Intelligent planning method and system for wire cutting machine cutting path based on image processing

Abstract

The present invention relates to the field of image processing technology, and discloses a method and system for intelligent planning of cutting paths for wire cutting machines based on image processing. The method comprises: performing multi-angle image acquisition and preprocessing on a workpiece to be processed to obtain a workpiece feature map; inputting the workpiece feature map into a perception region recognition network for processing to obtain a target processing region recognition map; performing feature enhancement on the target processing region recognition map to obtain an enhanced feature map and an importance weight map; performing cutting path planning based on the enhanced feature map and the importance weight map to obtain an initial cutting path; performing multi-objective optimization on the initial cutting path to obtain a target cutting path and a corresponding electric spark parameter configuration table. The present invention realizes the coordinated optimization of the cutting path and the electric spark parameters, balances the contradictions between processing accuracy, processing time and energy consumption, and enables the system to achieve optimal process matching on workpieces of different materials and geometric complexities.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to an image processing-based method and system for intelligent planning of cutting paths for wire cutting machines. Background Art

[0002] Traditional wire cutting path planning algorithms are no longer able to meet the demands of modern manufacturing. Relying primarily on simple edge detection techniques, these algorithms are unable to effectively extract features from complex workpieces. This is particularly true for multi-material structures and workpieces with variable cross-sections. They struggle to accurately identify primary and secondary features on the workpiece, resulting in low cutting accuracy.

[0003] In actual machining, traditional wire-cutting path planning methods fail to fully account for the specific geometric features of the workpiece, such as the machining requirements of target areas like sharp corners, fine holes, and narrow slits. This makes the electrode wire prone to vibration and unstable electric spark discharge when cutting in these areas. Furthermore, traditional methods have poor adaptability to different material properties and workpiece geometries, and are unable to perform differentiated processing. This leads to uneven electrode wire loss, high energy consumption during machining, and unstable surface roughness, seriously affecting machining quality and production efficiency. Summary of the Invention

[0004] The present invention provides a method and system for intelligent planning of cutting paths for wire cutting machines based on image processing. The present invention realizes the coordinated optimization of cutting paths and electric spark parameters, balances the contradictions among machining accuracy, machining time and energy consumption, and enables the system to achieve optimal process matching on workpieces of different materials and geometric complexity.

[0005] In a first aspect, the present invention provides an intelligent planning method for a wire cutting machine cutting path based on image processing, the intelligent planning method for a wire cutting machine cutting path based on image processing comprising:

[0006] Perform multi-angle image acquisition and preprocessing on the workpiece to be processed to obtain the workpiece feature map;

[0007] Inputting the workpiece feature map into the perception area recognition network for processing to obtain a target processing area recognition map;

[0008] Performing feature enhancement on the target processing area identification map to obtain an enhanced feature map and an importance weight map;

[0009] Performing cutting path planning according to the enhanced feature map and the importance weight map to obtain an initial cutting path;

[0010] The initial cutting path is subjected to multi-objective optimization to obtain a target cutting path and a corresponding electric spark parameter configuration table.

[0011] In a second aspect, the present invention provides an intelligent planning system for cutting paths of wire cutting machines based on image processing, the intelligent planning system for cutting paths of wire cutting machines based on image processing comprising:

[0012] An acquisition module is used to acquire and pre-process multi-angle images of the workpiece to be processed to obtain a feature map of the workpiece;

[0013] A region recognition module is used to input the workpiece feature map into a perception region recognition network for processing to obtain a target processing region recognition map;

[0014] A feature enhancement module is used to perform feature enhancement on the target processing area identification map to obtain an enhanced feature map and an importance weight map;

[0015] A path planning module, configured to perform cutting path planning based on the enhanced feature map and the importance weight map to obtain an initial cutting path;

[0016] The multi-objective optimization module is used to perform multi-objective optimization on the initial cutting path to obtain a target cutting path and a corresponding electric spark parameter configuration table.

[0017] In the technical solution provided by the present invention, through the multi-angle acquisition and adaptive preprocessing technology of industrial cameras, it is possible to fully capture the geometric feature information of the workpiece, effectively overcome the information loss problem caused by traditional single-view acquisition, and provide high-quality workpiece feature maps for subsequent accurate identification. The dual-stream feature extraction architecture of the perception area recognition network is adopted to achieve the coordinated extraction of macro-geometric features and micro-texture details, solve the problem of information transmission loss when processing sharp angles and small features in traditional methods, and accurately identify the target areas that require differentiated processing. Through the supplementation and fusion mechanism of global and local features, the system can pay special attention to small feature areas such as sharp corners and fine holes, accurately depict the complex contour features of the workpiece, and solve the problem of insufficient recognition of small features by traditional methods. Intelligent path planning based on enhanced feature maps and importance weight maps can implement differentiated control point density settings and feed speed adjustments for different geometric characteristic areas, ensure a stable and efficient cutting process, and significantly improve the quality of complex contour processing. By establishing a multi-objective optimization model that comprehensively considers geometric constraints and process constraints, the coordinated optimization of cutting path and EDM parameters is achieved, and the contradictions between machining accuracy, machining time and energy consumption are balanced, so that the system can achieve optimal process matching on workpieces of different materials and geometric complexity. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0019] Figure 1 A schematic diagram of an embodiment of a method for intelligent planning of cutting paths for a wire cutting machine based on image processing in an embodiment of the present invention;

[0020] Figure 2 Schematic diagram of an embodiment of an intelligent planning system for cutting paths of wire cutting machines based on image processing in an embodiment of the present invention. DETAILED DESCRIPTION

[0021] An embodiment of the present invention provides an image processing-based intelligent planning method and system for cutting paths of wire cutting machines. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or that are inherent to these processes, methods, products or devices.

[0022] For ease of understanding, the specific process of the embodiment of the present invention is described below. Figure 1 In one embodiment of the present invention, an intelligent planning method for a wire cutting machine cutting path based on image processing includes:

[0023] Step S101: Capture and preprocess multi-angle images of the workpiece to be processed to obtain a workpiece feature map;

[0024] It is understandable that the execution subject of the present invention can be a wire cutting machine cutting path intelligent planning system based on image processing, or a terminal or a server, which is not limited here. The embodiment of the present invention is described by taking the server as the execution subject as an example.

[0025] Specifically, a high-precision, multi-angle image acquisition system was constructed. This system is based on an industrial-grade, high-resolution camera with a resolution of at least 4096 × 3072 pixels and equipped with a ring-shaped LED light source to ensure shadow-free and uniform illumination. The system is fixed above the machine tool worktable and maintained at a vertical distance of 350 mm to capture workpiece images that meet geometric accuracy requirements. Furthermore, raw image datasets from multiple sides of the workpiece are acquired through multi-angle rotation, ensuring image integrity and detail representation of complex geometric shapes at different angles, resulting in a raw image sequence. An adaptive Gaussian filter is applied to the raw image sequence, using a dynamic kernel size adjustment mechanism to control the filter kernel size between 3 × 3 and 7 × 7 based on the local noise distribution of the image. This effectively suppresses random high-frequency noise while preserving edge structure information, generating a denoised image dataset. Histogram equalization is then performed on the denoised image dataset, and a piecewise linear mapping method is used to enhance image contrast and grayscale gradients at the edges. A bicubic interpolation algorithm is then applied for geometric correction to eliminate deformation errors introduced by image distortion, resulting in a rectified image dataset with a uniform perspective and enhanced detail. To extract the workpiece's main area, the rectified image is segmented. A modified Otsu thresholding method is used, combined with morphological processing such as closing operations and erosion and dilation, to extract a continuous and complete workpiece contour map. Coordinate registration and image fusion techniques are then used to integrate contour information from different angles into a single image representation. Edge detection is performed on the workpiece contour map, and an improved Canny algorithm is used to calculate an edge intensity map combining gradient magnitude and direction information. This map enhances the gradient response of features such as corners, holes, slopes, and slender edges. The edge intensity map is uniformly converted to a standard size and pixel values are normalized to generate a workpiece feature map.

[0026] Step S102: inputting the workpiece feature map into the perception area recognition network for processing to obtain a target processing area recognition map;

[0027] Specifically, the standardized workpiece feature map is input into the two parallel substructures of the perception area recognition network. The residual feature extraction branch adopts a four-level cascaded residual block architecture. Each level of residual block adopts a bottleneck structure, which includes a deep module consisting of convolutional layers, normalization layers and ReLU activation functions. It has the ability to compress and expand feature channels layer by layer, thereby performing step-by-step spatial structure abstraction and deep feature extraction of the image, and generating a main feature map with high-level semantic information expression capabilities. The main feature map effectively identifies geometric key areas such as sharp corners, inner holes, and slits; at the same time, the workpiece feature map texture preservation branch is composed of three levels of convolutional layers with different receptive field scales. The convolution kernels of each level adopt different sizes of 3×3, 5×5 and 7×7, respectively, and are supplemented by a multi-scale filtering mechanism to retain edge details and local texture changes, thereby generating an auxiliary feature map with multi-scale resolution capabilities. Based on heterogeneous feature pyramid technology, a multi-level skip connection pathway is established between the main feature map and the auxiliary feature map. This connection mechanism allows the interactive fusion of low-level texture features and high-level semantic features in a multi-scale structure, avoiding information isolation and feature gaps, and forming an enhanced feature set. A channel-level adaptive weight calculation mechanism is applied to the enhanced feature set. By performing average pooling, fully connected mapping, and softmax normalization on each feature channel, the weight coefficient of each channel in process judgment is calculated, generating a process-optimized feature representation that combines regional saliency and process sensitivity. The process-optimized feature representation is input into a pixel-level classification decoder, which uses a combination of deconvolution upsampling and skip connections to gradually restore spatial resolution, perform precise boundary localization, and screen target regions. A softmax classification layer is used to evaluate the probability of the processing region for each pixel, and outputs a target processing region identification map.

[0028] Step S103: performing feature enhancement on the target processing area identification map to obtain an enhanced feature map and an importance weight map;

[0029] Specifically, the target processing area identification map is used as input. Through complex feature sampling, N local patch images centered on the key points of the target area are extracted from the image. Each patch is 32×32 pixels in size and covers complex geometric features such as sharp corners, internal hole edges, and sharp transitions, forming a set of complex workpiece feature patches with structural recognition value. This patch set is input into the identity-related feature reconstruction module, which consists of global and local attention supplementary branches. This module extracts local geometric representations layer by layer through a three-level convolutional structure and introduces an attention mechanism in the feature channel dimension. The response strength of each feature channel is differentially modeled through the local perception path and the global context path, generating a feature importance weight matrix. This matrix assigns different response strengths to the features of different regions in each patch, effectively amplifying the recognition sensitivity of key processing areas. On this basis, the original target processing region identification map is simultaneously fed into the enhanced hierarchical feature fusion branch. The initial feature extraction layer extracts basic spatial structural features through standard convolution to generate a basic feature map. This is followed by a three-layer hierarchical feature extraction module. Each layer utilizes a dense connection strategy to ensure feature integrity while maintaining an expanded receptive field, resulting in a multi-scale feature set that encompasses low-level details, mid-level structure, and high-level semantics. To fully leverage the expressive power of different feature levels, a differentiated weighted integration mechanism is employed for this multi-scale feature set. Global average pooling and fully connected layers are used to calculate attention weights for each scale feature map. Fusion is then performed based on these weights, resulting in a fused feature representation that integrates boundary perception, morphological understanding, and processing intent. During the fusion stage, the fused feature representation is combined with the aforementioned feature importance weight matrix across branches. The channel response value at each position in the fused feature map is multiplicatively coupled with the corresponding weight matrix value. This information is then integrated across scales through weighted accumulation to generate an enhanced feature map. The channel attention responses generated during the fusion process are then independently output as importance weight maps.

[0030] The base feature map is fed into the first hierarchical feature extraction layer of the enhanced hierarchical feature fusion branch for feature decomposition. Standard convolutional modules are used to extract local spatial variation information and construct a preliminary first-layer feature map. To expand the receptive field and enhance the dimensionality of feature representation, a residual feature expansion process is applied to this first-layer feature map. This process alternates between 1×1 and 3×3 convolution kernels, and employs skip connections to maintain the continuity of the original feature path, generating the first-layer enhanced feature map. The first-layer enhanced feature map is then fed into the second hierarchical feature extraction layer. In this layer, in addition to performing conventional spatial abstraction convolutions, a skip connection channel is constructed directly from the first-layer enhanced feature map to the third layer. This channel maintains the resolution of the original texture features in a full-channel manner. Furthermore, a self-attention mechanism is introduced within the second layer to perform intra-channel attention filtering on the second-layer feature map. Inter-channel importance calculations are used to suppress redundant or non-critical information, generating the second-layer attention-weighted feature map. The skip-path data of the second-layer attention-weighted feature map and the first-layer enhanced feature map are fed into the third-layer hierarchical feature extraction layer, which focuses on feature complementarity and fusion. This layer utilizes a densely connected structure to spatially align and dimensionally fuse cross-layer data to generate the third-layer fused feature map. The first-layer enhanced feature map, the second-layer attention-weighted feature map, and the third-layer fused feature map are fed into the multi-scale aggregation module, which uses feature channel compression, spatially weighted fusion, and multi-scale response superposition to form a multi-scale feature set.

[0031] Step S104: performing cutting path planning based on the enhanced feature map and the importance weight map to obtain an initial cutting path;

[0032] Specifically, an edge extraction operation is performed on the enhanced feature map, and an improved Canny edge detection algorithm based on gradient response enhancement is used to obtain an initial set of edge points. This extraction process preserves the main contour segments and fully addresses sharp corners, hole edges, and curvature changes in the image, ensuring complete recognition of complex areas. Subsequently, a weighted adjustment process is performed on the edge points using an importance weight map. Each edge point's position is fine-tuned by superimposing its position coefficient on its corresponding position in the weight map. This generates a first set of edge points that reflects geometric feature accuracy and process priority. This first set of edge points is prioritized, and a priority function is constructed based on the importance response value, local edge continuity, and geometric stability of the corresponding region. The sorted results are then reorganized into a second set of edge points, which is reorganized based on local structural continuity, ensuring a consistent foundation for contour semantics during the fitting process. Based on this, curve fitting is performed on the second set of edge points, using a third-order B-spline interpolation algorithm to construct a first contour curve from the discrete edge points. This curve exhibits C² continuity and ensures smooth motion during wire cutting. To improve the geometric adaptability of the path to complex structures, the control point density of the first contour curve is subdivided based on an importance weight map. The control point distribution density is increased in areas with high weight response values, while the density is maintained at a low level in straight contour areas. This generates a second contour curve that accurately covers critical machining structures such as tiny holes, sharp corner transitions, and concave areas. Based on the geometric configuration of the second contour curve, the entry point is optimized. By analyzing the material thickness gradient, the local edge curvature distribution, and the estimated internal stress field of the workpiece, a comprehensive scoring function is constructed. The location on the contour curve with the lowest stress concentration, the smoothest curvature, and the moderate material thickness is selected as the optimal machining starting position, ensuring optimal heat-affected zone and path stability during the initial machining phase. An adaptive feed rate scheduling mechanism is established based on the curvature function of the second contour curve and the regional distribution information of the importance weight map. This feed rate distribution function is constructed by reducing the feed rate in areas with high curvature or high weight and increasing the cutting speed in straight, low-weight areas. Taking the second contour curve as the path basis, the optimized processing starting point is used as the path starting point, and the cutting speed of each segment on the path is assigned in segments according to the cutting feed speed function. An initial cutting path with continuous structure, adjustable speed and reasonable starting point is constructed, which includes the electrode wire center trajectory. The offset parameters are added according to the material properties to construct the electrode wire compensation trajectory and form standardized path data.

[0033] Step S105: Perform multi-objective optimization on the initial cutting path to obtain a target cutting path and a corresponding EDM parameter configuration table.

[0034] Specifically, based on actual machining requirements, machining accuracy requirements, workpiece contour tolerance, wire tension limit, and maximum cutting speed are set as hard constraints in the optimization model. A joint objective function is constructed for accuracy, time, and energy consumption, forming a well-structured multi-objective optimization model. Image features are discretized from the enhanced feature map, and the image space is divided into several grid cells using a fixed-size two-dimensional grid. Each grid cell is assigned a differentiated weight coefficient based on the grayscale response intensity of the corresponding region in the importance weight map, forming a feature importance matrix with spatial regional significance annotation capabilities. This matrix is used to determine the impact of each segment in the path on machining quality and control strategy. The initial cutting path is input into the optimization model, and a particle swarm optimization process is performed based on the feature importance matrix. In this process, each particle encodes both the coordinate vectors of the path control points and the local cutting parameter vectors, such as feed rate and wire compensation. The optimization algorithm incorporates an adaptive inertia weight strategy to dynamically adjust the particle search range and convergence speed. The weight decays nonlinearly with the number of iterations, effectively avoiding local optimal traps and obtaining an iteratively updateable path optimization solution. To ensure the physical continuity of the path and motion stability during machining, a curvature penalty factor is added as a regularization term in the optimization scheme to suppress sudden changes in curvature between adjacent path segments. By calculating the local curvature changes at the control points and constraining the accumulation of their squared differences, a target cutting path with high curvature consistency and smoothness is obtained. To ensure the use of the most appropriate EDM parameters in different workpiece areas, a material-parameter mapping function is constructed based on the workpiece material characteristic parameter matrix and the feature importance matrix. The material parameter matrix includes indicators such as conductivity, thermal conductivity, and hardness, and the feature importance matrix provides regional machining complexity weights. The two are established through interpolation fitting or neural mapping models. The optimal voltage and current combination parameters are dynamically calculated at each position point on the target cutting path to generate an EDM parameter configuration table. This configuration table has spatial position resolution capabilities and can achieve precise matching of discharge power for different machining areas in the control system.

[0035] In embodiments of the present invention, multi-angle acquisition and adaptive preprocessing technology using industrial cameras can comprehensively capture workpiece geometric feature information, effectively overcoming the information loss caused by traditional single-view acquisition and providing high-quality workpiece feature maps for subsequent accurate identification. A dual-stream feature extraction architecture using a perception region recognition network achieves the coordinated extraction of macroscopic geometric features and microscopic texture details, addressing the information loss inherent in traditional methods when processing sharp angles and fine features, and accurately identifying target areas requiring differentiated processing. By complementing and fusing global and local features, the system can pay special attention to small feature areas such as sharp corners and fine holes, accurately characterizing complex workpiece contours and addressing the inadequate recognition of small features by traditional methods. Intelligent path planning based on enhanced feature maps and importance weight maps enables differentiated control point density settings and feed rate adjustments for regions with different geometric characteristics, ensuring a stable and efficient cutting process and significantly improving the quality of complex contour processing. By establishing a multi-objective optimization model that comprehensively considers geometric and process constraints, the system achieves coordinated optimization of cutting paths and EDM parameters, balancing the trade-offs between machining accuracy, machining time, and energy consumption, enabling the system to achieve optimal process matching for workpieces of varying materials and geometric complexity.

[0036] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0037] Use an industrial camera to capture high-resolution images of the workpiece at multiple angles to obtain a dataset of original workpiece images, and then perform adaptive Gaussian filtering on the dataset to obtain a dataset of de-noised images.

[0038] Performing histogram equalization and geometric correction processing on the denoised image dataset to obtain a corrected image dataset, and segmenting and fusing the corrected image dataset to obtain a workpiece contour map;

[0039] Edge detection processing is performed on the workpiece contour map to obtain an edge intensity map, which is uniformly converted into a standard size and pixel value normalized to obtain a workpiece feature map.

[0040] Specifically, a high-resolution industrial camera system is used to capture multi-angle images of the workpiece being processed. The system consists of a main camera with a resolution of at least 4096 × 3072 pixels, an autofocus lens, a high-color rendering LED ring light source, and a precision rotation angle control platform. During the acquisition process, the camera is fixed to the reference axis above the wire-cutting machine, maintaining its optical axis perpendicular to the workpiece surface. The platform simultaneously rotates the workpiece at multiple angles in equiangular steps, capturing a full-field image at each angle. Over a complete rotation angle sequence, a raw image dataset covering all side profiles and detailed structures of the workpiece is generated. Preliminary denoising is performed on the raw image dataset. An adaptive Gaussian filtering mechanism is introduced to achieve local image smoothing while preserving edge features. For each raw image frame, a local variance-aware strategy is used to dynamically adjust the filter kernel size, using a larger 7 × 7 convolution kernel in noisy areas and a smaller 3 × 3 or 5 × 5 kernel in areas with clearer texture, resulting in a uniform and effectively denoised image dataset. During the filtering process, the kernel function uses a zero-mean two-dimensional Gaussian distribution model, and its standard deviation σ is automatically adjusted based on local brightness changes to ensure edge integrity in structural areas. The denoised image dataset is sequentially input into the image enhancement module, where histogram equalization is performed. This process optimizes grayscale uniformity by redistributing pixel grayscale values, thereby enhancing image contrast in low-light or highly reflective areas, thereby improving the overall grayscale hierarchy and texture clarity. After histogram equalization, a geometric correction mechanism is introduced, utilizing calibration data based on the camera's intrinsic and extrinsic parameter matrices to perform an anti-distortion transformation on the image. Specifically addressing the radial and tangential distortion issues inherent in industrial cameras, this method establishes an inverse mapping between pixel coordinates and world coordinates, and combines a bicubic interpolation algorithm to perform sub-pixel reconstruction of the original pixel positions, generating a geometrically accurately corrected image dataset. After geometric correction, to extract the main contours of the workpiece structure, the image datasets undergo region segmentation and view fusion. Region segmentation employs an improved Otsu thresholding algorithm combined with morphological operations, effectively separating the workpiece from the background and eliminating artifacts such as adhesions, holes, and fractures. For structural fusion between multi-angle images, a multi-image fusion strategy based on affine transformation and feature point registration is employed. Contour lines extracted from different angles are aligned and superimposed to form a continuous, complete workpiece contour map with the potential for 3D depth perception. Edge detection is then performed on the contour map. This edge detection stage utilizes an improved Canny algorithm based on image gradient field analysis. This algorithm considers the intensity gradient at edge points and integrates directional information, response connectivity, and noise-to-background response ratio. An edge intensity map is extracted using non-maximum suppression and a dual-threshold tracking mechanism. This edge intensity map contains the complete workpiece geometric contour, inner and outer boundaries, sharp corner locations, and curvature variation regions, and is a key component in constructing a spatial representation of the workpiece's features.Image normalization is performed on the edge intensity map. Size normalization resizes the original image to a uniform 512×512 pixel size, making it suitable for fixed-input network models and preventing misalignment of network layer input and output dimensions due to size inconsistencies. Pixel value normalization linearly maps grayscale values from the original [0, 255] to [0, 1], making subsequent convolution operations more stable and accelerating network convergence. The edge map after size and pixel normalization constitutes the standardized workpiece feature map.

[0041] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0042] The workpiece feature map is input into the perception region recognition network, which includes a residual feature extraction branch and a texture preservation branch;

[0043] The workpiece feature map is progressively extracted with deep features through the four-level cascade residual block in the residual feature extraction branch to obtain the main feature map;

[0044] The workpiece feature map is input into the texture preservation branch and processed by cascading three levels of convolutional layers of different scales and multi-scale filtering to obtain an auxiliary feature map;

[0045] Based on the heterogeneous feature pyramid technology, a multi-level skip connection path between the main feature map and the auxiliary feature map is established to obtain an enhanced feature set;

[0046] The enhanced feature set is subjected to a channel adaptive weight calculation mechanism preset by the wire cutting process to obtain a process optimized feature representation;

[0047] The process optimization feature representation is input into the pixel-level classification decoder for boundary positioning and target area screening to obtain the target processing area identification map.

[0048] Specifically, a dual-branch perceptual region recognition network with geometric perception, texture preservation, and region discrimination capabilities was constructed. Taking a standardized workpiece feature map as input, this network aims to automatically identify key regions suitable for wire-cut machining (WEDM) in the image and provide accurate and coherent descriptions of their boundaries. The network consists of two parallel functional subnetworks: a residual feature extraction branch that extracts deep semantic structure from the image, and a texture preservation branch that preserves local high-frequency details and edge texture information. In the residual feature extraction branch, the input image is subjected to layer-by-layer feature extraction via a four-level cascade of residual blocks. Each residual block consists of a convolutional layer followed by batch normalization, ReLU activation, and shortcut connections. Spatial features at different scales are extracted through alternating combinations of 1×1 and 3×3 convolutional kernels. The residual structure prevents gradient vanishing and feature degradation, ensuring information transparency. As the network deepens, this branch effectively captures large-scale structural relationships, geometric configuration changes, and semantic aggregation effects of the target region, forming a highly expressive main feature map. At the same time, the original workpiece feature map is input into the texture-preserving branch, which consists of three cascaded convolutional layers of different scales, using 3×3, 5×5, and 7×7 convolution kernels, respectively, to capture texture details in different receptive fields. This branch is supplemented by a multi-scale filtering processing module. Local response normalization and local directional enhancement mechanisms are used to enhance boundary details and texture discontinuities, thereby enhancing the ability to respond to process-sensitive areas such as sharp corners, holes, and grooves while maintaining spatial resolution. The output is an auxiliary feature map that is extremely sensitive to edge contours. To achieve feature synergy and semantic complementarity between the two branches, a fusion mechanism based on a heterogeneous feature pyramid structure is adopted to construct a multi-level skip connection pathway between the main and auxiliary feature maps. Edge information in the shallow auxiliary feature map is directly transferred to the mid- and high-level semantic maps of the main branch via skip connections. After upsampling and alignment of the feature channels, a cascaded splicing operation is performed to form an enhanced feature set. Based on the enhanced feature set, a channel adaptive weight calculation mechanism is introduced. By performing global average pooling on each channel feature map and sending the pooling result into a two-layer fully connected neural network and a softmax normalization layer, the relative importance coefficient of each channel in the feature expression is calculated. The channel attention mechanism dynamically reweights the features according to the different workpiece geometric structures and texture complexity, improves the response of key channels and suppresses redundant information, forming a process optimization feature representation.This process optimization feature representation is then fed into a pixel-level classification decoder built into the decoding path. This decoder primarily consists of a symmetrical upsampling module, a skip connection module, and a pixel-by-pixel Softmax classifier. The decoder's workflow involves progressively upsampling the deep process optimization features back to the original image size. At each level, local feature maps from shallow layers in the encoding path are fused together to restore spatial positioning accuracy and enhance boundary semantic consistency. A 1×1 convolution is then used to construct a channel compression layer, outputting the classification probability of each pixel under the "processing / non-processing" label. This is normalized using a Softmax function, so that the corresponding probability value for each pixel lies between [0, 1], representing the confidence level that the pixel belongs to the target processing area. The entire pixel classification output is then mapped into a target processing area identification map of the same size as the original image, where each pixel value represents its confidence level in being determined to be a processable area.

[0049] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0050] Extract N local patch images centered on the target point from the target processing area recognition map to form a workpiece complex feature patch set;

[0051] The complex feature patch set of the workpiece is input into the identity-related feature reshaping module of the global and local attention supplementary branches for feature weight analysis to obtain the feature importance weight matrix;

[0052] The target processing area identification map is input into the initial feature extraction layer of the enhanced hierarchical feature fusion branch for feature extraction to obtain a basic feature map;

[0053] The basic feature map is input into the three-layer hierarchical feature extraction layer of the enhanced hierarchical feature fusion branch for processing to obtain a multi-scale feature set;

[0054] The multi-scale feature sets are differentiated and weighted integrated to obtain a fused feature representation, and the feature importance weight matrix and the fused feature representation are combined across branches to obtain an enhanced feature map and an importance weight map.

[0055] Specifically, representative geometric keypoints are extracted from the target processing region identification map as the centers for local patch extraction. The target processing region identification map represents the probability of each pixel belonging to the processing region in the form of a confidence score. A confidence threshold is set to screen out pixels with high confidence and complex structure. Criteria such as edge gradient, curvature extremum, and connected domain center are then used to extract N representative target points within structurally salient regions. Centered on each target point, a corresponding image patch is cropped from the original image using a fixed-size window to construct a set of complex workpiece feature patches covering typical complex structures (such as junctions, closed holes, and sharp corners). This set of complex workpiece feature patches is then fed into an identity-related feature reconstruction module, which consists of two complementary branches: global and local attention. The local path extracts fine-grained texture features from each patch through stacked convolutional layers, while the global path extracts structural representations and geometric contextual information from the patch through spatial compression, self-attention, and channel-attention mechanisms. In the local branch, three layers of convolutional kernels are arranged in an alternating pattern of 3×3, 5×5, and 3×3, respectively, and a Leaky ReLU activation function is used to enhance responsiveness to edge and texture changes. In the global branch, each patch undergoes global average pooling and is compressed into a channel vector. This is then fed into two fully connected layers to extract inter-channel nonlinear mapping features. A sigmoid function is used to generate channel weight coefficients, which are then back-projected back into the original space to form a weight map. The local path emphasizes boundary sensitivity, while the global path emphasizes structural consistency. The two are fused through feature stacking and channel fusion to generate a unified feature importance weight matrix. Simultaneously, the target processing region identification map is fed into the initial feature extraction layer of the enhanced hierarchical feature fusion branch. This layer is a standard convolutional layer. Using multi-channel convolutional kernels, the entire image is encoded and modeled in the first round, extracting a basic feature map that includes boundary distribution, grayscale gradients, and region shape. This map preserves the spatial structural resolution of the image. The basic feature map is input into a three-layer hierarchical feature extraction module. This module uses a densely connected structure to construct a progressive feature path. The intermediate features output by each layer are not only fed into the next layer but also transmitted to higher layers via skip connections, thus maintaining the multi-layer transmission capability of local details. The first layer uses a small convolution kernel size to capture microstructural features, the second layer focuses on shape region modeling, and the third layer focuses on integrating contextual structure with large-scale semantics. This progressive relationship forms a multi-scale feature set, in which each scale feature map corresponds to different structural scale information. Differentiated weighted integration is performed on the multi-scale feature set. Channel pooling and channel attention coefficient calculation are performed on the output feature maps of each layer. They are then mapped to a common dimension using a soft fusion mechanism. Differentiated fusion coefficients are generated based on the response significance and spatial activity of different channels. A weighted accumulation mechanism is used to generate a fused feature representation. The fused feature representation is then combined with the feature importance weight matrix across branches.A channel-by-channel multiplication is performed on the channel features at each spatial location in the fused feature map, multiplied by the channel weights of the corresponding weight matrix to form a weighted feature response. Normalization is then performed across the spatial dimension to construct a feature response map with consistent response strength, generating an enhanced feature map. The output channel weight coefficient map from this process is retained as an importance weight map, which records a comprehensive expression of the participation, salience, and geometric complexity of each area of the workpiece during the feature generation process.

[0056] In a specific embodiment, the step of inputting the basic feature map into the three-layer hierarchical feature extraction layer of the enhanced hierarchical feature fusion branch for processing to obtain the multi-scale feature set may specifically include the following steps:

[0057] The basic feature map is input into the first hierarchical feature extraction layer of the enhanced hierarchical feature fusion branch for feature decomposition to obtain the first layer feature map;

[0058] Apply residual feature expansion processing to the first layer feature map, extract multi-scale texture information by alternating 1×1 and 3×3 convolution kernels to obtain the first layer enhanced feature map;

[0059] The first-layer enhanced feature map is input into the second-layer hierarchical feature extraction layer of the enhanced hierarchical feature fusion branch, and a skip connection channel is constructed from the first-layer enhanced feature map to the third layer to obtain the second-layer feature map and its skip path;

[0060] Perform self-attention feature filtering on the second-layer feature map to obtain the second-layer attention weighted feature map;

[0061] The skip path data of the second-layer attention weighted feature map and the first-layer enhanced feature map are input into the third-layer hierarchical feature extraction layer of the enhanced hierarchical feature fusion branch for feature complementary fusion to obtain the third-layer fusion feature map;

[0062] Multi-scale feature aggregation is performed on the first-layer enhanced feature map, the second-layer attention-weighted feature map, and the third-layer fusion feature map to obtain a multi-scale feature set.

[0063] Specifically, the base feature map is fed into the first-layer hierarchical feature extraction module of the enhanced hierarchical feature fusion branch. This module serves as the low-level entry point for the entire fusion architecture, performing primary structural decoupling and feature decomposition. Standard convolutional layers extract shallow geometric texture variations, boundary contour trends, and initial spatial relationships, forming a first-layer feature map containing coarse-grained semantic contours. Residual feature expansion is applied to the first-layer feature map, constructing a chain of local residual units by alternating 1×1 and 3×3 convolutional kernels. The 1×1 convolutional kernel in this architecture compresses and maps the channel dimension, reducing parameter size and improving nonlinear transformation capabilities. The 3×3 convolutional kernel captures spatial structural relationships and enhances the perception of small-scale features such as pore edges and corner points. While alternating these two convolutional kernels, a standard residual connection is introduced, fusing the input feature map with the convolutional stack output element-wise. This approach preserves low-level feature representation pathways while accumulating deep-level responses, generating an enhanced first-layer feature map with enhanced spatial resolution and texture discrimination. The first-layer enhanced feature map is fed into the second-layer hierarchical feature extraction layer of the enhanced hierarchical feature fusion branch. A skip connection is constructed internally from the first-layer enhanced feature map to the subsequent third-layer fusion module. This is used to compensate for low-level structural information at higher semantic levels and prevent semantic loss. The second-layer hierarchical extraction layer itself utilizes a deeper receptive field and a more complex channel combination strategy to enhance the recognition of mesoscale structure, making it suitable for identifying transitional regions within regional structures, such as surface gradients and boundary inflections. To ensure feature selectivity in this extraction process, a self-attention mechanism is applied to the output second-layer feature map. This mechanism constructs channel description vectors through global average pooling, which are then fed into a set of nonlinear transformation networks to generate channel response weights. These weights are applied to the original feature map to form a second-layer attention-weighted feature map. The second-layer attention-weighted feature map and the first-layer enhanced feature map are fed into the third-layer hierarchical feature extraction layer via a skip channel. The third layer, as the highest layer of the fusion branch, focuses on modeling semantic configuration and regional layout. Its internal structure introduces a feature complementation mechanism, that is, channel alignment, size standardization and feature splicing operations are performed on the two input feature maps respectively. On this basis, the shallow local features are integrated with the deep global expression through the convolutional fusion structure, which improves the synergy of features at different scales and structural levels, and obtains the third-layer fused feature map. Multi-scale feature aggregation operation is performed on the feature maps of the above three key levels. This operation performs channel compression processing on the first-layer enhanced feature map, the second-layer attention-weighted feature map and the third-layer fused feature map. After unifying the dimensions, they are fused according to the weighted strategy in the spatial position. The fusion coefficient is set according to the source level of the feature map and its local response strength. At the same time, the regional weighting parameter is introduced in combination with the spatial attention mechanism to realize the redistribution of spatial dimension information and form a multi-scale feature set.

[0064] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0065] Perform edge extraction on the enhanced feature map to obtain multiple initial edge points, and perform weighted adjustment on the multiple initial edge points in combination with the importance weight map to obtain a first edge point set;

[0066] Prioritizing the first edge point set to obtain a second edge point set, and performing curve fitting processing on the second edge point set to obtain a first contour curve;

[0067] Performing control point density subdivision processing on the first contour curve based on the importance weight map to obtain a second contour curve;

[0068] According to the second contour curve, the entry point is optimized by comprehensively considering the material thickness gradient, edge curvature distribution and internal stress distribution of the workpiece to obtain the optimal starting position for processing;

[0069] The feed rate is planned according to the geometric characteristics of the second contour curve and the regional weight distribution of the importance weight map, and a cutting feed rate distribution function is constructed;

[0070] An initial cutting path is generated based on the second contour curve, the optimal machining starting position and the cutting feed rate distribution function.

[0071] Specifically, the workpiece contour boundary is extracted from the enhanced feature map. An improved edge detection algorithm leverages the grayscale gradient information in the image to identify areas of drastic edge changes. Initial edge points are extracted, distributed along the workpiece's outer contour boundary, inner hole boundary, and localized geometric abrupt changes, forming an edge point set. These edge points are then combined with an importance weight map. By reading the response value of each edge point in the weight map, the spatial position and processing weight of the edge point are fine-tuned, shifting edge points in high-weight areas toward the structural core and giving them higher priority, while edge points in low-weight areas are slightly weakened. This results in a first set of edge points. This first set of edge points is then sorted. Based on the spatial distribution of edge points in the image, and incorporating multiple factors such as each point's process weight, curvature characteristics, and connectivity with surrounding points, the sorting results ensure that a logically closed and contoured point sequence is formed in the subsequent curve fitting process. Based on this, a second set of edge points is fitted to a geometrically continuous first contour curve using curve fitting techniques. The distribution density of the curve's control points is readjusted based on the regional response information in the importance weight map. In high-weighted areas, such as sharp corners, slits, and micropores, more control points are added to improve fitting accuracy. In flat areas, the number of control points is appropriately simplified to reduce computational redundancy, resulting in a second contour curve with greater structural adaptability and enhanced machining guidance. Based on the second contour curve, entry point optimization is performed by comprehensively considering the material thickness gradient, edge curvature distribution, and internal stress distribution within the workpiece. The starting position avoids areas of sudden thickness changes, high-curvature corners, and stress concentrations, prioritizing areas with stable structures, gentle curvature, and low stress gradients to ensure balanced heat distribution and good electrode wire stability during the entry process, thereby selecting the optimal machining starting point. A cutting feed rate scheduling strategy is developed based on the geometric characteristics of the second contour curve and the response strength of the corresponding areas in the importance weight map. In geometrically complex areas with high weights, the feed rate is appropriately reduced to improve machining accuracy and avoid problems such as ablation, wire drawing, and edge burrs. In areas with simple structures, smooth boundaries, and low weights, the feed rate is increased to enhance machining efficiency. The speed scheduling mechanism is dynamically adjusted based on each position point of the path, so that the entire machining process can meet both structural accuracy and optimize overall time consumption. Based on the second contour curve, the initial cutting path is constructed by combining the determined starting position and the corresponding feed speed scheduling information. This path includes the center motion trajectory of the electrode wire, and also performs appropriate trajectory compensation according to the machining requirements of different areas, and presets different feed speeds. Through the above steps, the initial path obtained has continuity in structure, adaptability in process, and flexibility in speed control.

[0072] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0073] A multi-objective optimization model is established by taking machining accuracy requirements, workpiece contour tolerance, electrode wire tension limit and maximum cutting speed as constraints;

[0074] Discretize the image features of the enhanced feature map to obtain multiple grid units, and assign different weight values to each grid unit according to the importance weight map to obtain the feature importance matrix;

[0075] Particle swarm optimization is performed on the initial cutting path based on the feature importance matrix and the multi-objective optimization model. The particle encoding contains the coordinates of the path control points and the cutting parameters. An adaptive inertia weight strategy is introduced to obtain an iterative path optimization scheme.

[0076] Adding a curvature penalty factor to the path optimization iteration scheme limits the curvature change of adjacent path segments to obtain the target cutting path;

[0077] A material-parameter mapping function is constructed based on the workpiece material characteristic parameter matrix and the feature importance matrix. The optimal voltage and current parameter combinations at different positions of the target cutting path are dynamically calculated through the material-parameter mapping function to obtain the EDM parameter configuration table.

[0078] Specifically, a constraint system is defined based on actual engineering requirements. The machining accuracy requirement determines the upper limit of the density of path control points and the lower limit of fitting accuracy. The workpiece contour tolerance determines the maximum allowable deviation from the original contour. The wire tension limit constrains the path curvature, minimum bend radius, and angular inflection strength. The maximum cutting speed limits the upper limit of the machining advance rate at each segment along the path. These multiple factors not only impose specific constraints on the spatial geometry but also form linked constraints on trajectory smoothness, velocity stability, and electrical parameter rationality at the dynamic execution level of the process. Based on these parameters, a multi-objective optimization model is established with the core objectives of minimizing path smoothness, minimizing machining time, minimizing energy consumption, and maximizing process stability. Each optimization objective is assigned a corresponding weight during the solution process, allowing the system to adjust the optimization focus based on the priority of different machining stages. Image features are discretized within the enhanced feature map. The entire enhanced feature map is divided into uniform two-dimensional grid cells, forming a regular grid structure. Within each grid cell, the texture complexity, gradient response, and structural mutation characteristics of the image region are extracted. A processing weight is then applied to each cell using an importance weight map, ensuring that high-complexity or critical processing areas receive greater attention and sensitivity to parameter adjustments during the optimization process. This process results in a feature importance matrix based on image coordinates and weighted by process priorities. The initial cutting path is optimized using the feature importance matrix and the established multi-objective optimization model as input. In this stage, a particle swarm optimization algorithm is introduced as the core search tool. The coordinates of path control points and process parameter values are co-encoded into particle vectors, forming optimization units that can be iteratively adjusted in the search space. Each particle updates its position and velocity in each iteration based on its historical optimal state and the global optimal state. Simultaneously, its comprehensive fitness is evaluated in the path space under multiple objective indicators, including machining accuracy, energy consumption, and velocity smoothness. In order to improve convergence efficiency and avoid falling into local optimality, an adaptive inertia weight mechanism is introduced. That is, as the number of iterations increases, the inertia factor of the particle gradually decays, so that the early search has the ability of broad exploration and the later convergence has the ability of local refinement, thereby improving the globality and stability of the overall optimization results. During the execution of particle swarm optimization, in order to prevent geometric mutations or physically unprocessable states in the path, an additional curvature penalty term is added to the fitness function to limit the curvature fluctuation value between adjacent control points of the path. The path sequence corresponding to the particle is geometrically analyzed to identify the angle changes and curvature deviations between continuous segments. When the curvature mutation of a certain segment exceeds the preset threshold, its fitness is automatically penalized, forcing the optimization algorithm to complete the search update while maintaining the smoothness of the path. The introduction of the curvature penalty term ensures that the target cutting path obtained not only meets the structural accuracy requirements, but also has a good wire tension control foundation and path continuity advantages at the physical execution level.After the target path is determined, the optimal electrical parameter combination is configured for each location along the path based on the properties of the workpiece material. A material-parameter mapping model is then constructed. This model uses workpiece material parameters such as conductivity, thermal conductivity, hardness, melting point, and specific heat capacity as input variables. Combined with the response values of the target path location in the feature importance matrix, the two types of information are combined to generate a mapping function. This function is implemented using empirical formulas, interpolation methods, or lightweight neural networks. Its output is the optimal discharge control parameters, such as voltage, current, or pulse width, for the current region. The goal of this mapping mechanism is to allocate low-energy, high-precision discharge strategies to high-importance areas, while allowing high-energy, high-speed cutting strategies to low-complexity areas. This maximizes overall cutting efficiency while ensuring boundary accuracy and surface quality. After completing the above path space optimization and parameter space configuration, a target cutting path and EDM parameter configuration table is generated. This table is indexed by point along the path, and each point contains key information such as spatial position, target speed, voltage and current parameters, and discharge timing settings.

[0079] The above describes the wire cutting machine cutting path intelligent planning method based on image processing in the embodiment of the present invention. The following describes the wire cutting machine cutting path intelligent planning system based on image processing in the embodiment of the present invention. Figure 2 In one embodiment of the present invention, an intelligent planning system for cutting paths of a wire cutting machine based on image processing includes:

[0080] The acquisition module 201 is used to acquire and pre-process multi-angle images of the workpiece to be processed to obtain a workpiece feature map;

[0081] The region recognition module 202 is used to input the workpiece feature map into the perception region recognition network for processing to obtain a target processing region recognition map;

[0082] A feature enhancement module 203 is used to perform feature enhancement on the target processing area identification map to obtain an enhanced feature map and an importance weight map;

[0083] A path planning module 204 is configured to plan a cutting path based on the enhanced feature map and the importance weight map to obtain an initial cutting path;

[0084] The multi-objective optimization module 205 is used to perform multi-objective optimization on the initial cutting path to obtain a target cutting path and a corresponding EDM parameter configuration table.

[0085] Through the collaborative efforts of the aforementioned components, multi-angle acquisition and adaptive preprocessing technology using industrial cameras can comprehensively capture workpiece geometric feature information, effectively overcoming the information loss caused by traditional single-view acquisition and providing high-quality workpiece feature maps for subsequent precise identification. A dual-stream feature extraction architecture using a perception region recognition network enables the coordinated extraction of macroscopic geometric features and microscopic texture details, resolving the information loss problem associated with traditional methods when processing sharp angles and small features, and accurately identifying target areas requiring differentiated processing. Through the supplementation and fusion of global and local features, the system can pay special attention to small feature areas such as sharp corners and fine holes, accurately depicting the complex contour features of the workpiece and addressing the inadequate recognition of small features by traditional methods. Intelligent path planning based on enhanced feature maps and importance weight maps enables differentiated control point density settings and feed speed adjustments for areas with different geometric characteristics, ensuring a stable and efficient cutting process and significantly improving the quality of complex contour processing. By establishing a multi-objective optimization model that comprehensively considers geometric constraints and process constraints, the coordinated optimization of cutting path and EDM parameters is achieved, and the contradictions between machining accuracy, machining time and energy consumption are balanced, so that the system can achieve optimal process matching on workpieces of different materials and geometric complexity.

[0086] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0087] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for enabling an image processing-based wire cutting machine cutting path intelligent planning device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program code.

[0088] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that the technical solutions described in the above embodiments can still be modified, or some of the technical features thereof can be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. An intelligent planning method for cutting paths of wire cutting machines based on image processing, characterized in that: include: Perform multi-angle image acquisition and preprocessing on the workpiece to be processed to obtain the workpiece feature map; The workpiece feature map is input into the perception area recognition network for processing to obtain the target processing area recognition map; Perform feature enhancement on the target processing area recognition map to obtain an enhanced feature map and an importance weight map; Cutting path planning is performed according to the enhanced feature map and the importance weight map to obtain an initial cutting path, including: edge extraction of the enhanced feature map to obtain multiple initial edge points, and weighted adjustment of the multiple initial edge points in combination with the importance weight map to obtain a first edge point set; priority sorting of the first edge point set to obtain a second edge point set, and curve fitting processing of the second edge point set to obtain a first contour curve; control point density subdivision processing of the first contour curve based on the importance weight map to obtain a second contour curve; based on the second contour curve, entry point optimization is performed by comprehensively considering the material thickness change gradient, edge curvature distribution and internal stress distribution factors of the workpiece to obtain the optimal processing starting position; feed speed planning is performed according to the geometric characteristics of the second contour curve and the regional weight distribution of the importance weight map, and a cutting feed speed distribution function is constructed; an initial cutting path is generated based on the second contour curve, the optimal processing starting position and the cutting feed speed distribution function; The initial cutting path is optimized with multiple objectives to obtain the target cutting path and the corresponding EDM parameter configuration table, including taking the machining accuracy requirements, workpiece contour tolerance, electrode wire tension limit and maximum cutting speed as constraints to establish a multi-objective optimization model; the image features of the enhanced feature map are discretized to obtain multiple grid units, and different weight values are assigned to each grid unit according to the importance weight map to obtain a feature importance matrix; particle swarm optimization is performed on the initial cutting path based on the feature importance matrix and the multi-objective optimization model, and the particle encoding contains the coordinates of the path control points and cutting parameters, and an adaptive inertia weight strategy is introduced to obtain a path optimization iterative scheme; a curvature penalty factor is added to the path optimization iterative scheme to limit the curvature change of adjacent path segments to obtain the target cutting path; a material-parameter mapping function is constructed based on the workpiece material characteristic parameter matrix and the feature importance matrix. The function is implemented through empirical formulas, interpolation methods or lightweight neural networks, and the optimal voltage, current or pulse width parameter combinations at different positions of the target cutting path are dynamically calculated through the material-parameter mapping function to obtain the EDM parameter configuration table.

2. The method for intelligent planning of cutting paths of wire cutting machines based on image processing according to claim 1, characterized in that: Perform multi-angle image acquisition and preprocessing on the workpiece to obtain the workpiece feature map, including: Use an industrial camera to capture high-resolution images of the workpiece at multiple angles to obtain a dataset of original workpiece images, and then perform adaptive Gaussian filtering on the dataset to obtain a dataset of de-noised images. Performing histogram equalization and geometric correction processing on the denoised image dataset to obtain a corrected image dataset, and segmenting and fusing the corrected image dataset to obtain a workpiece contour map; Edge detection processing is performed on the workpiece contour map to obtain an edge intensity map, which is uniformly converted into a standard size and pixel value normalized to obtain a workpiece feature map.

3. The method for intelligent planning of cutting paths of wire cutting machines based on image processing according to claim 1, characterized in that: The workpiece feature map is input into the perception area recognition network for processing to obtain the target processing area recognition map, including: The workpiece feature map is input into the perception region recognition network, which includes a residual feature extraction branch and a texture preservation branch; The workpiece feature map is progressively extracted with deep features through the four-level cascade residual block in the residual feature extraction branch to obtain the main feature map; The workpiece feature map is input into the texture preservation branch and the three-level convolution layers of different scales are cascaded to perform multi-scale filtering processing to obtain an auxiliary feature map; Based on the heterogeneous feature pyramid technology, a multi-level skip connection path between the main feature map and the auxiliary feature map is established to obtain an enhanced feature set; The enhanced feature set is subjected to a channel adaptive weight calculation mechanism preset by the wire cutting process to obtain a process optimized feature representation; The process optimization feature representation is input into the pixel-level classification decoder for boundary positioning and target area screening to obtain the target processing area identification map.

4. The method for intelligent planning of cutting paths of wire cutting machines based on image processing according to claim 1, characterized in that: Perform feature enhancement on the target processing area recognition map to obtain an enhanced feature map and an importance weight map, including: Extract N local patch images centered on the target point from the target processing area recognition map to form a workpiece complex feature patch set; The complex feature patch set of the workpiece is input into the identity-related feature reshaping module of the global and local attention supplementary branches for feature weight analysis to obtain the feature importance weight matrix; The target processing area identification map is input into the initial feature extraction layer of the enhanced hierarchical feature fusion branch for feature extraction to obtain a basic feature map; The basic feature map is input into the three-layer hierarchical feature extraction layer of the enhanced hierarchical feature fusion branch for processing to obtain a multi-scale feature set; The multi-scale feature sets are differentiated and weighted integrated to obtain a fused feature representation, and the feature importance weight matrix and the fused feature representation are combined across branches to obtain an enhanced feature map and an importance weight map.

5. The method for intelligent planning of cutting paths of wire cutting machines based on image processing according to claim 4, characterized in that: The basic feature map is input into the three-layer hierarchical feature extraction layer of the enhanced hierarchical feature fusion branch for processing to obtain a multi-scale feature set, including: The basic feature map is input into the first hierarchical feature extraction layer of the enhanced hierarchical feature fusion branch for feature decomposition to obtain the first layer feature map; Apply residual feature expansion processing to the first layer feature map, extract multi-scale texture information by alternating 1×1 and 3×3 convolution kernels to obtain the first layer enhanced feature map; The first-layer enhanced feature map is input into the second-layer hierarchical feature extraction layer of the enhanced hierarchical feature fusion branch, and a skip connection channel is constructed from the first-layer enhanced feature map to the third layer to obtain the second-layer feature map and its skip path; Perform self-attention feature filtering on the second-layer feature map to obtain the second-layer attention weighted feature map; The skip path data of the second-layer attention weighted feature map and the first-layer enhanced feature map are input into the third-layer hierarchical feature extraction layer of the enhanced hierarchical feature fusion branch for feature complementary fusion to obtain the third-layer fusion feature map; Multi-scale feature aggregation is performed on the first-layer enhanced feature map, the second-layer attention-weighted feature map, and the third-layer fusion feature map to obtain a multi-scale feature set.

6. An intelligent cutting path planning system for wire cutting machine based on image processing, characterized in that: For implementing the method for intelligent planning of a wire cutting machine cutting path based on image processing according to any one of claims 1 to 5, the intelligent planning system for a wire cutting machine cutting path based on image processing comprises: An acquisition module is used to acquire and pre-process multi-angle images of the workpiece to be processed to obtain a workpiece feature map; The region recognition module is used to input the workpiece feature map into the perception region recognition network for processing to obtain the target processing region recognition map; A feature enhancement module is used to perform feature enhancement on the target processing area recognition map to obtain an enhanced feature map and an importance weight map; A path planning module is used to plan the cutting path based on the enhanced feature map and the importance weight map to obtain the initial cutting path; The multi-objective optimization module is used to perform multi-objective optimization on the initial cutting path to obtain the target cutting path and the corresponding EDM parameter configuration table.

Images