuristic factors are injected into the ant population sequentially according to the order of the parameter mapping table, driving the ants to independently complete path construction. When the number of iterations in the ant population reaches the preset control threshold, a support vector machine is used to perform binary classification on the second-order differences, thereby obtaining the fluctuating convergence segment. Furthermore, the desired heuristic factor corresponding to the smallest first-order difference within the fluctuating convergence segment is taken as the optimal desired heuristic factor. Finally, based on the optimal desired heuristic factor, a probabilistic transition algorithm is used to iterate the ant colony, ultimately obtaining the optimal global access sequence.

[0140] In a preferred embodiment of the present invention, the specific implementation is as follows: After the local partitioning module has finished executing, a global sequence serialization taboo table is defined. , For the first Node and the The set of paths between nodes. It should be noted that consecutive nodes involved in the path optimization process of the ant colony algorithm are random nodes, represented here by d and q. In the ant colony optimization algorithm... In the next iteration, there are Probability of ant seeking the best at any given moment:

[0141] (1);

[0142] in, This is the set of candidate nodes that can be accessed by the global ant in a single iteration, i.e., the set of sub-partition sequences that have not yet been visited during the pathfinding process. , For all selectable nodes, The pheromone heuristic factor in the global ant colony optimization process. For the first The expected heuristic factor at the next iteration, where let Represents the global dynamic optimal expectation heuristic factor. Time-related nodes With nodes Global heuristic function matrix Defined as ,in For nodes With nodes European distance between them for Time-related nodes With node set The global heuristic function matrix, for Time-related nodes With nodes The pheromone matrix for Time-related nodes With node set The pheromone matrix Let represent the global heuristic function matrix relating node d and node set q at time t.

[0143] Pick If the boundary threshold is heterogeneous regulation, then In the process of secondary heterogeneous regulation, the global expectation heuristic factor The dynamic mapping relationship can be represented as:

[0144] (2);

[0145] in, This serves as the index variable for the dynamic parameters of the heuristic factor. Furthermore, the pheromone update process for global ant colony optimization can be described as follows:

[0146] (3);

[0147] (4);

[0148] (5);

[0149] Where w represents the ant colony size. The initial pheromone concentration, After the Xth iteration, The pheromone concentration on path dq at any given time. This represents the increase in pheromones during that period. For the first The amount of pheromone released by an ant when it travels along path dq in the x-th iteration. For the first The objective function value of a single ant in the x-th iteration. Q is the globally optimized pheromone evaporation factor, and Q is the globally optimized pheromone intensity. Also, considering its redundancy with the expected heuristic factor, Q is set to 1 here.

[0150] Decomposing the iteration number x yields:

[0151] , (6);

[0152] For any number of iterations x, define the first difference of the global optimization objective function between the second and third iterations. and its second-order difference function They are represented as follows:

[0153] (7);

[0154] (8);

[0155] To achieve high-precision prediction and segmentation under small sample conditions, this invention proposes a support vector machine prediction model (FFD-SVM) that integrates feedforward discrete passband diagnostics. Its mathematical modeling is as follows:

[0156] , (9);

[0157] in, The output classifies the second-order difference gain magnitude of the SVM over the entire iteration cycle. This is set as a binary category label space. =1, -1, represents the threshold parameter of the feedforward discrete passband filter.

[0158] (10);

[0159] in, For the passband reference bias operator, For a sample size of Threshold gain coefficient at time, kernel function of SVM It can be characterized as:

[0160] (11);

[0161] in, These are the training data matrix and the label vector, respectively. The weight bandwidth parameter of the radial basis function. and These are the structural parameters of the SVM model.

[0162] In the optimal hyperparameter optimization stage, for Category input variables Initialize the index counter ,satisfy:

[0163] , (12);

[0164] The dynamic parameter index variable can be derived through derivation. :

[0165] , (13);

[0166] Then, obtain the optimal expected heuristic factor of its mapping:

[0167] (14).

[0168] Optionally, generating the laser cleaning scanning path for the surface of the workpiece to be cleaned based on the optimal global access sequence includes:

[0169] Map the node order in the optimal global access sequence to the execution order of the corresponding sub-partitions;

[0170] The local paths of each sub-partition are extracted sequentially according to the execution order of the sub-partitions, and the local paths are concatenated to obtain the concatenated overall path.

[0171] Speed ​​planning is performed on the spliced ​​overall path to generate the laser cleaning scanning path.

[0172] Specifically, a correspondence is established between node sequences and sub-partitions, mapping the node order in the optimal global access sequence to the execution order of the corresponding sub-partitions, thus providing clear guidance for subsequent path splicing. This mapping ensures that the laser cleaning head accesses each sub-partition in the optimal global order, avoiding unnecessary round trips and repetitive operations, significantly improving cleaning efficiency. Next, the local paths of each sub-partition are extracted sequentially according to their execution order and spliced ​​to form the spliced ​​overall path. Finally, speed planning is performed on the spliced ​​overall path to generate the final laser cleaning scanning path. In this embodiment of the invention, speed planning involves rationally adjusting the laser head's movement speed based on the specific requirements of the cleaning task, the performance parameters of the laser equipment, and the material characteristics of the workpiece. For example, when cleaning difficult-to-remove contaminants, the speed may need to be reduced to ensure cleaning effectiveness; while when cleaning large areas of lightly contaminated areas, the speed can be appropriately increased to improve efficiency. Speed ​​planning not only improves cleaning quality but also reduces energy consumption and equipment wear.

[0173] This optional embodiment effectively solves the problems of path discontinuity, low efficiency, and inconsistent cleaning results in traditional path planning methods. Through precise node mapping, path splicing, and speed planning, a highly efficient scanning path suitable for laser cleaning is generated, significantly improving the practicality and reliability of laser intelligent cleaning technology.

[0174] In a preferred embodiment of the present invention, an application example of cleaning residual paint layers from an aircraft is used to further illustrate the present invention. The specific steps are as follows:

[0175] Step one: Extract image features and coordinates of the target. First, fix the workpiece to be maintained. As described in the invention, use the image acquisition module to sample, map coordinates, and binarize the surface of the workpiece within the field of view, generating a binarized feature result containing spatial coordinate information, and generating the corresponding target feature binarized image, as shown below. Figure 2 As shown, the white area represents the topcoat area, and the black area represents the surface damage area.

[0176] Step two: Based on the required maintenance scheme, the binary feature map is topologically partitioned and the target region path is filled using the local path heuristic module. This example demonstrates a custom maintenance scheme for aircraft paint cleaning residues, which simplifies the topology of each connected component into sequentially indexable sub-intervals, and then partitions each sub-partition... Figure 2The white area in the middle is used for scanning path coverage. In this example, the inner spiral heuristic is used to fill the path of the sub-partitions, and the data is uploaded to the global adaptive optimization module described in this invention in the form of a global representative node index (that is, each sub-partition is regarded as a global node with sequence information and coordinate information, and the representative coordinates of each sub-partition are explained as above). The partition and local path coverage diagram is shown below. Figure 3 As shown.

[0177] Step three uses the path inspiration within the sub-partition obtained in step two as a transition node for global-local mode switching, realizing continuous and smooth planning of the planar target coverage path, and demonstrating the ant colony algorithm's hyperparameter adaptive optimization process in global planning. Combined with... Figure 4 As shown, Figure 4 Presenting the observation results of parameter competition under the native heterogeneous regulation of ant colonies: By balancing the identifiability of the sampling model features with the overall system operation time, a regulation threshold is set. The number of trials is 60, with a grouping step size coefficient of 3 (as shown in Formula 6), meaning that every 3 observations of the same parameter are divided into 1 group, making the control process equally subdivided into 20 competing groups. An incremental β parameter mapping table is constructed within the value range of 0.01-20, and the competitive parameters are sequentially mapped to the 20 competing groups. At this time, the increase in the number of competitions is equivalent to the corresponding increase in the β parameter. According to Formulas 3-5, a failure operation is performed on the pheromone update mechanism and iterative optimization process during the control process.

[0178] In the native heterogeneous competition process, the embodiments of the present invention address... Figure 4 The trend analysis of the curve change is as follows: The change of the β parameter can be directly mapped to the expected heuristic weight, and its impact on the path search distance is as follows: Both excessively large and excessively small β values ​​are detrimental to path distance optimization. In the absence of pheromone intervention, when the β value is large, the ant colony tends to adopt a prior probability selection strategy, reducing randomness and significantly enhancing local optima; when the β value is small, it tends to favor random search, leading to an increase in path search distance and significant fluctuations in the iteration results within the group. Analysis of the intra-group fluctuation characteristics reveals that a significant step change usually occurs in the third observation within the initial competitive group. Figure 4 The first six groups of step characteristics and their directions of change are marked with dashed boxes and arrows. This phenomenon indicates a cross-correlation between the β value and the randomness of ant colony search. Based on Formula 2, forward differencing is performed on the third observation of each group, and the calculated change characteristic curve is shown below. Figure 5 As shown by the red dashed line, observations indicate that the fluctuation state of the first-order change in path search can be divided into three characteristic intervals as the β value increases: the oscillation region, the fast convergence region, and the stable region. The convergence region has a relatively short convergence time and can be approximated as the critical boundary. The early part of the stable region typically corresponds to the index range of the ideal parameters in the traditional ant colony algorithm, indicating that the dynamic optimization of the global path search in ant colonies is constrained by the trade-off between its prior and stochastic parameters.

[0179] To achieve feature extraction of ideal hyperparameters, a corresponding second-order difference curve is generated according to Formula 8 to explicitly represent the change vector of the first-order change in the path search. For example... Figure 5 As shown, the overall trend of the second-order difference curve, represented by the solid line, is basically consistent with that of the first-order difference curve, and it is easier to identify features—the second-order difference curve allows for more accurate segmentation of fluctuation amplitudes and enables the location and extraction of characteristic extreme points of the first-order change curve. In summary... Figure 4 and Figure 5 The path search random characteristic curve is used as a joint constraint condition with the path search distance and its corresponding change. In this example, based on the determination of the effective parameter feature point selection range, near the second wave peak where the fluctuation tends to be gentle, the third feature point corresponds to the stable amplitude segment of the second difference curve. The ant colony path search effect is optimal under the hyperparameter mapping corresponding to this point.

[0180] Based on the above conclusions regarding ideal hyperparameter localization, and following the methodological framework of Equations 9-14, the hyperparameter extraction results are as follows: Figure 6 The diagram illustrates the ideal hyperparameter indexing process in sequence: Support Vector Machine (SVM) classification prediction - feedforward passband diagnostics - feature point counter indexing. The default settings for the SVM kernel function prediction parameter group are: =1, =1, =3, set the feedforward diagnostic passband threshold parameter group as follows: =1200, k=1, the prediction and classification error is further improved. By implementing the feature point counting index method on the stable fluctuation state, the adaptive optimization solution of the ant colony β factor can be effectively derived.

[0181] Corresponding to the above optimization results, Figure 7 This paper presents a dynamic adaptive optimization curve based on the native heterogeneous competition-cooperation mechanism of ant colonies. The curve is divided into three stages: native heterogeneous regulation, hyperparameter competition extraction, and ant colony cooperative pathfinding. It is evident that the pheromone interaction and iterative update mechanism among ants is activated immediately after the optimal hyperparameters are determined, and the mapping process of the ideal hyperparameters effectively suppresses local optima. Specifically, the program sets the sum of the number of observations for native heterogeneous hyperparameter regulation and the number of cooperative iterations to 300, ensuring coverage of the effective iteration cycle of the cooperative stage. In this example scenario, the default settings for each hyperparameter are α=1, ρ=0.3, and num=5. The optimized ideal mapping β factor is β=3. After effective iteration by the ant colony, the ideal global path solution is obtained. At this point, the effective iteration count is 100, and the path distance is 7601. The global planning is illustrated below. Figure 8 As shown, the dashed line represents the global motion path. The convergence path cost and convergence effect obtained demonstrate the accuracy and effectiveness of the method described in this invention.

[0182] Combination Figure 9 As shown, a laser cleaning path planning system of the present invention includes:

[0183] The imaging acquisition and processing unit is used to acquire images of the surface of the workpiece to be cleaned, perform coordinate registration and binarization processing, and obtain a binarized image of the surface of the workpiece to be cleaned.

[0184] The local path determination unit is used to regularly partition the surface of the workpiece to be cleaned according to the binarized image to obtain multiple sub-partitions, and to perform path coverage for each sub-partition to obtain the local path of each sub-partition.

[0185] A modeling unit is used to map the starting point of the local path in each sub-partition to obtain a global representative point set of all local paths, and to generate a traveling salesman problem model based on the global representative point set.

[0186] The solving unit is used to solve the traveling salesman problem model using the native heterogeneous competition-cooperative ant colony optimization algorithm to obtain the optimal global access sequence for all the local paths;

[0187] The path generation unit is used to generate a laser cleaning scanning path for the surface of the workpiece to be cleaned based on the optimal global access sequence.

[0188] The laser cleaning path planning system of the present invention has the same advantages over the prior art as the laser cleaning path planning method described above, and will not be repeated here.

[0189] Combination Figure 10 As shown, an electronic device according to the present invention includes a memory and a processor;

[0190] The memory is used to store computer programs;

[0191] The processor is configured to implement the laser cleaning path planning method as described above when executing the computer program.

[0192] The electronic device of the present invention has the same advantages over the prior art as the laser cleaning path planning method described above, and will not be repeated here.

[0193] While the present invention has been disclosed above, its scope of protection is not limited thereto. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention, and all such changes and modifications will fall within the scope of protection of the present invention.

Claims

1. A laser cleaning path planning method, characterized in that, include: The surface of the workpiece to be cleaned is imaged, and coordinate registration and binarization are performed to obtain a binarized image of the surface of the workpiece to be cleaned. The surface of the workpiece to be cleaned is regularly partitioned according to the binarized image to obtain multiple sub-partitions, and path coverage is performed for each sub-partition to obtain the local path of each sub-partition; Map the starting point of the local path in each sub-partition to obtain a global representative point set for all local paths, and generate a traveling salesman problem model based on the global representative point set. The Traveling Salesman Problem model is solved using the native heterogeneous competition-cooperative ant colony optimization algorithm to obtain the optimal global access sequence for all local paths; including: constructing a parameter mapping table of expected heuristic factors based on ant population size, initial pheromone concentration, pheromone evaporation factor, and preset control threshold and preset grouping step size coefficient of the native heterogeneous competition stage. In native heterogeneous competition, keeping the initial pheromone concentration unchanged, the expected heuristic factors are injected into the ant population in the order of the parameter mapping table, driving the ants to independently complete path construction, and obtaining the total path cost, the first-order difference and the second-order difference of the total path cost; When the number of iterations of the ant population reaches a preset control threshold, the second-order difference is classified by support vector machine to obtain the fluctuation convergence segment, and the expected heuristic factor corresponding to the smallest first-order difference in the fluctuation convergence segment is taken as the optimal expected heuristic factor. Based on the optimal expectation heuristic factor, the ant colony is iterated using a probability transition algorithm to obtain the optimal global access sequence; Based on the optimal global access sequence, a laser cleaning scanning path is generated for the surface of the workpiece to be cleaned.

2. The laser cleaning path planning method according to claim 1, characterized in that, The process of imaging and acquiring the surface of the workpiece to be cleaned, performing coordinate registration and binarization to obtain a binarized image of the workpiece surface includes: Digital imaging is performed on the surface of the workpiece to be cleaned to obtain the original grayscale image of the surface of the workpiece to be cleaned; The original grayscale image is registered with the actual spatial position of the surface of the workpiece to be cleaned by using a coordinate transformation algorithm to obtain a registered grayscale image. Adaptive thresholding or maximum inter-class variance segmentation is performed on the registered grayscale image to obtain the binarized image of the surface of the workpiece to be cleaned.

3. The laser cleaning path planning method according to claim 1, characterized in that, The process involves regularly partitioning the surface of the workpiece to be cleaned based on the binarized image to obtain multiple sub-partitions, including: Connectivity analysis is performed on the foreground pixels in the binarized image to obtain multiple connected components to be cleaned; For each of the connected components to be cleaned, extract the minimum bounding contour region of the connected component to be cleaned; The threshold for partition size is determined based on the scanning range of the field lens of the laser cleaning device, the cleaning accuracy requirements, and the path planning efficiency requirements. Based on the partition size threshold, the connected component to be cleaned is adaptively divided to obtain multiple sub-partitions.

4. The laser cleaning path planning method according to claim 3, characterized in that, The step of performing path coverage for each sub-partition to obtain the local path of each sub-partition includes: The scanning reference direction of the local path is determined based on the shape of the sub-partition and the adaptability of the scanning direction of the laser cleaning device. Based on the scanning reference direction and the path coverage algorithm corresponding to the shape of the sub-partition, the starting point, ending point, and path node sequence in the sub-partition are obtained; An initial local path is generated for each sub-partition based on the starting point, the ending point, and the path node sequence in the sub-partition. The initial local path is optimized by using a Bézier curve to obtain the local path.

5. The laser cleaning path planning method according to claim 4, characterized in that, The process of mapping the starting point of the local path in each sub-partition to obtain a global representative point set for all local paths includes: Obtain the local coordinates of the starting point of each sub-partition in the local coordinate system of the sub-partition; Based on the local coordinates of the starting point, determine the coordinate registration relationship between the image pixels of the sub-partition and their actual spatial positions; Based on the coordinate registration relationship, the local coordinates of the starting point are converted into global coordinates in the global coordinate system; The global coordinates of the starting points of all sub-partitions are summarized to form the global representative point set.

6. The laser cleaning path planning method according to claim 1, characterized in that, The step of generating a traveling salesman problem model based on the global representative point set includes: Each of the starting points in the global representative point set is taken as a node in the traveling salesman problem model, and the Euclidean distance between any two nodes is calculated. Using the Euclidean distance as the path cost between the two nodes, traverse all the nodes to determine the total path cost; With the goal of minimizing the total path cost, a model of the Traveling Salesman Problem is constructed, which includes a set of nodes, a path cost matrix, and traversal constraints.

7. The laser cleaning path planning method according to claim 6, characterized in that, The step of generating the laser cleaning scanning path for the surface of the workpiece to be cleaned based on the optimal global access sequence includes: Map the node order in the optimal global access sequence to the execution order of the corresponding sub-partitions; The local paths of each sub-partition are extracted sequentially according to the execution order of the sub-partitions, and the local paths are concatenated to obtain the concatenated overall path. Speed ​​planning is performed on the spliced ​​overall path to generate the laser cleaning scanning path.

8. A laser cleaning path planning system, characterized in that, include: The imaging acquisition and processing unit is used to acquire images of the surface of the workpiece to be cleaned, perform coordinate registration and binarization processing, and obtain a binarized image of the surface of the workpiece to be cleaned. The local path determination unit is used to regularly partition the surface of the workpiece to be cleaned according to the binarized image to obtain multiple sub-partitions, and to perform path coverage for each sub-partition to obtain the local path of each sub-partition. A modeling unit is used to map the starting point of the local path in each sub-partition to obtain a global representative point set of all local paths, and to generate a traveling salesman problem model based on the global representative point set. The solution unit is used to solve the traveling salesman problem model using the native heterogeneous competition-cooperative ant colony optimization algorithm to obtain the optimal global access sequence for all local paths; including: constructing a parameter mapping table of expected heuristic factors based on the ant population size, initial pheromone concentration, pheromone evaporation factor, and preset control threshold and preset grouping step size coefficient of the native heterogeneous competition stage. In native heterogeneous competition, keeping the initial pheromone concentration unchanged, the expected heuristic factors are injected into the ant population in the order of the parameter mapping table, driving the ants to independently complete path construction, and obtaining the total path cost, the first-order difference and the second-order difference of the total path cost; When the number of iterations of the ant population reaches a preset control threshold, the second-order difference is classified by support vector machine to obtain the fluctuation convergence segment, and the expected heuristic factor corresponding to the smallest first-order difference in the fluctuation convergence segment is taken as the optimal expected heuristic factor. Based on the optimal expectation heuristic factor, the ant colony is iterated using a probability transition algorithm to obtain the optimal global access sequence; The path generation unit is used to generate a laser cleaning scanning path for the surface of the workpiece to be cleaned based on the optimal global access sequence.

9. An electronic device, characterized in that, Including memory and processor; The memory is used to store computer programs; The processor is configured to implement the laser cleaning path planning method as described in any one of claims 1 to 7 when executing the computer program.

Images