An edge feature matching-based irregular glass cutting method and system

Abstract

The application provides a special-shaped glass cutting method and system based on edge feature matching, and belongs to the technical field of glass processing. The special-shaped glass cutting method based on edge feature matching comprises the following steps: collecting multi-view images of glass to be cut to generate an original image dataset containing complete edge contour information; performing image enhancement and denoising processing on the original image dataset to extract a clear edge pixel set; constructing an edge feature vector based on the edge pixel set and generating an edge key point descriptor through a scale invariant feature transformation algorithm; matching the edge key point descriptor with a template descriptor in a preset reference feature library to determine an optimal matching template and corresponding local geometric transformation parameters; and performing spatial correction on the optimal matching template according to the local geometric transformation parameters to generate an initial cutting path.

Description

Technical Field

[0001] This invention belongs to the field of glass processing technology and relates to a method and system for cutting irregularly shaped glass based on edge feature matching. Background Technology

[0002] Due to their unique contours and functional adaptability, irregularly shaped glass has been widely used in fields such as architectural decoration, automobile manufacturing, and consumer electronics. With the continuous rise in demand for personalized customization, higher requirements have been placed on cutting precision and edge integrity.

[0003] Traditional processing methods typically rely on preset trajectories or physical molds for positioning and cutting, which is suitable for mass standardized production, but its adaptability is limited when dealing with parts without drawings or non-standard irregular shapes.

[0004] In recent years, intelligent cutting technology that integrates machine vision and CNC systems has gradually developed, and some solutions have attempted to introduce image processing algorithms to assist in path generation.

[0005] For example, the laser cutting trajectory planning method for irregular curved glass with publication number CN120734562A (publication date: October 3, 2025) achieves thermal deformation compensation by coupling instantaneous and historical thermal effect modeling, and dynamically optimizes the scanning strategy based on local geometric complexity, thus making progress in improving the accuracy of laser cutting.

[0006] However, the trajectory generation of this method still relies on pre-established CAD models or offline acquired 3D point cloud data, and does not integrate an online perception and feature matching mechanism for the actual glass edge contour. In application scenarios lacking standard templates or original design files, such as glass processing for artistic creation or cultural relic restoration, its responsiveness to real edge morphology is limited, which restricts the realization of adaptive cutting. Existing technologies still need to be improved in terms of robust extraction of edge features, real-time matching accuracy, and dynamic path generation. Summary of the Invention

[0007] The purpose of this invention is to address the aforementioned problems in existing technologies by proposing a method for cutting irregularly shaped glass based on edge feature matching.

[0008] The objective of this invention can be achieved through the following technical solutions:

[0009] In a first aspect, to solve the above-mentioned technical problems, the present invention provides a method for cutting irregularly shaped glass based on edge feature matching, comprising:

[0010] Multi-view image acquisition is performed on the glass to be cut to generate a raw image dataset containing complete edge contour information;

[0011] Image enhancement and denoising processing are performed on the original image dataset to extract a set of clear edge pixels;

[0012] An edge feature vector is constructed based on the set of edge pixels, and an edge key point descriptor is generated by a scale-invariant feature transformation algorithm.

[0013] The edge key point descriptor is matched with the template descriptor in the preset reference feature library to determine the optimal matching template and the corresponding local geometric transformation parameters;

[0014] The optimal matching template is spatially corrected based on the local geometric transformation parameters to generate an initial cutting path;

[0015] The initial cutting path is optimized based on edge continuity constraints to generate the final cutting path;

[0016] The final cutting path is sent to the CNC cutting equipment, which controls the laser head to perform the cutting operation along the final cutting path.

[0017] Preferably, the glass to be cut is subjected to multi-view image acquisition to generate an original image dataset containing complete edge contour information, including:

[0018] Place the glass to be cut on an image acquisition platform with uniform backlighting;

[0019] The industrial camera is controlled to take pictures at equal angle intervals around the glass to be cut along a preset circular track to obtain two-dimensional grayscale images from multiple perspectives.

[0020] Perspective correction is performed on images from each viewpoint, and then an image stitching algorithm is used to fuse them into a panoramic edge image;

[0021] The panoramic edge image is subjected to resolution normalization processing to generate the original image dataset.

[0022] Preferably, image enhancement and denoising processing is performed on the original image dataset to extract a set of clear edge pixels, including:

[0023] An adaptive histogram equalization algorithm is applied to the original image dataset to improve local contrast.

[0024] A bilateral filter is used to perform edge-preserving smoothing on the enhanced image;

[0025] Preliminary edge pixels are extracted using the Canny edge detection operator;

[0026] A morphological closing operation is performed on the initial edge pixels to connect the broken edges, and isolated noise regions with an area smaller than a preset threshold are removed to generate the edge pixel set.

[0027] Preferably, an edge feature vector is constructed based on the set of edge pixels, and an edge key point descriptor is generated using a scale-invariant feature transformation algorithm, including:

[0028] A Gaussian difference pyramid is constructed on the edge pixel set to detect extreme points in the scale space;

[0029] Remove low-contrast and unstable extreme points located in the edge response direction, and retain stable key points;

[0030] Assign a principal direction to each stable keypoint and calculate a gradient direction histogram in its neighborhood to generate a 128-dimensional edge keypoint descriptor.

[0031] All edge key point descriptors are sorted according to their spatial location to form the edge feature vector.

[0032] Preferably, the edge key point descriptor is matched with template descriptors in a preset reference feature library to determine the optimal matching template and the corresponding local geometric transformation parameters, including:

[0033] Load a subset of template descriptions from multiple historical successful cutting cases from the preset reference feature library;

[0034] The nearest neighbor distance ratio criterion is used to perform coarse matching between the edge key point descriptors and each template descriptor;

[0035] The random sampling consensus algorithm is applied to the coarse matching results to remove mismatched pairs;

[0036] The affine transformation matrix is ​​calculated based on the remaining correctly matched point pairs and used as the local geometric transformation parameters.

[0037] The template with the smallest reprojection error and the largest number of matching point pairs is selected as the optimal matching template.

[0038] Preferably, spatial correction is performed on the optimal matching template based on the local geometric transformation parameters to generate an initial cutting path, including:

[0039] Read the standard cutting trajectory coordinate sequence stored in the optimal matching template;

[0040] Substitute each coordinate point in the standard cutting trajectory coordinate sequence into the affine transformation matrix to perform coordinate transformation;

[0041] The transformed coordinate sequence is interpolated and encrypted to ensure that the distance between adjacent points is no greater than a preset step size threshold.

[0042] The encrypted coordinate sequence is used as the initial cutting path.

[0043] Preferably, path optimization based on edge continuity constraints is performed on the initial cutting path to generate the final cutting path, including:

[0044] Project the initial cutting path onto the image plane containing the edge pixel set;

[0045] Calculate the Euclidean distance from each point on the path to the nearest edge pixel, and mark abnormal point segments whose deviation distance is greater than the preset tolerance threshold;

[0046] The abnormal point segments are locally reconstructed using the B-spline curve fitting method to make them conform to the actual edge direction;

[0047] A speed look-ahead control algorithm is executed on the reconstructed path to generate a smooth motion command sequence containing acceleration constraints, which serves as the final cutting path.

[0048] Secondly, the present invention provides an irregularly shaped glass cutting system based on edge feature matching, comprising:

[0049] The image acquisition module is used to acquire multi-view images of the glass to be cut and generate a raw image dataset containing complete edge contour information.

[0050] The image preprocessing module is used to perform image enhancement and denoising processing on the original image dataset and extract a set of clear edge pixels;

[0051] The feature extraction module is used to construct an edge feature vector based on the edge pixel set and generate an edge key point descriptor through a scale-invariant feature transformation algorithm.

[0052] The template matching module is used to match the edge key point descriptor with the template descriptor in the preset reference feature library to determine the optimal matching template and the corresponding local geometric transformation parameters.

[0053] The path generation module is used to perform spatial correction on the optimal matching template according to the local geometric transformation parameters and generate an initial cutting path.

[0054] The path optimization module is used to perform path optimization based on edge continuity constraints on the initial cutting path to generate the final cutting path;

[0055] The cutting execution module is used to send the final cutting path to the CNC cutting equipment and control the laser head to perform cutting operations along the final cutting path.

[0056] Thirdly, the present invention also provides an electronic device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the irregular glass cutting method based on edge feature matching described in any one of the above.

[0057] Fourthly, the present invention also provides a computer-readable storage medium comprising a stored computer program, wherein, when the computer program is executed, it controls the device on which the computer-readable storage medium is located to perform the irregular glass cutting method based on edge feature matching described above.

[0058] Compared with existing technologies, this invention has the following advantages: By constructing an online perception and feature matching mechanism based on actual edge contours, this invention eliminates the dependence on original CAD models or offline 3D point cloud data, enabling adaptive cutting of what is seen is what is cut in non-standardized processing scenarios without drawings. The image acquisition module adopts a multi-view circular shooting and panoramic stitching strategy to ensure complete capture of complex curved surface edges; the feature extraction module uses a scale-invariant feature transformation algorithm to generate robust key point descriptors, effectively coping with changes in illumination and local occlusion; the template matching module combines nearest neighbor distance ratio and random sampling consistency algorithm to significantly improve matching accuracy and anti-interference ability; the path optimization module introduces edge continuity constraints and B-spline local reconstruction to make the cutting trajectory closely fit the real edge shape, avoiding overcutting or undercutting. The entire system realizes closed-loop adaptive control from physical object to digital path to physical processing, greatly improving the processing accuracy and efficiency of irregularly shaped glass in highly customized scenarios such as artistic creation and cultural relic restoration. Attached Figure Description

[0059] Figure 1 This is a schematic diagram of the process for cutting irregularly shaped glass based on edge feature matching;

[0060] Figure 2 This is a schematic diagram of an irregular glass cutting system module based on edge feature matching.

[0061] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0062] like Figure 1 As shown in the diagram, the first embodiment of the present invention provides a flowchart of a method for cutting irregularly shaped glass based on edge feature matching, including the following steps:

[0063] S11: Acquire multi-view images of the glass to be cut to generate an original image dataset containing complete edge contour information;

[0064] S12, perform image enhancement and denoising processing on the original image dataset to extract a set of clear edge pixels;

[0065] S13, construct an edge feature vector based on the edge pixel set, and generate an edge key point descriptor through a scale-invariant feature transformation algorithm;

[0066] S14, Match the edge key point descriptor with the template descriptor in the preset reference feature library to determine the optimal matching template and the corresponding local geometric transformation parameters;

[0067] S15, Spatial correction is performed on the optimal matching template according to the local geometric transformation parameters to generate an initial cutting path;

[0068] S16, perform path optimization based on edge continuity constraints on the initial cutting path to generate the final cutting path;

[0069] S17, The final cutting path is sent to the CNC cutting equipment to control the laser head to perform the cutting operation along the final cutting path.

[0070] In step S11, the process of acquiring multi-view images of the glass to be cut to generate an original image dataset containing complete edge contour information includes:

[0071] Place the glass to be cut on an image acquisition platform with uniform backlighting;

[0072] The industrial camera is controlled to take pictures at equal angle intervals around the glass to be cut along a preset circular track to obtain two-dimensional grayscale images from multiple perspectives.

[0073] Perspective correction is performed on images from each viewpoint, and then an image stitching algorithm is used to fuse them into a panoramic edge image;

[0074] The panoramic edge image is subjected to resolution normalization processing to generate the original image dataset.

[0075] For example, the image acquisition platform is equipped with a high color rendering index LED backlight, the light-emitting surface of which is parallel to the bottom surface of the glass to be cut, ensuring uniform distribution of transmitted light; the industrial camera is mounted on a circular guide rail driven by a programmable servo motor, the center of the guide rail coincides with the geometric center of the glass, and the optical axis of the camera lens always points perpendicularly to the center of the glass; the shooting interval is 30°, and a total of 12 frames of images are acquired; after homography matrix correction, each frame of image is registered by phase correlation method, and a weighted average fusion strategy is used to generate a distortion-free panoramic image.

[0076] Preferably, the resolution normalization process scales the panoramic image to 2048×2048 pixels to fit the fixed input size requirements of the subsequent image processing module.

[0077] In one possible implementation, if the maximum outer circle diameter of the glass to be cut is 500mm, the radius of the annular track is set to 600mm to ensure that the imaging field of view completely covers the glass outline and retains a 10% edge redundancy area, thus avoiding outline truncation due to positioning deviation.

[0078] In step S12, image enhancement and denoising processing is performed on the original image dataset to extract a set of clear edge pixels, including:

[0079] An adaptive histogram equalization algorithm is applied to the original image dataset to improve local contrast.

[0080] A bilateral filter is used to perform edge-preserving smoothing on the enhanced image;

[0081] Preliminary edge pixels are extracted using the Canny edge detection operator;

[0082] A morphological closing operation is performed on the initial edge pixels to connect the broken edges, and isolated noise regions with an area smaller than a preset threshold are removed to generate the edge pixel set.

[0083] Preferably, the neighborhood window size of the adaptive histogram equalization algorithm is set to 64×64 pixels, and the contrast limiting factor is 3.0; the spatial standard deviation of the bilateral filter... Set to 3.0, grayscale standard deviation Set to 50; set the high and low thresholds of the Canny operator to 50 and 150 respectively; use a 5×5 circular structuring element for the morphological closing operation, and set the area threshold to 20 pixels.

[0084] It should be noted that after the above processing, the broken edges are effectively connected, the tiny holes are filled, and the sharp corners and curvature abrupt changes of the real contour are preserved, providing high-quality input for subsequent feature extraction.

[0085] In step S13, the step of constructing an edge feature vector based on the edge pixel set and generating an edge keypoint descriptor using the Scale Invariant Feature Transform (SIFT) algorithm includes:

[0086] A Gaussian difference pyramid is constructed on the edge pixel set to detect extreme points in the scale space;

[0087] Remove low-contrast and unstable extreme points located in the edge response direction, and retain stable key points;

[0088] Assign a principal direction to each stable keypoint and calculate a gradient direction histogram in its neighborhood to generate a 128-dimensional edge keypoint descriptor.

[0089] All edge key point descriptors are sorted according to their spatial location to form the edge feature vector.

[0090] It is worth noting that during the construction of the Gaussian difference pyramid, the scaling factor... Starting from version 1.0, according to The pyramid structure is constructed by doubling the number of layers, resulting in 4 groups of 8-layer pyramids. The contrast threshold is set to 0.03, and the edge response threshold is set to 10. The main direction is determined by the peak direction in the 36-directional histogram. If multiple peaks exist and the difference is less than 80% of the main peak, multiple keypoint copies are generated. The 128-dimensional descriptor is obtained by dividing the 16×16 neighborhood into 4×4 sub-blocks and calculating an 8-directional gradient histogram for each block.

[0091] In step S14, matching the edge key point descriptor with the template descriptor in the preset reference feature library to determine the optimal matching template and the corresponding local geometric transformation parameters includes:

[0092] Load a subset of template descriptions from multiple historical successful cutting cases from the preset reference feature library;

[0093] The nearest neighbor distance ratio criterion is used to perform coarse matching between the edge key point descriptors and each template descriptor;

[0094] The random sampling consensus algorithm is applied to the coarse matching results to remove mismatched pairs;

[0095] The affine transformation matrix is ​​calculated based on the remaining correctly matched point pairs and used as the local geometric transformation parameters.

[0096] The template with the smallest reprojection error and the largest number of matching point pairs is selected as the optimal matching template.

[0097] Wherein, the nearest neighbor distance ratio threshold is set to 0.75; the number of iterations of the random sampling consensus algorithm is set to 1000, and the inlier distance tolerance is set to 3 pixels; the affine transformation matrix is ​​in the form of:

[0098]

[0099] in, , , , Control linear transformations such as scaling, rotation, and shearing. , The translation transformation is controlled by a matrix that describes the spatial pose relationship between the actual edge and the reference contour. The reprojection error is defined as the root mean square distance between the matching point after the inverse transformation and the original position, and is used to evaluate the accuracy of the spatial transformation matrix.

[0100] Preferably, the preset reference feature library is stored in a non-volatile memory, and each template contains the original edge image, key point descriptor, standard cutting trajectory coordinate sequence and process parameter metadata.

[0101] In step S15, the step of spatially correcting the optimal matching template according to the local geometric transformation parameters to generate an initial cutting path includes:

[0102] Read the standard cutting trajectory coordinate sequence stored in the optimal matching template;

[0103] Substitute each coordinate point in the standard cutting trajectory coordinate sequence into the affine transformation matrix to perform coordinate transformation;

[0104] The transformed coordinate sequence is interpolated and encrypted to ensure that the distance between adjacent points is no greater than a preset step size threshold.

[0105] The encrypted coordinate sequence is used as the initial cutting path.

[0106] In one possible implementation, the standard cutting trajectory coordinate sequence is a discrete point sequence. A new sequence of points is obtained after affine transformation. Cubic spline interpolation is used to refine the transformed trajectory, with a step size threshold set to 0.1 mm to ensure the smoothness of the CNC system's motion.

[0107] In step S16, performing path optimization based on edge continuity constraints on the initial cutting path to generate the final cutting path includes:

[0108] Project the initial cutting path onto the image plane containing the edge pixel set;

[0109] Calculate the Euclidean distance from each point on the path to the nearest edge pixel, and mark abnormal point segments whose deviation distance is greater than the preset tolerance threshold;

[0110] The abnormal point segments are locally reconstructed using the B-spline curve fitting method to make them conform to the actual edge direction;

[0111] A speed look-ahead control algorithm is executed on the reconstructed path to generate a smooth motion command sequence containing acceleration constraints, which serves as the final cutting path.

[0112] The preset tolerance threshold is set to 0.3 mm; the B-spline curve order is set to 3; the control points are fitted with edge pixels in the neighborhood of abnormal point segments using the least squares method; the speed look-ahead control algorithm dynamically adjusts the feed speed according to the curvature change to ensure that the laser power matches the cutting speed.

[0113] In step S17, the final cutting path is sent to the CNC cutting equipment, and the laser head is controlled to perform the cutting operation along the final cutting path.

[0114] The CNC cutting equipment includes a three-axis motion platform, a CO2 laser generator, a focusing lens group, and a closed-loop feedback system. The three-axis motion platform enables high-precision positioning of the laser head, the CO2 laser generator provides cutting energy, the focusing lens group converges the laser beam to the cutting point, and the closed-loop feedback system corrects motion errors in real time. The final cutting path is transmitted to the motion controller in standard G-code format via industrial Ethernet (such as EtherCAT). The motion controller parses the G-code and generates corresponding motion control commands, including process parameters such as laser power, frequency, and auxiliary gas pressure. Based on the thickness and material of the glass to be cut, the system automatically retrieves the optimal parameter combination from a preset process parameter database to ensure cutting quality and efficiency. During the cutting process, the closed-loop feedback system monitors the position and speed of the laser head in real time. If a deviation or abnormality exceeding a preset threshold is detected, the system automatically pauses cutting and issues an alarm to prevent workpiece damage.

[0115] For example, for a 5mm thick soda-lime glass, the laser power is set to 80W, the frequency to 20kHz, and the nitrogen-assisted pressure to 0.3MPa; the maximum acceleration of the motion platform is set to 2g to ensure trajectory accuracy at high-speed corners.

[0116] In summary, this invention acquires complete edge contours through multi-view image acquisition, extracts robust edge features after enhancement and denoising, generates key point descriptors using scale-invariant feature transformation, and performs high-precision matching with historical templates. Based on the matching results, spatial correction is performed to generate an initial path, and then local optimization is performed in combination with actual edge continuity constraints to finally generate a cutting trajectory that fits the real contour. This method does not rely on the original CAD model or 3D point cloud and achieves adaptive cutting that is exactly what you see in the drawing and non-standard irregular glass processing scenarios.

[0117] like Figure 2As shown, the second embodiment of the present invention provides an irregularly shaped glass cutting system based on edge feature matching, comprising:

[0118] Image acquisition module 10 is used to acquire multi-view images of the glass to be cut and generate an original image dataset containing complete edge contour information.

[0119] Image preprocessing module 20 is used to perform image enhancement and denoising processing on the original image dataset and extract a set of clear edge pixels;

[0120] Feature extraction module 30 is used to construct an edge feature vector based on the edge pixel set and generate an edge key point descriptor through a scale-invariant feature transformation algorithm;

[0121] The template matching module 40 is used to match the edge key point descriptor with the template descriptor in the preset reference feature library to determine the optimal matching template and the corresponding local geometric transformation parameters.

[0122] The path generation module 50 is used to perform spatial correction on the optimal matching template according to the local geometric transformation parameters to generate an initial cutting path.

[0123] The path optimization module 60 is used to perform path optimization based on edge continuity constraints on the initial cutting path to generate the final cutting path;

[0124] The cutting execution module 70 is used to send the final cutting path to the CNC cutting equipment and control the laser head to perform cutting operations along the final cutting path.

[0125] It should be noted that the irregular glass cutting system based on edge feature matching provided in this embodiment of the invention is used to execute all the process steps of the irregular glass cutting method based on edge feature matching in the above embodiment. The working principle and beneficial effects of the two are one-to-one, so they will not be described again.

[0126] This invention also provides an electronic device. The electronic device includes a processor, a memory, and a computer program stored in the memory and executable on the processor, such as an image processing and path planning program. When the processor executes the computer program, it implements the steps described in the various embodiments of the edge feature matching-based irregular glass cutting method, for example... Figure 1 The step S11 shown. Alternatively, when the processor executes the computer program, it implements the functions of each module / unit in the above-described device embodiments, such as the path optimization module.

[0127] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the electronic device.

[0128] The electronic device may be an industrial control computer, an embedded industrial control computer, or a server. The electronic device may include, but is not limited to, a processor and a memory. Those skilled in the art will understand that the above-described components are merely examples of electronic devices and do not constitute a limitation on the electronic device. It may include more or fewer components than described above, or combine certain components, or different components. For example, the electronic device may also include an image acquisition interface, a motion control card, a network communication module, etc.

[0129] The processor referred to can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. General-purpose processors can be multi-core processors, supporting parallel processing of image pyramid construction and feature matching tasks.

[0130] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the electronic device by running or executing the computer programs and / or modules stored in the memory, and by calling data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, image processing library, motion control driver, etc.; the data storage area may store the original image, edge pixel set, feature descriptor, reference template library, cutting path, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as solid-state drive (SSD), eMMC storage chip, SD card, etc.

[0131] Wherein, if the modules / units integrated in the electronic device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), etc.

[0132] It should be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can be specifically implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0133] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

[0134] To enable those skilled in the art to fully understand and implement this invention, the specific implementation principles of this invention are further supplemented below with a specific application scenario.

Claims

1. A method for cutting irregularly shaped glass based on edge feature matching, characterized in that, include: Multi-view image acquisition is performed on the glass to be cut to generate a raw image dataset containing complete edge contour information; Image enhancement and denoising processing are performed on the original image dataset to extract a set of clear edge pixels; An edge feature vector is constructed based on the set of edge pixels, and an edge key point descriptor is generated by a scale-invariant feature transformation algorithm. The edge key point descriptor is matched with the template descriptor in the preset reference feature library to determine the optimal matching template and the corresponding local geometric transformation parameters; The optimal matching template is spatially corrected based on the local geometric transformation parameters to generate an initial cutting path; The initial cutting path is optimized based on edge continuity constraints to generate the final cutting path; The final cutting path is sent to the CNC cutting equipment, which controls the laser head to perform the cutting operation along the final cutting path.

2. The method for cutting irregularly shaped glass based on edge feature matching according to claim 1, characterized in that, Multi-view image acquisition is performed on the glass to be cut to generate a raw image dataset containing complete edge contour information, including: Place the glass to be cut on an image acquisition platform with uniform backlighting; The industrial camera is controlled to take pictures at equal angle intervals around the glass to be cut along a preset circular track to obtain two-dimensional grayscale images from multiple perspectives. Perspective correction is performed on images from each viewpoint, and then an image stitching algorithm is used to fuse them into a panoramic edge image; The panoramic edge image is subjected to resolution normalization processing to generate the original image dataset.

3. The method for cutting irregularly shaped glass based on edge feature matching according to claim 1, characterized in that, Image enhancement and denoising processing are performed on the original image dataset to extract a set of sharp edge pixels, including: An adaptive histogram equalization algorithm is applied to the original image dataset to improve local contrast. A bilateral filter is used to perform edge-preserving smoothing on the enhanced image; Preliminary edge pixels are extracted using the Canny edge detection operator; A morphological closing operation is performed on the initial edge pixels to connect the broken edges, and isolated noise regions with an area smaller than a preset threshold are removed to generate the edge pixel set.

4. The method for cutting irregularly shaped glass based on edge feature matching according to claim 1, characterized in that, An edge feature vector is constructed based on the aforementioned set of edge pixels, and an edge key point descriptor is generated using a scale-invariant feature transformation algorithm, including: A Gaussian difference pyramid is constructed on the edge pixel set to detect extreme points in the scale space; Remove low-contrast and unstable extreme points located in the edge response direction, and retain stable key points; Assign a principal direction to each stable keypoint and calculate a gradient direction histogram in its neighborhood to generate a 128-dimensional edge keypoint descriptor. All edge key point descriptors are sorted according to their spatial location to form the edge feature vector.

5. The method for cutting irregularly shaped glass based on edge feature matching according to claim 1, characterized in that, The edge key point descriptors are matched with template descriptors in a preset reference feature library to determine the optimal matching template and the corresponding local geometric transformation parameters, including: Load a subset of template descriptions from multiple historical successful cutting cases from the preset reference feature library; The nearest neighbor distance ratio criterion is used to perform coarse matching between the edge key point descriptors and each template descriptor; The random sampling consensus algorithm is applied to the coarse matching results to remove mismatched pairs; The affine transformation matrix is ​​calculated based on the remaining correctly matched point pairs and used as the local geometric transformation parameters. The template with the smallest reprojection error and the largest number of matching point pairs is selected as the optimal matching template.

6. The method for cutting irregularly shaped glass based on edge feature matching according to claim 1, characterized in that, Based on the local geometric transformation parameters, the optimal matching template is spatially corrected to generate an initial cutting path, including: Read the standard cutting trajectory coordinate sequence stored in the optimal matching template; Substitute each coordinate point in the standard cutting trajectory coordinate sequence into the affine transformation matrix to perform coordinate transformation; The transformed coordinate sequence is interpolated and encrypted to ensure that the distance between adjacent points is no greater than a preset step size threshold. The encrypted coordinate sequence is used as the initial cutting path.

7. The method for cutting irregularly shaped glass based on edge feature matching according to claim 1, characterized in that, Perform path optimization based on edge continuity constraints on the initial cutting path to generate the final cutting path, including: Project the initial cutting path onto the image plane containing the edge pixel set; Calculate the Euclidean distance from each point on the path to the nearest edge pixel, and mark abnormal point segments whose deviation distance is greater than the preset tolerance threshold; The abnormal point segments are locally reconstructed using the B-spline curve fitting method to make them conform to the actual edge direction; A speed look-ahead control algorithm is executed on the reconstructed path to generate a smooth motion command sequence containing acceleration constraints, which serves as the final cutting path.

8. A non-circular glass cutting system based on edge feature matching, comprising: The image acquisition module is used to acquire multi-view images of the glass to be cut and generate a raw image dataset containing complete edge contour information. The image preprocessing module is used to perform image enhancement and denoising processing on the original image dataset and extract a set of clear edge pixels; The feature extraction module is used to construct an edge feature vector based on the edge pixel set and generate an edge key point descriptor through a scale-invariant feature transformation algorithm. The template matching module is used to match the edge key point descriptor with the template descriptor in the preset reference feature library to determine the optimal matching template and the corresponding local geometric transformation parameters. The path generation module is used to perform spatial correction on the optimal matching template according to the local geometric transformation parameters and generate an initial cutting path. The path optimization module is used to perform path optimization based on edge continuity constraints on the initial cutting path to generate the final cutting path; The cutting execution module is used to send the final cutting path to the CNC cutting equipment and control the laser head to perform cutting operations along the final cutting path.

9. An electronic device comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor, when executing the computer program, implements the method as claimed in any one of claims 1-7.

10. A computer-readable medium, characterized in that, It stores a computer program thereon, wherein the program, when executed by a processor, implements the method as described in any one of claims 1-7.

Images