A cutting parameter automatic generation method and system for single-point incremental forming

Abstract

The application provides a cutting parameter automatic generation method and system for single-point incremental forming, and belongs to the technical field of mechanical parameter automatic generation. The method comprises the following steps: collecting image data information of a workpiece surface; constructing a data analysis model, analyzing the image data information of the workpiece surface, and realizing three-dimensional surface reconstruction of two-dimensional image data; comparing the differences of the three-dimensional reconstruction results, and generating corresponding machining data parameters; generating control instructions according to the data parameters; and executing the control instructions to realize the machining of the target object. Through the analysis of three-dimensional data, the data of the target form and the initial form of the workpiece are obtained, and the corresponding parameter machining data is automatically generated, so that the cutting parameter automatic generation is realized, the error phenomenon in the manual calculation process is reduced, and the work efficiency is improved.

Description

Technical Field

[0001] This invention relates to the technical field of automatic generation of mechanized parameters, and in particular to a method and system for automatic generation of cutting parameters for single-point progressive forming. Background Technology

[0002] With the advancement of automation technology, redundant manual labor is gradually being replaced by mechanized operation methods. In the processing of metal workpieces, single-point progressive forming is a forming tool on a CNC machine tool that controls the shape of the workpiece through a coded computer program. It utilizes parameters such as rotational speed, feed speed, and axial feed amount to form the three-dimensional envelope surface of plate-like parts layer by layer in the horizontal and vertical directions according to a set trajectory, thereby realizing the plastic forming method of metal workpieces.

[0003] In existing technologies, the operating parameters for metal workpieces that need to be processed are often calculated and input manually. Compared with the traditional manual processing of metal workpieces, this method only improves the time efficiency of using machinery to replace manual cutting operations. When faced with the required processing parameters, a large amount of labor costs are still required for calculation. Moreover, manual calculation methods are often prone to errors, which leads to increased waste of workpieces. Summary of the Invention

[0004] Purpose of the invention: To propose an automatic generation method and system for cutting parameters oriented towards single-point progressive forming, in order to solve the above-mentioned problems existing in the prior art. By analyzing three-dimensional data, the target shape and initial shape of the workpiece are obtained, and the corresponding parameter processing data are automatically generated, thereby realizing the automatic generation of cutting parameters, reducing the error phenomenon in the manual calculation process, and improving the work efficiency.

[0005] Technical solution: Firstly, an automatic generation method for cutting parameters oriented towards single-point progressive forming is proposed, which specifically includes the following steps:

[0006] Step 1: Acquire two-dimensional image data information from different angles using an information acquisition device; the two-dimensional image data information includes: image data information of the workpiece to be processed and image data information of the target object;

[0007] Step 2: Preprocess and analyze the acquired two-dimensional image data;

[0008] Step 3: Obtain the three-dimensional point coordinates corresponding to the two-dimensional image data information using the built-in parameters of the calibrated information acquisition device;

[0009] Step 4: Extract features from the preprocessed and analyzed image data;

[0010] Step 5: Perform stereo matching after feature point extraction;

[0011] Step 6: Combine the stereo matching results and the calibrated intrinsic and extrinsic parameters to achieve three-dimensional reconstruction of the two-dimensional image data;

[0012] Step 7: Compare the 3D reconstructed shapes of the workpiece to be processed and the target object to obtain the differences between them;

[0013] Step 8: Generate the operation parameters required to transform the workpiece to the target object based on the differences; the operation parameters include: cutting speed, feed rate, and depth of cut;

[0014] Step 9: Transmit the operation parameters to the cutting platform and generate the corresponding control commands;

[0015] Step 10: Complete the cutting operation according to the generated control instructions.

[0016] In some possible implementations of the first aspect, when preprocessing and analyzing the acquired two-dimensional image data, a three-dimensional sparse point cloud is obtained by extracting and matching features of the two-dimensional image data and estimating the pose of the information acquisition device based on the motion reconstruction structure; subsequently, a final three-dimensional dense point cloud is generated based on the acquired three-dimensional sparse point cloud.

[0017] The process of obtaining the pose of the information acquisition device specifically includes the following steps:

[0018] Step 3.1: Receive the pre-processed and analyzed two-dimensional image data;

[0019] Step 3.2: Extract image features from the two-dimensional image data; specifically, this includes the following steps:

[0020] Step 3.2.1: Identify scale- and rotation-invariant points of interest using the Gaussian differential function;

[0021] Step 3.2.2: At each candidate location of interest point, determine the location and scale using a fine-fit model;

[0022] Step 3.2.3: Assign at least one direction to each keypoint location based on the local gradient direction of the image;

[0023] Step 3.2.4: Measure the local gradient of the image in the neighborhood around each key point and at the selected scale, and use the transformed gradient representation to obtain the feature vector describing the key point, thereby realizing the extraction of image features.

[0024] In this process, when matching the extracted image features, the similarity between features is calculated to perform image matching; in the second image Select the first matching image The expression for the feature point is:

[0025]

[0026] In the formula, Represents feature points in the first image; Indicates the feature points in the second image;

[0027] After obtaining the feature points The feature points with the highest similarity Then, by traversing through the second image, the secondary similarity points are found. When the nearest distance Divide by the second closest distance When the feature point is less than the preset threshold, With feature points Match successful; among them, Representing feature points to feature point The distance between them; Representing feature points similarity The distance.

[0028] Step 3.3: Match the extracted image features to obtain the matching degree between each image and other images;

[0029] Step 3.4: Sort the matching results in descending order and select the image with the highest ranking;

[0030] Step 3.5: Obtain the pose of the information acquisition device and the spatial position of the points using epipolar geometry.

[0031] Step 3.6: Perform sparse reconstruction based on the obtained spatial point locations.

[0032] In some feasible methods of the first aspect, the process of three-dimensional reconstruction of two-dimensional image data by combining stereo matching results and calibrated intrinsic and extrinsic parameters specifically includes the following steps:

[0033] Step 6.1: Construct a depth map feature extraction network and receive image data from different viewpoints and the parameters of the corresponding information acquisition devices;

[0034] Step 6.2: Extract features from image data from different viewpoints using a depth map feature extraction network;

[0035] Step 6.3: Using the parameters of the information acquisition equipment, map the features from different perspectives to the reference perspective through homography transformation;

[0036] Step 6.4: Construct the cost volume using the mapped features;

[0037] Step 6.5: Obtain the depth map of the reference image using regularized cost volume;

[0038] Step 6.6: Based on the depth map, generate a three-dimensional dense point cloud;

[0039] Step 6.7: Complete the 3D reconstruction of the 2D image data based on the majority of 3D dense point clouds.

[0040] In some implementations of the first aspect, the cutting speed represents the instantaneous speed of the cutting tool relative to the main motion of the workpiece; the feed rate represents the distance the cutting tool moves along the feed direction per revolution of the workpiece; and the depth of cut represents the perpendicular distance between the machined surface and the surface to be machined. The expression for the cutting speed is:

[0041]

[0042] Based on the cutting speed, feed rate, and depth of cut, the expression for the feed rate is obtained as follows:

[0043]

[0044] In the formula, Indicates the workpiece rotation speed; Indicates cutting speed; Indicates the diameter of the surface to be machined on the workpiece; Indicates the feed per tooth; Indicates the number of teeth; This indicates the spindle speed.

[0045] Secondly, an automatic cutting parameter generation system for single-point progressive forming is proposed to realize the automatic generation method of cutting parameters. The system specifically includes:

[0046] The data acquisition module includes information acquisition equipment, used to acquire the required two-dimensional image data during actual operations;

[0047] The data processing module is used to receive two-dimensional image data acquired by the data acquisition module and to preprocess the two-dimensional image data.

[0048] The 3D reconstruction module includes a sparse reconstruction module and a dense reconstruction module, which are used to convert 2D image data into 3D image data.

[0049] The difference comparison module is used to receive the 3D image data reconstructed by the 3D reconstruction module and perform shape difference comparison.

[0050] The parameter generation module generates difference parameters based on the comparison results of the difference comparison module.

[0051] The data transmission module outputs the difference parameters generated by the parameter generation module through the constructed data transmission link;

[0052] The instruction generation module receives the differential parameters transmitted by the data transmission module and generates the corresponding control instructions.

[0053] The job execution module triggers the cutting process based on control commands and completes the cutting operation.

[0054] Thirdly, a computer-readable storage medium is proposed, on which computer program instructions are stored. When the computer program instructions are executed by a processor, a method for automatically generating cutting parameters is implemented.

[0055] Beneficial effects: This invention proposes an automatic generation method and system for cutting parameters oriented towards single-point progressive forming. By analyzing three-dimensional data, it obtains data on the target shape and initial shape of the workpiece and automatically generates corresponding parameter processing data, realizing automatic generation of cutting parameters, reducing the error phenomenon in the manual calculation process, and improving work efficiency. Attached Figure Description

[0056] Figure 1 This is a data processing flowchart of the present invention.

[0057] Figure 2 This is a schematic diagram of the improved module of the present invention.

[0058] Figure 3 This is a schematic diagram of the sparse reconstruction process of the present invention. Detailed Implementation

[0059] In the following description, numerous specific details are set forth in order to provide a more thorough understanding of the invention. However, it will be apparent to those skilled in the art that the invention can be practiced without one or more of these details. In other instances, certain technical features well-known in the art have not been described in order to avoid obscuring the invention.

[0060] In one embodiment, during personalized processing, users often have a specific desired object shape, but the parameters required for the corresponding workpiece processing cannot be obtained during the actual operation. Therefore, manual parameter input is the only option, which not only increases the time cost of the calculation process but also fails to achieve the expected accuracy. This embodiment addresses the parameter problem in the cutting process by proposing an automatic cutting parameter generation method for single-point progressive forming. Through analysis of the target object shape and the initial workpiece shape, the cutting parameters required for processing the initial workpiece are obtained, such as... Figure 1 As shown, this method specifically includes the following steps:

[0061] Step 1: Acquire two-dimensional image data from different angles using an information acquisition device; the acquired image data includes: image data of the workpiece to be processed and image data of the target object;

[0062] Step 2: Preprocess and analyze the acquired two-dimensional image data;

[0063] Step 3: Obtain the three-dimensional point coordinates corresponding to the two-dimensional image data information using the built-in parameters of the calibrated information acquisition device;

[0064] Specifically, when preprocessing and analyzing the acquired two-dimensional image data, feature extraction and matching of the two-dimensional image data are performed, and the pose of the information acquisition device is estimated based on the motion reconstruction structure to obtain a three-dimensional sparse point cloud; subsequently, the final three-dimensional dense point cloud is generated based on the acquired three-dimensional sparse point cloud.

[0065] The process of obtaining the pose of the information acquisition device specifically includes the following steps:

[0066] Step 3.1: Receive the pre-processed and analyzed two-dimensional image data;

[0067] Step 3.2: Extracting image features from two-dimensional image data; the process of extracting image features from two-dimensional image data specifically includes the following steps:

[0068] Step 3.2.1: Identify scale- and rotation-invariant points of interest using the Gaussian differential function;

[0069] Step 3.2.2: At each candidate location of interest point, determine the location and scale using a fine-fit model;

[0070] Step 3.2.3: Assign at least one direction to each keypoint location based on the local gradient direction of the image;

[0071] Step 3.2.4: Measure the local gradient of the image in the neighborhood around each key point and at the selected scale, and use the transformed gradient representation to obtain the feature vector describing the key point, thereby realizing the extraction of image features.

[0072] In a preferred embodiment, during the feature extraction process, since there are differences in brightness, rotation, and scale between the received multi-view images, a scale-invariant feature transformation method is used for feature extraction, thereby reducing the influence of scale factors by utilizing scale space detection.

[0073] Step 3.3: Match the extracted image features to obtain the matching degree between each image and other images; when matching the extracted image features, calculate the similarity between features and perform image matching; in the second image Select the first matching image The expression for the feature point is:

[0074]

[0075] In the formula, Represents feature points in the first image; Indicates the feature points in the second image;

[0076] After obtaining the feature points The feature points with the highest similarity Then, by traversing through the second image, the secondary similarity points are found. When the nearest distance Divide by the second closest distance When the feature point is less than the preset threshold, With feature points Match successful; among them, Representing feature points to feature point The distance between them; Representing feature points similarity The distance.

[0077] Step 3.4: Sort the matching results in descending order and select the image with the highest ranking;

[0078] Step 3.5: Obtain the pose of the information acquisition device and the spatial position of the points using epipolar geometry.

[0079] Step 3.6: Perform sparse reconstruction based on the obtained spatial point locations.

[0080] In a preferred embodiment, after feature extraction and feature matching are completed, the matching degree between each two-dimensional image data and other image data is obtained. Then, in order to improve the effect of sparse reconstruction, a pair of seed images are first selected to initialize the motion recovery structure. The essential matrix is ​​estimated according to the matching relationship between the seed images. Then, the matrix is ​​decomposed to estimate the position corresponding to the other seed image. Finally, the points in the space are restored by triangulation.

[0081] Since this embodiment primarily relies on image matching, the more matching feature points between image pairs, the easier it is for subsequent information acquisition devices to estimate their pose. Therefore, this embodiment employs a similar image selection method for the image data to be analyzed, eliminating images with few matching points to reduce feature matching time. The number of matching points between image pairs is determined using the final similarity score.

[0082] To improve the matching accuracy between similarity scores, this example uses the average similarity score as the final similarity criterion. The expression for the final similarity score is:

[0083]

[0084] In the formula, This indicates the similarity obtained using a grayscale histogram; This indicates the similarity obtained using a three-histogram method; This indicates the similarity using differential hashing; This indicates the similarity using mean hashing; This indicates the similarity using perceptual hashing.

[0085] Step 4: Extract features from the preprocessed and analyzed image data;

[0086] Step 5: Perform stereo matching after feature point extraction;

[0087] Step 6: Combine the stereo matching results and the calibrated intrinsic and extrinsic parameters to achieve 3D reconstruction of the 2D image data. The specific steps involved in achieving 3D reconstruction of the 2D image data by combining the stereo matching results and the calibrated intrinsic and extrinsic parameters are as follows:

[0088] Step 6.1: Construct a depth map feature extraction network and receive image data from different viewpoints and the parameters of the corresponding information acquisition devices;

[0089] Step 6.2: Extract features from image data from different viewpoints using a depth map feature extraction network;

[0090] Step 6.3: Using the parameters of the information acquisition equipment, map the features from different perspectives to the reference perspective through homography transformation;

[0091] Step 6.4: Construct the cost volume using the mapped features;

[0092] Step 6.5: Obtain the depth map of the reference image using regularized cost volume;

[0093] Step 6.6: Based on the depth map, generate a three-dimensional dense point cloud;

[0094] Step 6.7: Complete the 3D reconstruction of the 2D image data based on the majority of 3D dense point clouds.

[0095] In a preferred embodiment, by combining stereo matching results and calibrated intrinsic and extrinsic parameters, the three-dimensional reconstruction of two-dimensional image data is achieved by first using a constructed depth map feature extraction network to extract multi-scale features of the rabbit image; then, feature maps from different perspectives are aggregated to achieve a predicted depth map from coarse to fine.

[0096] Specifically, the process begins by inputting a reference image and several other images associated with the reference image, along with the corresponding information acquisition device parameters. After inputting all images into the feature extraction network, each image will generate feature maps at three resolutions. Following the generation of these multi-scale feature maps, a coarse-to-fine approach is used in the aggregation module to generate high-resolution depth maps.

[0097] The depth map feature extraction network includes convolutional layers, pooling layers, fully connected layers, and activation functions. In a preferred embodiment, a multi-scale feature extraction network (FPN) is used to extract features from the image. It obtains feature maps of different scales by inputting an original image of one size. This network structure includes a transposed convolutional module, and a batch normalization layer and the ReLU activation function are added after the convolution with stride S. To improve the image feature processing performance, ConvNeXt is further introduced into the FPN structure. Figure 2 The diagram shown is a schematic of the ConvNeXt structure, where g represents the number of received feature map channels; Conv( ) represents a grouped convolution with a kernel of 7, input and output channels of g, and groups of g; Conv( () indicates that the convolution kernel is 1, the input channel is g, and the output channel is... Convolution; Conv( ) represents a convolution with a kernel of 1, input channels g, and output channels g; LayerNorm represents normalization.

[0098] Step 7: Compare the 3D reconstructed shapes of the workpiece to be processed and the target object to obtain the differences between them;

[0099] Step 8: Generate the operation parameters required for the workpiece to be processed to the target object based on the differences; the operation parameters include: cutting speed, feed rate and depth of cut; cutting speed represents the instantaneous speed of the cutting tool relative to the main motion of the workpiece; feed rate represents the distance the cutting tool moves along the feed direction for each revolution of the workpiece; depth of cut represents the vertical distance between the machined surface of the workpiece and the surface to be processed.

[0100] The expression for the cutting speed is:

[0101]

[0102] Based on the cutting speed, feed rate, and depth of cut, the expression for the feed rate is obtained as follows:

[0103]

[0104] In the formula, Indicates the workpiece rotation speed; Indicates cutting speed; Indicates the diameter of the surface to be machined on the workpiece; Indicates the feed per tooth; Indicates the number of teeth; This indicates the spindle speed.

[0105] Step 9: Transmit the operation parameters to the cutting platform and generate the corresponding control commands;

[0106] Step 10: Complete the cutting operation according to the generated control instructions.

[0107] In a further embodiment, when acquiring the target sparse point cloud through a series of image sequences from different perspectives and obtaining the pose of the information acquisition device, an incremental approach is used to perform global optimization after the calculation is completed.

[0108] Specifically, such as Figure 3 As shown, firstly, features of all input 2D image data are extracted, and the extracted image features are matched to obtain the matching degree between each image and other images. Secondly, a set of image data with the highest matching degree is selected, and the pose and spatial position of the information acquisition device are obtained using a geometric approach. Thirdly, a new image data is added to recalculate the pose and spatial position of the information acquisition device, thereby updating the parameters. A threshold is then set, and within the threshold range, the pose and spatial position of the information acquisition device are continuously updated iteratively. Finally, at the end of the iteration, bundle adjustment is used again for global optimization. In a preferred embodiment, after obtaining the parameters of the information acquisition device in each iteration, bundle adjustment is used for local optimization.

[0109] In the process of optimizing the reconstructed results using the bundle adjustment method, a set of spatial points is first set. Imaging from information acquisition devices at different viewpoints, let the projection matrix of the i-th camera be... ,point The image under the i-th information acquisition device is The purpose of reconstruction is to estimate the parameters of the information acquisition equipment. and spatial points ,in, In practical applications, noise interference often occurs, leading to errors or mistakes in the calculation process. Therefore, the bundle adjustment method proposed in this embodiment can minimize the distance between the reprojection point and the imaging point, optimize the reconstructed result, and provide a true maximum likelihood estimate.

[0110] In a further embodiment, existing technologies often use a fixed resolution to construct a single cost volume, which results in significant memory consumption. Furthermore, the input image is typically downsampled, leading to a low-resolution depth map as the final processing result. Subsequent generation of the 3D point cloud relies on fusing these low-resolution depth maps, impacting the quality of the 3D point cloud. Therefore, this embodiment employs a cascaded approach to construct the cost volume. First, the cost volume is constructed on a low-resolution feature map, and then the corresponding depth map is estimated. After obtaining the low-resolution depth map, it is upsampled using bilinear interpolation. This reduces the number of depth planes in the high-resolution feature map, significantly decreasing the cost volume in the depth plane dimension when constructing the cost volume from the high-resolution feature map, thus greatly reducing memory usage.

[0111] Specifically, the cascading method used in constructing the cost volume involves a three-stage cascading structure. The first stage inputs include: , , The corresponding output is The output image size is The second level of input includes: , , The corresponding output is The output image size is The third level of input includes: , , The corresponding output is The output image size is .in, Representing feature maps, Indicates the parameters of the information collection device, Indicates the number of depth planes, This represents the estimated depth map, where L represents the level. The process of obtaining the estimated depth map for level L specifically includes the following steps:

[0112] Step a: Combine feature maps The corresponding information acquisition device parameters and the number of depth planes are used to perform homography transformation on the feature maps of other perspectives, thereby realizing the mapping from other perspectives to the reference perspective;

[0113] Step b: Stack the mapped features together and perform adaptive view aggregation to obtain the cost volume;

[0114] Step c: Regularize the cost volume to obtain the probability distribution map;

[0115] Step d: Estimate the depth map from the reference viewpoint using the probability distribution map.

[0116] The cascaded adaptive module used in this embodiment allows the low-resolution cost volume to have greater depth, as well as smaller width and height.

[0117] In one embodiment, an automatic cutting parameter generation system for single-point incremental forming is proposed to implement an automatic cutting parameter generation method for single-point incremental forming. The system specifically includes the following modules:

[0118] The data acquisition module includes information acquisition equipment, used to acquire the required two-dimensional image data during actual operations;

[0119] The data processing module is used to receive two-dimensional image data acquired by the data acquisition module and to preprocess the two-dimensional image data.

[0120] The 3D reconstruction module includes a sparse reconstruction module and a dense reconstruction module, which are used to convert 2D image data into 3D image data.

[0121] The difference comparison module is used to receive the 3D image data reconstructed by the 3D reconstruction module and perform shape difference comparison.

[0122] The parameter generation module generates difference parameters based on the comparison results of the difference comparison module.

[0123] The data transmission module outputs the difference parameters generated by the parameter generation module through the constructed data transmission link;

[0124] The instruction generation module receives the differential parameters transmitted by the data transmission module and generates the corresponding control instructions.

[0125] The job execution module triggers the cutting process based on control commands and completes the cutting operation.

[0126] In one embodiment, a computer-readable storage medium is provided on which computer program instructions are stored.

[0127] Among them, when the computer program instructions are executed by the processor, a method for automatically generating cutting parameters is implemented.

[0128] As described above, although the invention has been shown and described with reference to specific preferred embodiments, it should not be construed as limiting the invention itself. Various changes in form and detail may be made without departing from the spirit and scope of the invention as defined in the appended claims.

Claims

1. A method for automatic generation of cutting parameters for single point incremental forming, characterized in that, Specifically comprising the following steps: Step 1, obtaining two-dimensional image data information of different angles through an information acquisition device; The two-dimensional image data information includes: workpiece image data information to be processed and target object image data information; Step 2, pre-processing and analyzing the obtained two-dimensional image data; Step 3, obtaining three-dimensional point coordinates corresponding to the two-dimensional image data information through the calibrated built-in parameters of the information acquisition device; Step 4, extracting features from the image data information after pre-processing and analysis; Step 5, performing stereo matching after completing feature point extraction; Step 6, combining the stereo matching result and the calibrated internal and external parameters to realize three-dimensional reconstruction of the two-dimensional image data; Step 7, comparing the morphology of the workpiece to be processed and the target object after three-dimensional reconstruction to obtain the difference between the two; Step 8, generating operation parameters required for the workpiece to be processed to the target object according to the difference; the operation parameters include: cutting speed, feed rate and cutting depth; the cutting speed represents the instantaneous speed of the cutting tool selected relative to the main motion of the workpiece; the feed rate represents the distance that the tool moves along the feed direction per revolution of the workpiece to be machined; the cutting depth represents the vertical distance between the machined surface and the to-be-machined surface of the workpiece to be machined; Step 9, transmitting the operation parameters to the cutting platform and generating corresponding control instructions; Step 10, completing the cutting operation according to the generated control instructions.

2. The method of claim 1, wherein, When pre-processing and analyzing the obtained two-dimensional image data, the pose of the information acquisition device is estimated based on the feature extraction and matching of the two-dimensional image data and the motion recovery structure, so as to obtain a three-dimensional sparse point cloud; Subsequently, a final three-dimensional dense point cloud is generated based on the obtained three-dimensional sparse point cloud.

3. The method of claim 2, wherein, The process of obtaining the pose of the information acquisition device specifically comprises the following steps: Step 3.1, receiving the two-dimensional image data after pre-processing and analysis; Step 3.2, extracting image features of the two-dimensional image data; Step 3.3, matching the extracted image features to obtain the matching degree between each image and other images; Step 3.4, arranging the matching degree results in descending order and selecting the highest ranked image; Step 3.5, obtaining the pose of the information acquisition device and the spatial point position by the method of epipolar geometry; Step 3.6, realizing sparse reconstruction according to the obtained spatial point position.

4. The method of claim 3, wherein, In the process of extracting image features of the two-dimensional image data, specifically comprising the following steps: Step 3.2.1, identifying interest points that are invariant to scale and rotation through a Gaussian differential function; Step 3.2.2, determining the position and scale through a fitting fine model at the candidate position of each interest point; Step 3.2.3, assigning at least one direction to each key point position based on the gradient direction of the image local; Step 3.2.4, measuring the gradient of the image local within the neighborhood around each key point and at the selected scale, and obtaining a feature vector describing the key point by using the transformed gradient representation, to realize the extraction of image features.

5. The method of claim 3, wherein the cutting parameter is automatically generated based on the single-point incremental forming. When the extracted image features are matched, the similarity between the features is calculated, and the matching between the images is performed; in the second image the expression of the feature point matched with the first image in the first image is: In the formula, denotes a feature point in the first image; denotes a feature point in the second image; After the feature point with the highest similarity is found , the second similar point in the second image is found by traversal . When the nearest distance is divided by the second nearest distance is less than a preset threshold, the feature point is successfully matched with the feature point ; wherein, denotes the distance between the feature point and the feature point . denotes the distance between the feature point and the second similar point .

6. The method of claim 1, wherein, In the three-dimensional reconstruction process of the two-dimensional image data in combination with the stereomatching result and the calibrated internal and external parameters, the following steps are specifically included: Step 6.1, constructing a depth map feature extraction network and receiving image data of different perspectives and parameters of the information acquisition device corresponding thereto; Step 6.2, extracting features of the image data of different perspectives by using the depth map feature extraction network; Step 6.3, mapping the features of different perspectives to a reference perspective by using the information acquisition device parameters through homography transformation; Step 6.4, constructing a cost volume by using the mapped features; Step 6.5, obtaining a depth map of the reference image by using the regularized cost volume; Step 6.6, generating a three-dimensional dense point cloud based on the depth map; Step 6.7, completing the three-dimensional reconstruction of the two-dimensional image data according to the three-dimensional dense point cloud.

7. The method of claim 1, wherein, The expression of the cutting speed is: According to the cutting speed, the feed rate and the cutting depth, the expression of the feed speed is obtained as: wherein represents the workpiece rotational speed; represents the cutting speed; represents the workpiece machined surface diameter; represents the feed per tooth; represents the number of teeth; represents the spindle rotational speed.

8. A cutting parameter automatic generation system for single point incremental forming, for implementing the cutting parameter automatic generation system according to any one of claims 1 to 7, characterized in that, Specifically, the following modules are included: A data acquisition module including an information acquisition device, configured to acquire the required two-dimensional image data in an actual operation process; A data processing module configured to receive the two-dimensional image data collected by the data acquisition module and pre-process the two-dimensional image data; A three-dimensional reconstruction module including a sparse reconstruction module and a dense reconstruction module, configured to realize the conversion of the two-dimensional image data to three-dimensional image data; A difference comparison module configured to receive the three-dimensional image data reconstructed by the three-dimensional reconstruction module and perform shape difference comparison; A parameter generation module configured to generate difference parameters according to the comparison result of the difference comparison module; A data transmission module configured to output the difference parameters generated by the parameter generation module through a constructed data transmission link; An instruction generation module configured to receive the difference parameters transmitted by the data transmission module and generate corresponding control instructions; An operation execution module configured to trigger a cutting operation process according to the control instructions and complete the cutting operation.

9. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer program instructions, and the computer program instructions are executed by the processor to realize the cutting parameter automatic generation method according to any one of claims 1-7. The computer readable storage medium stores computer program instructions, and the computer program instructions are executed by the processor to realize the cutting parameter automatic generation method according to any one of claims 1-7.

Images