A machine vision-based tool grinding method

Abstract

The application provides a tool grinding method based on machine vision, which comprises the following steps: S1, obtaining image data of a to-be-processed accessory after the accessory is fixed; S2, extracting accessory features of the to-be-processed accessory according to the obtained image data; S3, comparing and analyzing the extracted accessory features with a set processing target, and generating processing parameters according to the comparison and analysis result, wherein the processing parameters comprise feed parameters and grinding parameters; and S4, processing the to-be-processed surface according to the processing parameters, and repeating the steps S1-S3 until the current processing surface reaches the processing target after the processing is completed. The application helps to improve the intelligent level and efficiency of accessory grinding processing.

Description

Technical Field

[0001] This invention relates to the field of machine vision technology, and in particular to a tool grinding method based on machine vision. Background Technology

[0002] In traditional grinding processes, grinding usually involves a long processing time. During repeated grinding operations, operators typically need to stay near the machine tool and adjust the machine tool's processing parameters in real time based on the processing status of the workpiece to complete the grinding process.

[0003] However, the traditional grinding process relies too much on manual adjustment of tool parameters and observation of the finished surface. This manual operation method has significant limitations: relying too much on the operator's experience and reaction results in excessively high labor costs and low production efficiency, which cannot meet the needs of modern intelligent manufacturing. Summary of the Invention

[0004] To address the aforementioned problems, this invention aims to provide a tool grinding method based on machine vision.

[0005] The objective of this invention is achieved through the following technical solution:

[0006] This invention discloses a tool grinding method based on machine vision, comprising the following steps:

[0007] S1 acquires image data of the part to be processed after it has been fixed in place;

[0008] S2 extracts the part features of the part to be processed based on the acquired image data, where the part features include size information and image information of the surface to be processed;

[0009] S3 compares and analyzes the extracted part features with the set machining targets, and generates machining parameters based on the comparison and analysis results. The machining targets include part size standards and machined surface quality standards, and the machining parameters include feed parameters and grinding parameters; specifically including:

[0010] S31 calculates the dimensional characteristics of the workpiece based on its current dimensional information, compares and analyzes the obtained dimensional characteristics with the preset workpiece dimensional standard to obtain the dimensional deviation, and sets the corresponding feed parameters based on the dimensional deviation, including the feed amount.

[0011] When the dimensional deviation is greater than the preset first deviation value, adjust the current feed rate to the first feed rate;

[0012] When the dimensional deviation is less than or equal to the preset first deviation value and greater than or equal to the second deviation value, adjust the current feed rate to the second feed rate.

[0013] When the dimensional deviation is less than the preset second deviation value, adjust the current feed rate to the third feed rate;

[0014] The first feed rate is greater than the second feed rate; the second feed rate is greater than the third feed rate.

[0015] S32 When the dimensional deviation is less than the preset second deviation value, the quality of the surface to be processed is further obtained based on the current image information of the surface to be processed of the part to be processed. The obtained quality of the surface to be processed is compared and analyzed with the preset quality standard of the surface to be processed to obtain the quality deviation of the surface to be processed. The corresponding grinding parameters are set according to the quality deviation of the surface to be processed; the grinding parameters include the grinding speed.

[0016] S4 processes the surface to be processed according to the processing parameters, and after processing is completed, repeats the above steps S1-S3 until the current processed surface reaches the processing target.

[0017] Preferably, in step S1, the processing surface image and key side image data of the part to be processed are acquired based on the 3D camera.

[0018] Preferably, in step S2, edge detection processing is performed based on the acquired image data to obtain the edge feature information of the part to be processed;

[0019] Based on the comparison and analysis of the acquired edge feature information with the preset standard coordinate system, the dimensional information of the processing surface and key side of the part to be processed is obtained as the dimensional information of the part to be processed. The dimensional information includes the coordinate system position, length data, horizontal tilt angle, vertical tilt angle, etc. of each edge.

[0020] Based on the obtained edge feature information, the area to be processed of the part to be processed is further extracted to obtain the image of the surface to be processed.

[0021] Preferably, step S2 further includes:

[0022] The acquired image data of the parts to be processed is preprocessed to obtain the preprocessed image data. Then, edge detection processing is performed on the preprocessed image data to obtain the edge feature information of the parts to be processed.

[0023] Preferably, step S31 specifically includes:

[0024] Calculate the length characteristics of the surface to be machined based on the edge lengths of the surface to be machined and the critical side of the part to be machined; or, calculate the thickness characteristics of the surface to be machined based on the dimensional information of the critical side.

[0025] The length or thickness characteristics of the surface to be processed are compared with the preset part size standard to obtain the length or thickness dimensional deviation. Furthermore, the dimensional deviation is compared with the preset first deviation value and second deviation value.

[0026] When the dimensional deviation is greater than the preset first deviation value, adjust the current feed rate to the first feed rate;

[0027] When the dimensional deviation is less than or equal to the preset first deviation value and greater than or equal to the second deviation value, adjust the current feed rate to the second feed rate.

[0028] When the dimensional deviation is less than the preset second deviation value, adjust the current feed rate to the third feed rate;

[0029] Based on the horizontal or numerical inclination of the edge of the surface to be machined or the critical side, calculate the horizontal or numerical angular deviation between the surface to be machined and the standard coordinate system, and adjust the feed angle or feed path accordingly based on the angular deviation.

[0030] Preferably, step S32 specifically includes:

[0031] Based on the acquired image of the surface to be processed, a machine vision-based grinding surface roughness detection model is used to analyze the roughness evaluation value Ra(t) of the surface to be processed. The roughness evaluation value Ra(t) is compared with the roughness standard RaT to obtain the roughness deviation ΔRa = Ra(t) - RaT. When the roughness deviation ΔRa is greater than the set threshold ΔRaT, the grinding speed is adjusted to decrease; when the roughness deviation ΔRa is less than or equal to the set threshold ΔRaT, the grinding speed is adjusted to the standard grinding speed.

[0032] Preferably, step S3 further includes:

[0033] S33 performs image recognition-based defect analysis on the surface to be processed based on the acquired image information of the surface to be processed, detects abnormalities on the surface to be processed, and obtains the defect analysis results of the surface to be processed.

[0034] When the defect analysis result is abnormal, the image information of the surface to be processed corresponding to the abnormal analysis result is extracted and transmitted to the management terminal, specifically including:

[0035] A machine vision-based grinding surface defect recognition model analyzes and processes images of the surface to be processed, and identifies abnormal defects on the surface to be processed, including scratches, pits, and deformations.

[0036] Preferably, step S4 specifically includes:

[0037] If the dimensional deviation between the size of the surface to be processed and the dimensional standard is less than the preset second deviation value, and the current roughness evaluation value is less than the preset roughness standard, it is determined that the processing step of the current surface is completed, and the processing of the current surface is stopped, or the processing of the current surface is skipped.

[0038] The beneficial effects of this invention are as follows: This invention proposes a machine vision-based tool grinding method. During the grinding process, before each feed stroke or the start of each grinding step, image data of the workpiece's surface to be processed is acquired. Based on the obtained image data, feature extraction based on machine vision is performed. By obtaining the deviation between the dimensional features of the workpiece's surface and the target size, the feed rate is progressively controlled, thereby achieving real-time intelligent switching and control between roughing and fine machining. This helps improve the overall efficiency of grinding while ensuring machining precision. Simultaneously, in the fine machining stage, the quality features of the workpiece's surface are further extracted based on the image information. Grinding parameters are further intelligently adjusted according to these quality features to ensure the quality of the machined surface meets the machining requirements. Based on the above method, the CNC machining equipment is controlled to complete the grinding of parts. On the one hand, this improves the level of intelligent dynamic control and saves labor costs; on the other hand, the progressive and surface quality-based dynamic control method effectively improves the efficiency and accuracy of grinding. Attached Figure Description

[0039] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0040] Figure 1 A flowchart illustrating a tool grinding method based on machine vision, as shown in an exemplary embodiment of the present invention;

[0041] Figure 2 This is a flowchart illustrating step S3 of an exemplary embodiment of the present invention.

[0042] Figure 3 This is a schematic diagram of the implementation device framework structure as shown in an exemplary embodiment of the present invention. Detailed Implementation

[0043] The present invention will be further described in conjunction with the following application scenarios.

[0044] See Figure 1 An example illustrates a machine vision-based tool grinding method, which includes the following steps:

[0045] S1 acquires image data of the part to be processed after it has been fixed in place;

[0046] S2 extracts the part features of the part to be processed based on the acquired image data, where the part features include size information and image information of the surface to be processed;

[0047] S3 compares and analyzes the extracted part features with the set processing targets, and generates processing parameters based on the comparison and analysis results. The processing targets include part size standards and processing surface quality standards, and the processing parameters include feed parameters and grinding parameters.

[0048] S4 processes the surface to be processed according to the processing parameters, and after processing is completed, repeats the above steps S1-S3 until the current processed surface reaches the processing target.

[0049] Preferred, see Figure 2 Step S3 specifically includes:

[0050] S31 calculates the dimensional characteristics of the workpiece based on its current dimensional information, compares and analyzes the obtained dimensional characteristics with the preset workpiece dimensional standard to obtain the dimensional deviation, and sets the corresponding feed parameters based on the dimensional deviation, including the feed amount.

[0051] When the dimensional deviation is greater than the preset first deviation value, adjust the current feed rate to the first feed rate;

[0052] When the dimensional deviation is less than or equal to the preset first deviation value and greater than or equal to the second deviation value, adjust the current feed rate to the second feed rate.

[0053] When the dimensional deviation is less than the preset second deviation value, adjust the current feed rate to the third feed rate;

[0054] The first feed rate is greater than the second feed rate; the second feed rate is greater than the third feed rate.

[0055] When the dimensional deviation is less than the preset second deviation value, S32 further obtains the surface quality of the surface to be processed based on the current surface image information of the part to be processed, compares and analyzes the obtained surface quality with the preset surface quality standard to obtain the surface quality deviation, and sets the corresponding grinding parameters based on the surface quality deviation; wherein the grinding parameters include grinding speed.

[0056] The present invention proposes a machine vision-based tool grinding method. During the grinding process, before each feed stroke or grinding step, image data of the workpiece surface to be processed is acquired. Based on the obtained image data, machine vision-based feature extraction is performed. By obtaining the deviation between the dimensional features of the workpiece surface and the target size, the feed rate is progressively controlled, thereby achieving real-time intelligent switching between roughing and fine machining. This helps improve the overall efficiency of grinding while ensuring machining precision. Simultaneously, in the fine machining stage, the quality features of the workpiece surface are further extracted based on the image information. Grinding parameters are further intelligently adjusted according to these quality features to ensure the quality of the machined surface meets the machining requirements. This method enables control of CNC machining equipment to complete the grinding of parts. On the one hand, it improves the level of intelligent dynamic control and saves labor costs; on the other hand, the progressive and quality-based dynamic control method effectively improves grinding efficiency and machining accuracy.

[0057] In one exemplary scenario, see [link to example]. Figure 3 The tool grinding method described in the above embodiments can be implemented based on a processor or server, and more specifically, it can be implemented based on a CNC machining center with an intelligent processor, intelligent grinding equipment, etc. For example, by executing the above method in the processor of a CNC machining center, intelligent setting of grinding parameters of the CNC machining center can be achieved, thereby improving the intelligence level of the CNC machining equipment.

[0058] The image data of the parts to be processed is captured in real time by a camera set on the workbench. After the camera collects image data of the parts to be processed from different angles, it is transmitted to the processor for further analysis and processing.

[0059] In one exemplary scenario, cameras are positioned facing the end face and the side face to be processed, respectively, to capture images of the surface to be processed and the thickness of the side face of the part.

[0060] In one exemplary scenario, the parts to be processed can be components of different shapes or structures, such as cylinders, spheres, or flat plates, which can be adapted to different component processing scenarios.

[0061] Preferably, the processing objectives include the dimensional objectives of the parts to be processed and the quality standards of the surfaces to be processed.

[0062] Preferably, in step S1, the processing surface image and key side image data of the part to be processed are acquired based on the 3D camera.

[0063] Depending on the setup of the CNC machine tool's worktable, the camera can be positioned appropriately to target the workpiece being processed, thereby enabling the acquisition and reception of relevant image data.

[0064] Preferably, step S2 includes: performing edge detection processing based on the acquired image data to obtain edge feature information of the part to be processed;

[0065] Based on the comparison and analysis of the acquired edge feature information with the preset standard coordinate system, the dimensional information of the processing surface and key side of the part to be processed is obtained as the dimensional information of the part to be processed. The dimensional information includes the coordinate system position, length data, horizontal tilt angle, vertical tilt angle, etc. of each edge.

[0066] Based on the obtained edge feature information, the area to be processed of the part to be processed is further extracted to obtain the image of the surface to be processed.

[0067] In one exemplary scenario, the actual length of the corresponding edge can be calculated by combining the depth information from a 3D camera with the pixel length data of the edge of the part to be processed in the image.

[0068] In one exemplary scenario, based on the camera's location and the location of the part to be processed, a "+" sign for a standard coordinate system is set in the actual equipment (workbench). This sign is used to mark the horizontal and vertical reference angles of the actual coordinate system, and a standard scale is set on it. Based on the marked scale, the actual length of each edge of the part to be processed in the image can be identified by machine vision.

[0069] For example, based on the pixel distance between the two outer edges of the surface to be processed, the actual length between the two outer edges is calculated using the marked scale information, thus obtaining the size information of the surface to be processed.

[0070] By extracting features from the acquired image data, the relevant dimensional information of the workpiece to be processed can be accurately extracted, which serves as the basis for setting subsequent feed parameters. At the same time, the image of the surface to be processed is further extracted, which serves as the basis for further image analysis and setting grinding parameters based on the image analysis results.

[0071] In actual operating environments, the acquired image data of the parts to be processed often exhibits reflections due to uncontrollable angles caused by the metal material or residual grinding fluid. This reflection (e.g., sparks from the processing site are reflected / mapped into the image) results in localized blurring, affecting the accuracy of subsequent edge detection, dimensional analysis, or surface roughness analysis based on the image data. Adjusting the lens angle or removing water droplets from the surface of the parts to eliminate these effects requires significant additional time and resources, severely impacting processing efficiency. Therefore, to address the abnormal reflections in the image data of the parts caused by the environmental conditions, this invention specifically preprocesses the acquired image data to eliminate these abnormal effects through image processing.

[0072] Preferably, step S2 further includes:

[0073] The acquired image data of the parts to be processed is preprocessed to obtain the preprocessed image data. Then, edge detection processing is performed on the preprocessed image data to obtain the edge feature information of the parts to be processed.

[0074] In step S2, the acquired image data of the parts to be processed is preprocessed, specifically including:

[0075] Extract the grayscale information of each pixel in the image of the part to be processed;

[0076] A K×K detection window Φ with side length K=3 or K=5 is used to sequentially traverse each pixel in the image, calculating the attention factor of each region. The traversal method involves aligning the center of the detection window Φ with each pixel in the image. The function used to calculate the attention factor of the detection window is:

[0077]

[0078] In the formula, ATT(x,y) represents the attention factor of the detection window Φ(x,y) centered at pixel (x,y), (a,b) and (c,d) are pixels within the detection window range, hg(a,b) and hg(c,d) represent the gray values ​​at pixel positions (a,b) and (c,d) respectively, SDH(Φ) represents the standard deviation of gray values ​​of each pixel within the current detection window, ω1 represents the preset standard deviation correction value, where ω1∈[0.1,1]; SIN(x,y) represents the position adjustment factor of the current detection window, where SIN(x,y)=1 when the center position (x,y) of the detection window is within the preset area range, otherwise SIN(x,y)=0.5, K represents the side length of the detection window, and Num unED (Φ) represents the number of feature pixels within the current detection window, where the feature pixel determination function is: When a pixel (x, y) satisfies the judgment function, the pixel is marked as a feature pixel, where HT represents the preset gray-level difference threshold, HT∈[20,35]; Num(Φ) represents the number of pixels in the detection window. This represents the floor function;

[0079] Based on the obtained attention factor, when the attention factor ATT(x,y) is greater than the set attention standard value ATG, ATT(x,y)>ATG, the area covered by the current detection window is marked as the attention region, and the pixels within the attention region are marked as attention pixels. Furthermore, the attention feature values ​​of each attention pixel are labeled. In the formula, (i,j) represents the pixels within the coverage area when the center position of the detection window is (x,y);

[0080] After the detection window sequentially traverses each region of the image of the part to be processed, and marks the feature values ​​of the regions and pixels of interest in the current image, further enhancement processing is performed on the image, including:

[0081] Extract the luminance component sub-map PL, color component sub-map Pa, and color component sub-map Pb of the image to be processed in the Lab color space;

[0082] Based on the obtained luminance component sub-map PL, the luminance component values ​​of each pixel in the Lab color space are obtained;

[0083] Histogram statistics are performed on the feature values ​​of each pixel in the region of interest to obtain a histogram of feature values. The histogram contains the ranking information of each pixel and the number of pixels corresponding to each feature value.

[0084] Pixels are sorted from largest to smallest based on their light beam feature values, with larger feature values ​​ranking higher. The top 0-25% of pixels are then labeled as a class Sn. I The remaining 25% to 100% of pixels are marked as Class II pixels, Sn. II ;

[0085] First, a first brightness enhancement process is performed on the pixels in the region of interest. The first brightness enhancement process function used is as follows:

[0086]

[0087] In the formula, Lgb(x,y) represents the luminance component value of pixel (x,y) after the first enhancement process, Lg(x,y) represents the current luminance component value of pixel (x,y); LgK1 represents the preset stretching standard value, where LgK1∈[10,25];

[0088] After the first brightness enhancement process, a second brightness enhancement process is performed on each pixel of interest, sequentially from low to high, based on the pixel's feature values. The function used for the second brightness enhancement process is as follows:

[0089]

[0090] In the formula, Lga(x,y) represents the luminance component value of pixel (x,y) after the second luminance enhancement process. During the first luminance enhancement process, the current luminance component value of each pixel of interest is updated after processing. Lg(x,y) represents the current luminance component value of pixel (x,y), and pixel (a,b) is a pixel belonging to the eight-neighborhood of pixel (x,y). AT (a,b) represents the classification feature value of pixel (a,b), where when pixel (a,b) is a pixel of interest, key AT (a,b) = 3, otherwise key AT (a,b) = 1; Lg(a,b) represents the current luminance component value of pixel (a,b); Num 8q,unAT Num represents the total number of non-interested pixels within an eight-neighborhood centered at pixel (x,y); 8q,AT This represents the total number of pixels of interest within an eight-neighborhood centered at pixel (x,y);

[0091] After completing the second brightness enhancement process, a further overall brightness enhancement process is performed on all pixels. The overall brightness enhancement function used is as follows:

[0092] Lgc(x,y)=ω2×(Lg(x,y)+LgP1)

[0093] In the formula, Lgc(x,y) represents the luminance component value of pixel (x,y) after overall luminance enhancement processing, Lg(x,y) represents the current luminance component value of pixel (x,y); LgP1 represents the preset standard luminance component value, where LgP1∈[60,65], and ω2 represents the preset adjustment factor, where ω2∈[0.45,0.55].

[0094] Based on the updated luminance component sub-image PL, it is reconstructed with the corresponding color component sub-images Pa and Pb to obtain the preprocessed image data to be processed.

[0095] The present invention proposes a technical solution for preprocessing images of parts to be processed based on image processing technology. This solution adaptively adjusts local reflections caused by uncontrollable conditions in the processing environment and on the surface of the parts, improving image clarity and quality. Specifically, during preprocessing, based on the characteristics of regions of interest in the image, a region of interest detection scheme based on the image's grayscale information is proposed. This scheme accurately captures regions with significant local feature changes within the part to be processed area, using a detection window as a basis. (Considering that the surface of parts to be processed typically has relatively small local feature changes, this method can accurately detect abnormal regions in the image, which is more suitable for detecting areas prone to reflection in parts images compared to traditional methods based on grayscale changes or edge information.) Furthermore, to distinguish between regular edge regions and abnormal regions, a feature point-based adjustment component is added, resulting in a more distinct region-discriminating attention factor. Finally, based on the attention factor, the region of interest in the image is extracted, which can accurately detect general edge features and abnormal reflective features. Further brightness enhancement processing is performed on the obtained region of interest. First, based on the attention feature values ​​of each pixel in the region of interest, abnormal reflections are selectively eliminated in abnormal areas. Based on the eliminated features, and considering the influence characteristics of reflective phenomena, a regionalized brightness enhancement method from the outside in is proposed to further restore the influence of reflective areas. This achieves adaptive elimination and restoration of local abnormal illumination in the image of the part to be processed, helping to restore the true image situation and improve image quality. Finally, based on the overall enhancement approach, global brightness enhancement processing is performed on the entire image to improve the overall image quality, thereby improving the accuracy of subsequent edge detection, size analysis, or surface roughness analysis.

[0096] Preferably, step S31 specifically includes:

[0097] Calculate the length characteristics of the surface to be machined based on the edge lengths of the surface to be machined and the critical side of the part to be machined; or, calculate the thickness characteristics of the surface to be machined based on the dimensional information of the critical side.

[0098] The length or thickness characteristics of the surface to be processed are compared with the preset part size standard to obtain the length or thickness dimensional deviation. Furthermore, the dimensional deviation is compared with the preset first deviation value and second deviation value.

[0099] When the dimensional deviation is greater than the preset first deviation value, adjust the current feed rate to the first feed rate;

[0100] When the dimensional deviation is less than or equal to the preset first deviation value and greater than or equal to the second deviation value, adjust the current feed rate to the second feed rate.

[0101] When the dimensional deviation is less than the preset second deviation value, adjust the current feed rate to the third feed rate.

[0102] In one exemplary scenario, based on the material, size, and other characteristic parameters of the part to be processed, the first feed rate is set to 0.02-0.1mm; the second feed rate is set to 0.01-0.05mm; and the third feed rate is set to 0-0.005mm.

[0103] In an exemplary scenario, based on the current thickness information h = 10.56 mm of the machined surface, when the thickness information is compared with the target thickness h' = 10.3 mm, the dimensional deviation DT = 1.56 - 1.3 = 0.26 mm exceeds the preset first deviation value D1 = 0.1 mm. Then, the feed rate for the current machining cycle is set to the first feed rate Y = 0.05 mm. After two machining cycles, the thickness of the machined surface is measured again as h = 10.37 mm, and the dimensional deviation DT = 1.37 - 1.3 = 0.07 mm does not exceed the preset first deviation value D1 = 0.1 mm. If the thickness is 0.1mm but exceeds the second deviation value D2 = 0.011mm, then the feed rate for the current machining cycle is set to the second feed rate Y = 0.01mm. After 6 machining cycles, if the thickness of the machined surface is measured again to be 10.305mm and the dimensional deviation DT = 10.305 - 10.3 = 0.005mm, which is less than the second standard deviation value D2 = 0.011mm, then the feed rate for the current cycle is set to the third feed rate 0.002mm (the minimum feed rate of the current machining equipment), and the grinding of the current machined surface is completed after 5 machining cycles.

[0104] In the above embodiments of the present invention, the feed rate of the current machining cycle is adaptively adjusted according to the dimensional characteristics of the machined surface to achieve intelligent and adaptive control of the feed rate. When the size of the workpiece to be machined is still some distance from the target size, a larger feed rate is used; when the size is close to the target size, a fine feed rate is used to improve the precision of the machined surface. This stepped feed control method helps to effectively improve machining efficiency while ensuring machining accuracy.

[0105] Preferably, step S31 specifically includes:

[0106] Based on the horizontal or numerical inclination of the edge of the surface to be machined or the critical side, calculate the horizontal or numerical angular deviation between the surface to be machined and the standard coordinate system, and adjust the feed angle or feed path accordingly based on the angular deviation.

[0107] Based on the detected three-dimensional coordinate information of the edge, when the machining surface of the workpiece shows a horizontal or vertical angular deviation, the feed angle or feed path can be adjusted accordingly to adapt to the machining target. The feed path or feed angle referred to in the above implementation refers to the relative angle or relative path between the machining surface and the tool. In actual machining, this can be a method where the workpiece is fixed while the tool moves, or a method where the tool is fixed while the workpiece is moved.

[0108] Preferably, step S32 specifically includes:

[0109] Based on the acquired image of the surface to be processed, a machine vision-based grinding surface roughness detection model is used to analyze the roughness evaluation value Ra(t) of the surface to be processed. The roughness evaluation value Ra(t) is compared with the roughness standard RaT to obtain the roughness deviation ΔRa = Ra(t) - RaT. When the roughness deviation ΔRa is greater than the set threshold ΔRaT, the grinding speed is adjusted and reduced.

[0110] Furthermore, when the roughness deviation ΔRa is less than or equal to the set threshold ΔRaT, the grinding speed is adjusted to the standard grinding speed.

[0111] In one exemplary scenario, based on the material, size, and other characteristic parameters of the workpiece to be processed, the surface roughness standard RaT is set to 0.025-0.5μm; the standard grinding speed is adjusted to 20-40mm / min; and the reduced grinding speed is adjusted to 10-20mm / min.

[0112] Based on the image information of the surface to be processed, machine vision-based grinding surface roughness detection is performed. The roughness of the surface to be processed can be quantified based on the trained grinding roughness detection model, and the obtained roughness can be compared with the processing standard to achieve control of grinding speed. Based on the above method, a balance between processing speed and grinding precision can be achieved to meet the processing target requirements and improve the intelligence level of processing control.

[0113] In one exemplary scenario, the grinding speed is the relative displacement speed between the surface to be machined and the cutting tool. Depending on the actual situation, this can be a method where the workpiece is fixed while the cutting tool moves, or a method where the cutting tool is fixed while the workpiece is moved.

[0114] In one exemplary scenario, the machine vision-based grinding surface roughness detection model can be implemented using a pre-trained image processing model that detects roughness of the surface to be processed based on image information, and this invention does not impose specific limitations on this model. For example, in one exemplary scenario, the model is built based on a BP neural network. It performs grayscale processing on the image of the surface to be processed, performs grayscale histogram statistics on the grayscale image, obtains the surface contour curve of the surface to be processed, and constructs a grayscale co-occurrence matrix. Grayscale co-occurrence matrices of 0°, 45°, 90°, and 135° are constructed as feature parameters and input into the trained BP neural network. The BP neural network analyzes the input feature parameters and outputs the corresponding roughness estimate.

[0115] Preferably, step S3 of the method further includes:

[0116] S33 performs image recognition-based defect analysis on the surface to be processed based on the acquired image information of the surface to be processed, detects abnormalities on the surface to be processed, and obtains the defect analysis results of the surface to be processed.

[0117] When the defect analysis result is abnormal, the image information of the surface to be processed corresponding to the abnormal analysis result is extracted and transmitted to the management terminal.

[0118] The above embodiments of the present invention, based on the obtained image of the surface to be processed, further perform abnormality and defect analysis on the surface to be processed, which can identify defective parts to be processed in real time, and thus stop processing or perform targeted processing according to the identified abnormal defects. This helps to avoid wasting processing time (if the processed part has an irreparable defect, it is discarded in time and the next part is replaced for processing, avoiding the waste of processing steps for the damaged part) or to achieve the repair of the processed part (if the defect is repairable, the defective part is repaired through manual intervention, avoiding the waste of parts), and further improve the intelligence level of grinding process control.

[0119] Preferably, step S33 specifically includes:

[0120] A machine vision-based grinding surface defect recognition model analyzes and processes images of the surface to be processed to identify abnormal defects on the surface to be processed, including scratches, pits, deformations, etc.

[0121] The grinding surface defect recognition model is built on a deep learning neural network. The image of the surface to be processed is input into the model, and the neural network extracts multi-dimensional feature information of the image based on the input image. Convolution and pooling are performed on the extracted feature information to obtain a set of feature vectors. Based on the obtained feature vectors, a classifier is used for normalization and classification to finally obtain the recognition result of abnormal defects.

[0122] The grinding surface defect recognition model can be trained based on a training set consisting of multiple sets of part processing surface images with abnormal defect labels. After the model is trained based on the training set, the model is further tested using a test set. When the test results meet the standard, the trained grinding surface defect recognition model is obtained.

[0123] Preferably, step S4 specifically includes:

[0124] Based on the obtained parameters such as feed rate, grinding speed, and anomaly identification results, the machining parameters for the current stroke / process are set until the current stroke / process ends, at which point the process jumps back to step S1 to start execution again until the machined surface reaches the machining target.

[0125] Each successful feed is recorded as the completion of one machining cycle. Before each machining cycle begins, steps S1-S3 are executed to set the machining parameters for the current cycle until the target surface is achieved. This intelligent machining method helps save manpower and resources and improves machining efficiency.

[0126] Preferably, step S4 specifically includes:

[0127] If the dimensional deviation between the size of the surface to be processed and the dimensional standard is less than the preset second deviation value, and the current roughness evaluation value is less than the preset roughness standard, the processing procedure of the current surface is determined to be completed, and the processing of the current surface is stopped, or the processing procedure of the next surface is skipped.

[0128] It should be noted that the functional units / modules in the various embodiments of the present invention can be integrated into one processing unit / module, or each unit / module can exist physically separately, or two or more units / modules can be integrated into one unit / module. The integrated unit / module described above can be implemented in hardware or in the form of software functional units / modules.

[0129] From the above description of the embodiments, those skilled in the art will clearly understand that the embodiments described herein can be implemented in hardware, software, firmware, middleware, code, or any suitable combination thereof. For hardware implementation, the processor can be implemented in one or more of the following units: Application-Specific Integrated Circuit (ASIC), Digital Signal Processor (DSP), Digital Signal Processing Device (DSPD), Programmable Logic Device (PLD), Field-Programmable Gate Array (FPGA), processor, controller, microcontroller, microprocessor, other electronic units designed to implement the functions described herein, or combinations thereof. For software implementation, some or all of the processes of the embodiments can be implemented by a computer program instructing the associated hardware. During implementation, the program can be stored in a computer-readable medium or transmitted as one or more instructions or code on a computer-readable medium. Computer-readable media include computer storage media and communication media, wherein communication media include any medium that facilitates the transmission of a computer program from one place to another. Storage media can be any available medium accessible to a computer. Computer-readable media can include, but is not limited to, RAM, ROM, EEPROM, CD-ROM or other optical disk storage, magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible to a computer.

[0130] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit the scope of protection of the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should be able to analyze that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the essence and scope of the technical solutions of the present invention.

Claims

1. A tool grinding method based on machine vision, characterized in that, include: S1 acquires image data of the part to be processed after it has been fixed in place; S2 extracts the part features of the part to be processed based on the acquired image data, where the part features include size information and image information of the surface to be processed; S3 compares and analyzes the extracted part features with the set machining targets, and generates machining parameters based on the comparison and analysis results. The machining targets include part size standards and machined surface quality standards, and the machining parameters include feed parameters and grinding parameters; specifically including: S31 calculates the dimensional characteristics of the workpiece based on its current dimensional information, compares and analyzes the obtained dimensional characteristics with the preset workpiece dimensional standard to obtain the dimensional deviation, and sets the corresponding feed parameters based on the dimensional deviation, including the feed amount. When the dimensional deviation is greater than the preset first deviation value, adjust the current feed rate to the first feed rate; When the dimensional deviation is less than or equal to the preset first deviation value and greater than or equal to the second deviation value, adjust the current feed rate to the second feed rate. When the dimensional deviation is less than the preset second deviation value, adjust the current feed rate to the third feed rate; The first feed rate is greater than the second feed rate; the second feed rate is greater than the third feed rate. S32 When the dimensional deviation is less than the preset second deviation value, the quality of the surface to be processed is further obtained based on the current image information of the surface to be processed of the part to be processed. The obtained quality of the surface to be processed is compared and analyzed with the preset quality standard of the surface to be processed to obtain the quality deviation of the surface to be processed. The corresponding grinding parameters are set according to the quality deviation of the surface to be processed; the grinding parameters include the grinding speed. S4 processes the surface to be processed according to the processing parameters, and after processing is completed, repeats the above steps S1-S3 until the current processed surface reaches the processing target.

2. The tool grinding method based on machine vision according to claim 1, characterized in that, In step S1, the processing surface image and key side image data of the part to be processed are acquired based on the 3D camera.

3. The tool grinding method based on machine vision according to claim 2, characterized in that, Step S2 includes: performing edge detection processing based on the acquired image data to obtain the edge feature information of the part to be processed; Based on the comparison and analysis of the acquired edge feature information with the preset standard coordinate system, the size information of the processing surface and key side of the part to be processed is obtained as the size information of the part to be processed. The size information includes the coordinate system position, length data, horizontal tilt angle and vertical tilt angle of each edge. Based on the obtained edge feature information, the area to be processed of the part to be processed is further extracted to obtain the image of the surface to be processed.

4. The tool grinding method based on machine vision according to claim 3, characterized in that, Step S2 also includes: The acquired image data of the parts to be processed is preprocessed to obtain the preprocessed image data. Then, edge detection processing is performed on the preprocessed image data to obtain the edge feature information of the parts to be processed.

5. The tool grinding method based on machine vision according to claim 3, characterized in that, Step S31 specifically includes: Calculate the length characteristics of the surface to be machined based on the edge lengths of the surface to be machined and the critical side of the part to be machined; or, calculate the thickness characteristics of the surface to be machined based on the dimensional information of the critical side. The length or thickness characteristics of the surface to be processed are compared with the preset part size standard to obtain the length or thickness dimensional deviation. Furthermore, the dimensional deviation is compared with the preset first deviation value and second deviation value. When the dimensional deviation is greater than the preset first deviation value, adjust the current feed rate to the first feed rate; When the dimensional deviation is less than or equal to the preset first deviation value and greater than or equal to the second deviation value, adjust the current feed rate to the second feed rate. When the dimensional deviation is less than the preset second deviation value, adjust the current feed rate to the third feed rate; Based on the horizontal or numerical inclination of the edge of the surface to be machined or the critical side, calculate the horizontal or numerical angular deviation between the surface to be machined and the standard coordinate system, and adjust the feed angle or feed path accordingly based on the angular deviation.

6. The tool grinding method based on machine vision according to claim 5, characterized in that, Step S32 specifically includes: Based on the acquired image of the surface to be processed, a machine vision-based grinding surface roughness detection model is used to analyze the roughness evaluation value Ra(t) of the surface to be processed. The roughness evaluation value Ra(t) is compared with the roughness standard RaT to obtain the roughness deviation ΔRa = Ra(t) - RaT. When the roughness deviation ΔRa is greater than the set threshold ΔRaT, the grinding speed is adjusted to decrease; when the roughness deviation ΔRa is less than or equal to the set threshold ΔRaT, the grinding speed is adjusted to the standard grinding speed.

7. The tool grinding method based on machine vision according to claim 6, characterized in that, Step S3 also includes: S33 performs image recognition-based defect analysis on the surface to be processed based on the acquired image information of the surface to be processed, detects abnormalities on the surface to be processed, and obtains the defect analysis results of the surface to be processed. When the defect analysis result is abnormal, the image information of the surface to be processed corresponding to the abnormal analysis result is extracted and transmitted to the management terminal, specifically including: A machine vision-based grinding surface defect recognition model analyzes and processes images of the surface to be processed, and identifies abnormal defects on the surface to be processed, including scratches, pits, and deformations.

8. The tool grinding method based on machine vision according to claim 6, characterized in that, Step S4 specifically includes: If the dimensional deviation between the size of the surface to be processed and the dimensional standard is less than the preset second deviation value, and the current roughness evaluation value is less than the preset roughness standard, it is determined that the processing step of the current surface is completed, and the processing of the current surface is stopped, or the processing of the current surface is skipped.

Images