Gear forging control method based on visual guidance

Abstract

The invention provides a gear forging control method based on visual guidance, relates to the field of image processing, and is applied to gear forging quality control in the aviation field. The control method specifically comprises the following steps: shooting a gear panoramic image by using a camera; performing image analysis on the panoramic image to obtain a multi-region spliced image; respectively inputting the multi-region spliced image into a region quality detection model to obtain a multi-region quality result; the gear quality report is generated according to the multi-region quality result, the forging equipment is controlled to reprocess, and each region of the gear is independently detected and analyzed, so that refined quality control can be performed on each part of the gear, the overall quality of the gear is enhanced, and the performance and reliability of the gear in an extreme environment are improved.

Description

Technical Field

[0001] The invention relates to the field of image processing, and in particular to a visual guidance control method for gear forging in the aviation field. Background Art

[0002] As the global aviation market continues to expand, the demand for aircraft is rising year by year. In order to meet the growing market demand, the pace of aircraft production is also accelerating. Among the many aircraft parts, gears are a vital link because they play a core role in the aircraft's power transmission system. Due to the vital function of gears in aircraft, the quality control standards for their production have also been improved accordingly to ensure that they can maintain efficient and safe operation under extreme flight conditions.

[0003] In the manufacturing process of gears, traditional quality inspection methods mainly rely on laser measurement technology to detect the geometric size and shape accuracy of gears to ensure that they meet strict production specifications. This method can effectively measure parameters such as gear diameter, pitch, and tooth shape to verify whether they meet design requirements. However, these traditional technologies mainly focus on the accuracy of size and shape, but do not pay enough attention to the microstructure of gear materials and their performance in actual operating environments.

[0004] Aviation gears not only need to withstand conventional mechanical loads during use, but also need to cope with harsh environments such as high temperature, pressure changes, and corrosion generated during high-speed operation. These extreme conditions require that gears not only meet specifications in terms of size, but also meet higher standards in terms of material uniformity, wear resistance, and fatigue resistance. Traditional quality inspection methods cannot fully evaluate the performance of gears in actual use, especially the pressure and wear resistance of key parts such as tooth grooves and tooth tops, which are usually high-incidence areas for gear failures.

[0005] Due to the limitations of the accuracy and comprehensiveness of traditional methods, there is an urgent need for a gear forging control method that can provide more detailed and accurate detection. The present invention proposes a gear forging control method based on vision guidance. This method uses visual detection technology to perform separate detection and analysis on each area of ​​the gear. Through this method, each part of the gear can be finely controlled, which enhances the overall quality of the gear and improves its performance and reliability in extreme environments. Summary of the invention

[0006] The present invention provides a gear forging control method based on vision guidance, the method comprising the following steps: S1: Use a camera to capture a panoramic image of the gear; S2: performing image analysis on the panoramic image to obtain a multi-region stitching image; S3: inputting the multi-region stitching images into the regional quality detection model respectively to obtain multi-region quality results; S4: Generate a gear quality report based on the multi-region quality results and control the forging equipment for reprocessing.

[0007] The present invention provides a gear forging control system based on vision guidance, the system comprising: Photographing equipment: using a camera to take panoramic images of gears; Image stitching module: the image stitching module performs image analysis on the panoramic image to obtain a multi-region stitching image; Defect detection module: the defect detection module inputs the multi-region stitching images into the regional quality detection model to obtain multi-region quality results; Recommendation module: Generate a gear quality report based on the multi-area quality results and control the forging equipment for reprocessing.

[0008] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned gear forging control method based on vision guidance when executing the computer program.

[0009] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the computer program implements the above-mentioned gear forging control method based on vision guidance.

[0010] Compared with the prior art, the present invention aims to solve the problem of insufficient gear detection during the forging process in the prior art. The present invention analyzes the physical performance requirements of different positions in the gear work, segments different areas in a targeted manner, and performs image enhancement and defect detection based on the segmented different areas, so as to make the quality control more refined. The panoramic image taken once is used for multi-area detection, which improves the detection efficiency and execution convenience. The side image is targeted for affine transformation to improve the detection accuracy. In addition, the present invention splices the tooth tops and tooth grooves with the same physical performance requirements to obtain a complete image, and uses global features to analyze the consistency of each local area, so that the physical properties of each area of ​​the gear are close, avoiding the problem of incomplete processing of local areas. By adopting the above-mentioned forging quality control method, the present invention finely monitors the quality performance of each area, and improves the detection accuracy by using adaptive detection means for different areas. At the same time, it simplifies the complex procedures of multiple shots and multiple detections, and improves the convenience of application. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0012] Figure 1 is a flow chart of a gear forging control method based on vision guidance in this application; DETAILED DESCRIPTION

[0013] The embodiments of the present application are described in detail below with reference to the accompanying drawings.

[0014] The following describes the implementation methods of the present application through specific specific examples, and those skilled in the art can easily understand other advantages and effects of the present application from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. The present application can also be implemented or applied through other different specific implementation methods, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present application. It should be noted that, in the absence of conflict, the following embodiments and the features in the embodiments can be combined with each other. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without making creative work belong to the scope of protection of the present application.

[0015] It should be noted that various aspects of the embodiments within the scope of the appended claims are described below. It should be apparent that the aspects described herein can be embodied in a wide variety of forms, and any specific structure and / or function described herein is merely illustrative. Based on the present application, it should be understood by those skilled in the art that an aspect described herein can be implemented independently of any other aspect, and two or more of these aspects can be combined in various ways. For example, any number and aspect described herein can be used to implement the device and / or practice the method. In addition, other structures and / or functionalities other than one or more of the aspects described herein can be used to implement this device and / or practice this method.

[0016] Additionally, in the following description, specific details are provided to facilitate a thorough understanding of the examples. However, it will be understood by those skilled in the art that the examples can be practiced without these specific details.

[0017] The embodiment of this specification proposes a gear forging control method based on vision guidance, which includes the following steps: S1: Use a camera to capture a panoramic image of the gear; S2: performing image analysis on the panoramic image to obtain a multi-region stitching image; S3: inputting the multi-region stitching images into the regional quality detection model respectively to obtain multi-region quality results; S4: Generate a gear quality report based on the multi-region quality results and control the forging equipment for reprocessing.

[0018] Specifically, S1: using a camera to capture a panoramic image of the gear; The image acquisition area in the present invention is an independent quality inspection workshop, which is equipped with a stable and height-adjustable workbench. The surface of the workbench is flat and has a shockproof function to prevent vibration from affecting the image quality during the shooting process. A rotating table is installed on the workbench, and the gear will be fixed on the rotating table. By accurately controlling the rotation of the rotating table, every angle of the gear can be captured by the camera.

[0019] The camera is fixed on a bracket opposite the workbench, and a tripod or other stabilizing device is used to ensure the stability of the camera during shooting. The lens of the camera should be aligned with the center line of the gear and kept horizontal or slightly tilted to optimize the shooting angle. The camera used in the present invention should have a digital SLR camera with high resolution and wide dynamic range. The camera needs to be equipped with a zoom lens so that the focal length can be adjusted according to the size of the gear to ensure that every part of the gear can be clearly displayed in the image. Preferably, the camera device can be selected such as Canon EOS 5D Mark IV, which provides a resolution of 30.4 million pixels.

[0020] In terms of shooting parameter settings, the camera was set to manual mode to ensure that the exposure and focus remained consistent throughout the shooting process. The aperture was set between f / 8 and f / 11 to obtain a large depth of field and ensure that the entire gear was in focus. The shutter speed could be adjusted according to the lighting conditions on site, and the ISO was kept at a low level to reduce noise.

[0021] Before shooting, the rotating stage is first started to rotate the gear at a uniform speed. The camera is set to continuous shooting mode and automatically shoots at preset intervals (e.g., once every 2 degrees of rotation). In order to ensure the continuity of the image, the coverage area of ​​each shot should overlap to a certain extent. Preferably, the present invention sets the overlap area of ​​the current shot image frame to account for 20-30% of the previous frame.

[0022] When the gear rotates one circle, the shooting ends. At this point, all the captured image frames will be imported into the image processing software for stitching. In the process of stitching the panoramic image, use image processing software (such as Adobe Photoshop or professional panoramic stitching software PTGui) to import the continuously shot pictures. The software will automatically analyze the overlapping parts between each picture and stitch them accurately to generate a seamless gear panoramic image.

[0023] S2: performing image analysis on the panoramic image to obtain a multi-region stitching image; The image analysis of the panoramic image includes extracting broken lines in the panoramic image and detecting broken line defects; S2-1-1: preprocessing the panoramic image, wherein the preprocessing includes denoising, contrast enhancement and sharpening, and highlighting the broken line part in the panoramic image for easy identification; The denoising includes reducing image noise using Gaussian filtering: ; in, (x, y) is the panoramic image, is the Gaussian filter parameter, which controls the degree of blur. is the preprocessed image after Gaussian filtering; The contrast enhancement is used to improve the image foreground contrast, making the polyline more distinct from the background: ; in, is the average brightness of the panoramic image, is the enhancement factor, is the contrast-enhanced image; In one implementation, the preprocessing of the panoramic image is implemented using the following pseudo code.

[0024] def enhance_image(image_path): image = cv2.imread(image_path) if image is None: print("Error: Image not found.") return blurred = cv2.GaussianBlur(image, (5, 5), 0) image_yuv = cv2.cvtColor(blurred, cv2.COLOR_BGR2YUV) image_yuv[:,:,0] = cv2.equalizeHist(image_yuv[:,:,0]) contrast_enhanced = cv2.cvtColor(image_yuv, cv2.COLOR_YUV2BGR) sharpening_kernel = np.array ([[-1, -1, -1], [-1, 9, -1], [-1,-1, -1]]) sharpened = cv2.filter2D(contrast_enhanced, -1, sharpening_kernel) S2-1-2: extracting polylines from the preprocessed panoramic image; Use the Canny algorithm to extract edges in a panoramic image: ;E(x,y) is the binary image after edge detection by Canny algorithm, is the preprocessed panoramic image, They are the low and high thresholds for edge detection, controlling the sensitivity of the edge; Use dilation erosion to enhance the edges: ;in, represents the binary image after dilation and corrosion; K is the structural element; is the offset of the structural element; Perform polyline fitting on the local area of ​​the edge image to obtain a polyline set; for For each edge point (x, y) and each angle θ in the image, calculate ρ=xcosθ+ysinθ, and add 1 to the corresponding (θ, ρ) unit in the accumulator. Find the element value in the accumulator that is higher than the preset threshold. The (θ, ρ) corresponding to these elements is the fitted polyline; polyline set It is expressed as: ; where ρ is the vertical distance from the polyline to the origin of the image coordinate system, θ is the inclination angle of the line, represents the i-th extracted polyline, To extract the total number of broken lines; S2-1-3: performing fold line defect detection in the fold line set, wherein the defects include fold line missing, fold line bending, and local fold line irregularity; Polyline missing detection: For a polyline , define its point set as ,in, Represents a polyline The kth coordinate in Represents a polyline The number of coordinate points included; if the distance between any two adjacent points exceeds the preset threshold, the polyline is considered There is a break: ; in, is the preset missing threshold; Polyline bending detection: For a polyline , construct a fitting straight line , and calculate the polyline The vertical distance from the point to the fitted line. If the vertical distance from the polyline point to the fitted line exceeds the preset threshold, the polyline is considered to be curved: , in, is the preset bending threshold, and They represent the slope and intercept of the fitted line respectively; Polyline local irregularity detection, for a polyline , calculate its local curvature and identify the local irregularity defect of the broken line when the local curvature exceeds the preset threshold: , in, To preset the local irregularity threshold, The first derivative of The second derivative of .

[0025] S2-1-3: For each broken line , record the defect types of broken lines, broken line bending and local irregular broken lines, and record the corresponding positions of the defects, integrate all broken lines, and output the broken line set Corresponding polyline quality inspection report.

[0026] In one embodiment, defect detection for fold lines is implemented by the following code: def euclidean_distance(p1, p2): return np.sqrt((p2[0] - p1[0])**2 + (p2[1] - p1[1])**2) def check_continuity(points, d_max): for i in range(len(points) - 1): if euclidean_distance(points[i], points[i+1])>d_max: return False, i # Returns the location of the break return True, None def line_fit(points): x = points[:, 0] y = points[:, 1] A = np.vstack([x, np.ones(len(x))]).T m, b = np.linalg.lstsq(A, y, rcond=None)[0] return m, b def deviation_from_line(points, m, b): x = points[:, 0] y = points[:, 1] deviations = np.abs(y - (m * x + b)) return np.mean(deviations), np.max(deviations) def curvature(points): x = np.array(points[:, 0], dtype=float) y = np.array(points[:, 1], dtype=float) dx = np.gradient(x) dy = np.gradient(y) ddx = np.gradient(dx) ddy = np.gradient(dy) curvatures = np.abs(ddx * dy - ddy * dx) / (dx**2 + dy**2)**1.5 return curvatures The multi-region stitched image includes a gear plane image stitched by a plurality of tooth top planes and tooth groove planes and a gear side image stitched by a plurality of side surfaces from tooth tops to tooth grooves; The multi-region stitching image construction specifically includes: S2-2-1: Eliminate polyline collection , get a subset containing only vertical lines ,in, It means extracting the jth vertical line in sequence. is the total number of vertical broken lines extracted; S2-2-2: Divide the panoramic image into multiple regions according to the subset of vertical fold lines, each region consists of two adjacent vertical fold lines. The definition is: ,in, They represent the horizontal coordinates corresponding to the y positions on the jth and j+1th polylines, respectively. Represents the minimum and maximum values ​​of the y coordinates on the jth and j+1th polylines, respectively. Total Regions; S2-2-3: Classify the regions by the order of the broken lines according to the periodic characteristics of the image: Tip plane area: ,in, , and m is a positive integer; Tooth plane area: ,in, , and m is a positive integer; Side area from tooth tip to tooth groove: ,in, , and m is a positive integer; S2-2-3: Extract classified images based on image classification results: Image set of tooth addendum plane area: ; Alveolar plane region image set: ; Image set of the side area from tooth tip to tooth groove: ; S2-2-4: According to the spatial order in the original image, all the tooth top plane area image sets and tooth groove plane image sets are stitched together to construct a complete gear plane image The plane image Contains the position information of each tooth top plane and tooth groove plane that constitutes its image.

[0027] S3: Before the multi-region stitching images are input into the regional quality detection model, the multi-region stitching images and the side region image set from the tooth top to the tooth groove need to be image enhanced respectively: For the gear plane image composed of the top of the gear or the tooth groove, the tooth top plane and the tooth groove plane are the direct contact areas of the gear meshing, which require high hardness and smooth surfaces to reduce friction and wear and avoid friction heat or vibration caused by excessive roughness. At the same time, the shooting angle is frontal shooting, and the viewing angle is perpendicular to the tooth top and tooth groove plane. Therefore, in the captured image, the image surface texture is relatively uniform, and the defects may be relatively small and easily masked by background noise. In addition, due to vertical shooting and the requirement for a relatively smooth shooting surface, there may be local overexposure or shadows due to light reflection. Based on the above shooting problems, the gear plane image is adaptively histogram equalized, edge enhanced, and denoised: in, is the normalization coefficient, defines a neighborhood centered at (x, y), and are the spatial Gaussian function and the intensity Gaussian function, Neighborhood The coordinates within; For the side image of the gear obtained by splicing the side of the tooth top to the tooth groove, the tooth side plane is the key part of the gear to transmit torque, which bears shear force and contact stress to avoid local plastic deformation, and the tooth side surface needs to maintain a certain uniform roughness to ensure the formation of lubricating oil film and avoid high temperature caused by long-term contact wear. For the tooth side plane, the shooting angle is an oblique angle. Therefore, in the captured image, the brightness adjustment and spatial affine transformation of different depths of field are used to obtain the front view: ; in, is the perspective transformation matrix, is the gamma correction constant, Indicates that gamma correction is performed on the side area image set to make the brightness distribution of the side area more uniform. represents the front view image set of the side area obtained after the spatial affine transformation; The regional quality detection model includes a gear side region detection model, the gear side region detection model includes a backbone module, the backbone module includes an input layer and 4 intermediate processing layers Layer1-4, the input layer convolution kernel is 3*3, and the step size is 2; Layer 1 contains two residual modules. Each residual module includes 3*3 convolution, BatchNorm and skip connection connected in sequence. Layer 1 is used to extract low-level features such as edges and textures while maintaining a high spatial resolution. Layer 2 contains three depth-wise separable convolutions, using input features to perform jump connections with the output of the depth-wise convolution module. Layer 3 stacks 4 residual modules, each of which contains: a 1×1 convolution for dimensionality reduction and improving nonlinear expression ability, a 3×3 convolution for extracting spatial features, and a 1×1 convolution for dimensionality increase and skip connection; Layer 4 uses 2 dilated convolution modules, each of which contains a dilated convolution of dilation=2 and a 3×3 convolution. The dilated convolution is used to expand the receptive field and capture global features. The features processed by the backbone module are input into the feature enhancement network, which includes enhancement modules En_Block1 and En_Block2. The features processed by the backbone module are sequentially processed by enhancement modules En_Block1 and En_Block2, and then input into the enhancement module En_Block1 for processing to obtain the output of the feature enhancement network, wherein the enhancement modules En_Block1 and En_Block2 are defined as: ; ; The output features of the feature enhancement network are respectively input into the region proposal network and the multi-scale mapping network. The region proposal network generates a ROI map which is mapped to the feature map generated by the multi-scale mapping network to obtain a multi-scale feature map with region proposals. The multi-scale feature map is input into the defect detection head to output the defect type and position coordinates. The gear side regional defects include uneven texture, cracks, and pitting.

[0028] The regional quality detection model also includes a gear plane global detection model, which is used to detect global defects of tooth tops and tooth grooves: Calculate the global brightness gradient of the gear plane image and calculate the average brightness gradient and brightness gradient standard deviation , and the global color entropy is calculated at the same time; ; ; in, Global color entropy, is the normalized frequency of color value t in color channel c, is the Kronecker function, is the color value at (x, yc) after normalization, H and W are the image height and width; Calculate the image set of the tooth top plane area and alveolar plane area image set The local average brightness gradient, local brightness gradient standard deviation and local color entropy of each local image in Determine whether each local image is abnormal based on the preset brightness threshold and color threshold: , then it is determined that the brightness of the local image is abnormal; , then the local image color is considered abnormal; In one implementation, the above global defect judgment is implemented by executing the following code: def normalize_brightness(image): gray = cv2.cvtColor(image, cv2.COLOR_BGR2GRAY) mean, std = gray.mean(), gray.std() normalized = (gray - mean) / std return normalized def normalize_color(image): sum_channels = np.sum(image, axis=2, keepdims=True) normalized = image / (sum_channels + 1e-6) # To avoid division by zero return normalized def compute_gradient(image): grad_x = cv2.Sobel(image, cv2.CV_64F, 1, 0, ksize=3) grad_y = cv2.Sobel(image, cv2.CV_64F, 0, 1, ksize=3) gradient = np.sqrt(grad_x**2 + grad_y**2) return gradient def compute_histogram_entropy(image, bins=256): histograms = [] entropies = [] for channel in range(3): # Calculate for R, G, B channels separately hist, _ = np.histogram(image[:, :, channel], bins=bins, range=(0, 1), density=True) histograms.append(hist) entropy = -np.sum(hist * np.log(hist + 1e-6)) # Avoid log(0) entropies.append(entropy) return histograms, entropies def compute_kl_divergence(hist1, hist2): kl_div = np.sum(hist1 * np.log((hist1 + 1e-6) / (hist2 + 1e-6))) return kl_div def detect_defects(image, regions, T_G, T_color, alpha, beta): brightness = normalize_brightness(image) normalized_color = normalize_color(image) gradient = compute_gradient(brightness) G_mean = gradient.mean() G_std = gradient.std() global_histograms, global_entropies = compute_histogram_entropy(normalized_color) defects = [] for region in regions: x, y, w, h = region region_brightness = brightness[y:y+h, x:x+w] region_color = normalized_color[y:y+h, x:x+w] region_gradient = compute_gradient(region_brightness) G_mean_i = region_gradient.mean() G_std_i = region_gradient.std() S_G = abs(G_mean_i - G_mean) + abs(G_std_i - G_std) region_histograms, region_entropies = compute_histogram_entropy(region_color) region_entropy = np.mean(region_entropies) S_E = abs(region_entropy - global_entropy) S_C = 0 for c in range(3): S_C += compute_kl_divergence(region_histograms[c], global_histograms[c]) S_color = alpha * S_C + beta * S_E if S_G>T_G: defects.append((region,"brightness anomaly")) elif S_color>T_color: defects.append((region,"color abnormality")) return defects After stitching the images of the gear tooth top plane and tooth groove plane into a complete image, global features are extracted from the stitched image and defect detection is performed based on these features, thereby optimizing the efficiency, accuracy and process feedback capability of detection as a whole.

[0029] First, the plane images of the tooth tops and tooth grooves of all gears are spliced ​​into a complete image so that the overall surface characteristics of the gears can be presented uniformly. This splicing method can effectively integrate the key parts of the gear into a global perspective, making it easier to analyze the overall features uniformly. By displaying all the details of the tooth tops and tooth grooves in a complete image, the detection system can more intuitively capture the global distribution of light curvature and color. Light curvature reflects the regularity and continuity of the surface brightness changes with position, while color features reflect the uniformity and consistency of surface color distribution. For the processing of gears, the overall consistency of light curvature and color is an important indicator for evaluating the consistency of the processing process. If the light curvature or color features of a certain area are significantly different from those of other areas, it reflects an abnormality in the processing process. Feature extraction through this global perspective can more comprehensively discover possible defects, which is more efficient and robust than local detection of each tooth surface separately.

[0030] Secondly, global feature extraction after stitching images can significantly improve the stability and reliability of detection. In traditional detection methods, point-by-point detection is usually performed on local areas. The disadvantage of this method is that it is easy to ignore global information, which is crucial for identifying process-related problems. Through the complete image obtained by stitching, the distribution characteristics of global light curvature, such as the mean and standard deviation of the brightness gradient, can be extracted to measure the overall surface smoothness and processing consistency; at the same time, color features such as color histogram and color entropy can be used to evaluate whether the surface color distribution is uniform. Any anomaly in the processing process (such as uneven temperature, inconsistent pressure distribution, or material processing errors) will leave traces in the global distribution of light curvature and color. By comparing the deviation of the characteristics of each local area from the global characteristics, the abnormal area can be quickly located and the nature of these anomalies can be further analyzed.

[0031] Furthermore, this method based on global feature detection is also more advantageous for defect classification and positioning. By stitching images, the defect area can be directly mapped to the original gear surface position, thereby achieving precise spatial positioning. This positioning capability is very important for subsequent quality control and maintenance. For example, if the light curvature of a certain tooth groove is found to be abnormal, the specific position of the tooth groove can be directly located for targeted evaluation and treatment. At the same time, the extraction of global features can provide an important basis for the classification of defects. For example, light curvature abnormalities are usually related to material processing or surface flatness during processing, while color abnormalities may be more related to coating, heat treatment or surface contamination. By combining light curvature and color features, defects can be classified and analyzed more carefully, providing more targeted guidance for subsequent process improvements.

[0032] S4: generating a gear quality report according to the multi-region quality results, and controlling the forging equipment to perform reprocessing; The multi-region quality results include broken line defect detection results: broken line missing, broken line bending, local broken line irregularity; also include gear side inspection results of uneven texture, cracks, pitting and gear plane global inspection results of abnormal brightness and color. The multi-region quality results are annotated to the gear panoramic image and forging processing suggestions are given based on the inspection results.

[0033] The processing suggestions specifically include: The goal of broken line detection is to find geometric anomalies at the junction of adjacent contact surfaces of gears. These broken lines directly affect the meshing performance and transmission efficiency of gears. Missing fold line: Use supplementary processing method to repair the missing fold line through local forging to ensure the continuity of the junction of the contact surface. After processing, precision polishing is performed to eliminate possible surface roughness and prevent stress concentration on the meshing surface.

[0034] Fold line bending: Use cold correction method to adjust the bending area to ensure that the straightness of the fold line meets the design requirements. After the local correction is completed, heat treatment is required to restore the internal stress balance of the material. Local irregularity of fold line: Use precision grinding and micro-forging technology to repair the fluctuation or irregularity of the fold line in the local area, and then perform surface hardening treatment to improve the wear resistance and strength of the repaired area; The area from the tooth top to the tooth groove side is the main area for gears to transmit loads. Its surface quality directly determines the wear resistance and load-bearing capacity of the gears. Uneven texture: Use local polishing and micro-forging technology to eliminate uneven texture and ensure uniform texture of the tooth surface. After the repair is completed, the tooth surface is shot peened or heat treated to improve fatigue resistance. Cracks: Perform deep removal processing on the crack area. Preferably, perform local reforging to fill the cracks. After the cracks are removed, perform high-temperature tempering to restore the material properties of the gear. After processing, perform magnetic particle inspection or penetrant inspection to ensure that the cracks are completely eliminated. Pitting: The pitting area needs to remove the damaged surface layer and supplement the material through micro-forging technology; after repair, the repaired area is nitrided to enhance hardness and corrosion resistance; Abnormal brightness: Use precision polishing technology to restore the surface smoothness and reduce the friction coefficient; after polishing, spray or apply protective coating to the surface to improve wear resistance; Color abnormality: Perform local annealing on the color abnormality area to eliminate the problem of excessive heat treatment, and re-perform heat treatment (such as surface quenching or carburizing quenching) to ensure that the heat treatment hardness and depth meet the design requirements. After the treatment is completed, hardness testing can be performed to verify the heat treatment results.

[0035] Regarding the reprocessing priority, cracks are the most serious defects that may cause gear breakage and need to be removed and repaired first. The broken line problem directly affects the gear meshing performance and should be treated as the second priority. Repair pitting and uneven texture. These defects affect the surface performance of the gear and need to be processed after the key defects are repaired; repair brightness and color abnormalities, and for surface quality issues, finally perform polishing and heat treatment repairs uniformly.

[0036] The present invention provides a gear forging control system based on vision guidance, the system comprising: Photographing equipment: using a camera to take panoramic images of gears; Image stitching module: the image stitching module performs image analysis on the panoramic image to obtain a multi-region stitching image; Defect detection module: the defect detection module inputs the multi-region stitching images into the regional quality detection model to obtain multi-region quality results; Recommendation module: Generate a gear quality report based on the multi-area quality results and control the forging equipment for reprocessing.

[0037] An electronic device comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned gear forging control method based on vision guidance when executing the computer program.

[0038] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the computer program implements the above-mentioned gear forging control method based on vision guidance.

[0039] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0040] In this specification, the same or similar parts between the various embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the embodiments described later, the description is relatively simple, and the relevant parts can be referred to the partial description of the previous embodiments.

[0041] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A gear forging control method based on vision guidance, the method comprising the following steps: S1: Use a camera to capture a panoramic image of the gear; S2: performing image analysis on the panoramic image to obtain a multi-region stitching image; S3: inputting the multi-region stitching images into the regional quality detection model respectively to obtain multi-region quality results; S4: Generate a gear quality report based on the multi-region quality results and control the forging equipment for reprocessing.

2. The gear forging control method based on vision guidance according to claim 1 is characterized in that: The image analysis of the panoramic image includes extracting broken lines in the panoramic image and detecting broken line defects; S2-1-1: Panoramic image Performing preprocessing, the preprocessing including denoising, contrast enhancement and sharpening; S2-1-2: extracting polylines from the preprocessed panoramic image; Get polyline set; polyline set It is expressed as: ; represents the i-th extracted polyline, To extract the total number of broken lines; S2-1-3: performing fold line defect detection in the fold line set, wherein the defects include fold line missing, fold line bending, and local fold line irregularity; Polyline missing detection: For a polyline , define its point set as ,in, Represents a polyline The kth coordinate in Represents a polyline The number of coordinate points included; if the distance between any two adjacent points exceeds the preset threshold, the polyline is considered There is a break: ; in, is the preset missing threshold; Polyline bending detection: For a polyline , construct a fitting straight line , and calculate the polyline The vertical distance from the point to the fitted line. If the vertical distance from the polyline point to the fitted line exceeds the preset threshold, the polyline is considered to be curved: ;in, is the preset bending threshold, and represent the slope and intercept of the fitted line, respectively; Polyline local irregularity detection, for a polyline , calculate its local curvature and identify the local irregularity defect of the broken line when the local curvature exceeds the preset threshold: , in, To preset the local irregularity threshold, The first derivative of The second derivative of ; S2-1-4: For each broken line , record the defect types of broken lines, broken line bending and local irregular broken lines, and record the corresponding positions of the defects, integrate all broken lines, and output the broken line set Corresponding polyline quality inspection report.

3. The gear forging control method based on vision guidance according to claim 2 is characterized in that: The multi-region stitching image construction specifically includes: S2-2-1: Eliminate polyline collection , get a subset containing only vertical lines ,in, It means extracting the jth vertical line in sequence. is the total number of vertical broken lines extracted; S2-2-2: Divide the panoramic image into multiple regions according to the subset of vertical fold lines, each region consists of two adjacent vertical fold lines. The definition is: ,in, They represent the horizontal coordinates corresponding to the y positions on the jth and j+1th polylines, respectively. Represents the minimum and maximum values ​​of the y coordinates on the jth and j+1th polylines, respectively. Total Regions; S2-2-3: Classify the regions by the order of the broken lines according to the periodic characteristics of the image: Tip plane area: ,in, , and m is a positive integer; Tooth plane area: ,in, , and m is a positive integer; Side area from tooth tip to tooth groove: ,in, , and m is a positive integer; S2-2-3: Extract classified images based on image classification results: Image set of tooth top plane area: ; Alveolar plane region image set: ; Image set of the side area from tooth tip to tooth groove: ; S2-2-4: According to the spatial order in the original image, all the tooth top plane area image sets and tooth groove plane image sets are stitched together to construct a complete gear plane image The plane image Contains the position information of each tooth top plane and tooth groove plane that constitutes its image.

4. The gear forging control method based on vision guidance according to claim 3 is characterized in that: Before the multi-region stitched images are input into the regional quality detection model, the multi-region stitched images and the side region image set from the tooth top to the tooth groove need to be enhanced respectively: adaptive histogram equalization, edge enhancement and denoising of the gear plane image: ; ; ; in, is the normalization coefficient, defines a neighborhood centered at (x, y), and are the spatial Gaussian function and the intensity Gaussian function, Neighborhood The coordinates inside.

5. The gear forging control method based on vision guidance according to claim 4 is characterized in that: Perform brightness adjustment and spatial affine transformation on the image set of the side area from the tooth top to the tooth groove: ; in, is the perspective transformation matrix, is the gamma correction constant, Indicates that gamma correction is performed on the side area image set to make the brightness distribution of the side area more uniform. Represents the set of frontal images of the side area obtained after spatial affine transformation.

6. The gear forging control method based on vision guidance according to claim 1 is characterized in that: The regional quality detection model includes a gear side region detection model, the gear side region detection model includes a backbone module, the backbone module includes an input layer and four intermediate processing layers Layer1-4, the features processed by the backbone module are input into a feature enhancement network, the feature enhancement network includes enhancement modules En_Block1 and En_Block2, the features processed by the backbone module are sequentially processed by enhancement modules En_Block1 and En_Block2, and then input into the enhancement module En_Block1 for processing to obtain the output of the feature enhancement network, wherein the enhancement modules En_Block1 and En_Block2 are respectively defined as: ; ; in, Input feature map to the module, and represents element-wise addition and multiplication, represents the weight parameter; The output features of the feature enhancement network are respectively input into the region proposal network and the multi-scale mapping network. The region proposal network generates a ROI map which is mapped to the feature map generated by the multi-scale mapping network to obtain a multi-scale feature map with region proposals. The multi-scale feature map is input into the defect detection head to output the defect type and position coordinates. The gear side regional defects include uneven texture, cracks, and pitting.

7. The gear forging control method based on vision guidance according to claim 3 is characterized in that: The regional quality detection model also includes a gear plane global detection model, which is used to detect global defects of tooth tops and tooth grooves: Calculate the global brightness gradient of the gear plane image and calculate the average brightness gradient and brightness gradient standard deviation , and the global color entropy is calculated at the same time; ; ; in, Global color entropy, is the normalized frequency of color value t in color channel c, is the Kronecker function, is the color value at (x, y, c) after normalization, H and W are the image height and width; Calculate the image set of the tooth top plane area and alveolar plane area image set The local average brightness gradient, local brightness gradient standard deviation and local color entropy of each local image in; Determine whether each local image is abnormal based on the preset brightness threshold and color threshold: , then it is determined that the brightness of the local image is abnormal; , then the local image color is considered abnormal.

8. A vision-guided gear forging control system, used to execute the vision-guided gear forging control method according to any one of claims 1 to 7, characterized in that The system includes: Photographing equipment: using a camera to take panoramic images of gears; Image stitching module: the image stitching module performs image analysis on the panoramic image to obtain a multi-region stitching image; Defect detection module: the defect detection module inputs the multi-region stitching images into the regional quality detection model to obtain multi-region quality results; Recommendation module: Generate a gear quality report based on the multi-area quality results and control the forging equipment for reprocessing.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the vision-guided gear forging control method as described in any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the vision-guided gear forging control method according to any one of claims 1 to 7.

Images