A visual-based gear contact pattern acquisition and recognition method

Abstract

The application belongs to the field of mechanical transmission, and discloses a gear contact mark collection and identification method based on vision. The gear meshing video is collected in real time through an industrial camera, the video is frame extracted and converted into a rectangular image, the rectangular image is pretreated, the pretreated rectangular image is converted into an HSV color space and threshold segmented, and a contact mark area is separated out, the contact mark area is morphologically processed, the edge features of the contact mark area are extracted, and finally, contact mark information is calculated. Through the method, real-time dynamic online collection of contact mark feature information can be realized, and through the introduction of the HSV color space adaptive segmentation and multi-scale morphological processing algorithm, the feature information recognition accuracy under poor exposure conditions (overexposure and underexposure) is further improved.

Description

Technical Field

[0001] This invention relates to the field of mechanical transmission, and specifically to a vision-based method for collecting and identifying gear contact imprints. Background Technology

[0002] Gear transmission is the main way of transmitting power and motion in mechanical transmission, and it plays an important role in the transmission of mechanical motion. The stability and reliability of gear transmission are the basis for ensuring the normal operation of machinery. The contact imprint characteristics of the meshing area are an important indicator for evaluating the assembly quality of gear pairs. Therefore, how to achieve efficient and accurate extraction of contact imprint characteristics is of great significance for the accurate evaluation of the assembly quality of transmission systems.

[0003] With the rapid development of image processing theory, machine vision technology has been gradually applied to the identification of contact imprint feature information in gear meshing areas. CN118096665A discloses a visual measurement method, device, and storage medium for contact imprints of aerospace spiral bevel gears. It utilizes acquired laser point cloud data and performs matrix transformation, tooth surface fitting, tooth surface contour point cloud segmentation, and unfolding to automatically measure imprint parameters. CN118817296A discloses a gear meshing quality assessment method based on contact imprint image recognition and correction. It uses machine vision to design a processing method for identifying contact imprints in the meshing area of ​​gears, qualitatively evaluating the quality of gear meshing using the area ratio, tooth height ratio, tooth width ratio, and Euclidean distance between the centroids of the contact imprints and individual gear tooth surfaces. However, the above methods all acquire contact imprint images in a static state after the gears have completed running-in, making dynamic online recognition impossible. Furthermore, since exposure quality significantly affects the extraction of contact imprint feature information during dynamic acquisition, these patents do not address this issue. Summary of the Invention

[0004] To address the problems in existing technologies, dynamically identify gear contact mark features online, and reduce the impact of exposure quality, this application provides a vision-based method for acquiring and identifying gear contact marks.

[0005] The following technical solution will be adopted:

[0006] Real-time video capture of gear meshing is obtained using an industrial camera;

[0007] Frames are extracted from the video to obtain irregular images of the gear surface, and the irregular images are converted into rectangular images.

[0008] The rectangular image is preprocessed to filter out noise that interferes with recognition;

[0009] The preprocessed rectangular image is then converted to an HSV color space.

[0010] The rectangular image in the HSV color space is thresholded to separate the contact imprint region from the rectangular image;

[0011] The contact imprint area is subjected to morphological processing, which removes isolated noise points by opening operations and fills holes by closing operations.

[0012] The edge features of the contact imprint area are extracted;

[0013] Calculate contact imprint information, which includes contact imprint area, tooth width ratio, and tooth height ratio.

[0014] Furthermore, when extracting frames from the video to obtain irregular images of the gear surface, a phase compensation method based on angle increments is used for frame extraction, as detailed below:

[0015] Initially, the cumulative angle is 0°, and the first trigger angle is set to the position of the first selected tooth surface;

[0016] For each frame processed, the cumulative angle increases by one frame rotation angle. When the cumulative angle is greater than the initial trigger angle, it is determined that the current tooth surface has reached the frame extraction position, and frame extraction is triggered.

[0017] After triggering frame skipping, calculate the overshoot Os by which the actual angle exceeds the target angle:

[0018]

[0019] Where Os represents overshoot; Ca represents cumulative angle; and Nta represents frame-triggered angle.

[0020] The compensation amount C is calculated using modulo operations:

[0021]

[0022] By continuously updating the next frame-skipping trigger angle and the current cumulative angle, the calculation formulas for the frame-skipping trigger angle and the cumulative angle are as follows:

[0023]

[0024] Where Dpt represents the central angle corresponding to a single tooth face;

[0025] The formula for calculating the rotation angle Dps per frame is as follows:

[0026]

[0027] in, Rotational speed, in revolutions per minute; This refers to the video frame rate, measured in Hz.

[0028] The central angle corresponding to a single tooth surface The calculation formula is as follows:

[0029]

[0030] in This represents the number of teeth on the gear.

[0031] Furthermore, when converting the irregularly acquired image into a rectangular image, a perspective transformation algorithm is used.

[0032] Furthermore, when preprocessing the rectangular image to filter out noise that interferes with recognition, contrast enhancement and Gaussian blur algorithms are employed.

[0033] Furthermore, the preprocessed rectangular image undergoes color space conversion, specifically converting from RGB color space to HSV color space as follows:

[0034] Convert the RGB values ​​of a rectangular image to a range of 0 to 1;

[0035] The specific formula for calculating the HSV value of a rectangular image is as follows:

[0036]

[0037] in, This represents the calculated lightness component; This represents the calculated saturation component; This represents the calculated hue component; These represent the component values ​​of the red, green, and blue channels in the input RGB color model, respectively. This represents the function that takes the maximum value. This represents the function that takes the minimum value; when the calculated value is... If the value is less than 0, add 360 to the value;

[0038] Convert the values ​​corresponding to the rectangular image to a range of 0 to 255 to determine the red threshold range.

[0039] Furthermore, when performing threshold segmentation on the rectangular image in the HSV color space to separate the contact imprint region from the rectangular image, the specific steps are as follows:

[0040] The red threshold range is defined as follows:

[0041]

[0042] in, and Indicates two ranges of values. The value range of is [0, 10]; The value range of is [170, 179]; Represents rust red; Represents pure red;

[0043] Merge the rust-red mask, then merge the pure red area mask, and finally merge the two masks into a single red area mask.

[0044] Furthermore, the edge features of the contact imprint area are extracted as follows:

[0045] When extracting edge features of the contact imprint area, Canny edge detection is used, and the dual-threshold detection rules are as follows:

[0046] Pixels exceeding a high threshold are marked as strong edges;

[0047] Pixels below a low threshold are marked as non-edges;

[0048] Pixels in between are marked as weak edges if they are connected to strong edges, otherwise they are considered noise removal.

[0049] Furthermore, the high threshold is 150; the low threshold is 50.

[0050] Furthermore, the contact imprint information includes the contact imprint area, tooth width ratio, and tooth height ratio.

[0051] The contact imprint is binarized to obtain a A digital image (two-dimensional pixel matrix), in which, and All are positive integers greater than 1. This indicates the height of the image, which is the number of rows of pixels in the image. This indicates the width of the image, which is the number of columns of pixels in the image.

[0052] The contact imprint is binarized to obtain a digital images, in which, and All are positive integers greater than 1. This indicates the height of the image, which is the number of rows of pixels in the image. This indicates the width of the image, which is the number of columns of pixels in the image.

[0053] Will The digital image undergoes morphological processing, and the sum of the pixels in the processed digital image is identified as the contact imprint area. The expression is:

[0054]

[0055] in, This indicates the location of the processed digital image at coordinates ( , The pixel value at () and It is a digital image pixel coordinate index. Indicates the column number, from arrive ; Indicates the line number, from arrive ;

[0056] The length of the major axis of the circumscribed ellipse of the meshing region minor axis length Establish a rectangular region that encloses the entire tooth surface. The pixel size of this rectangular region in the image is [missing information]. ,in Represents the number of pixels along the width of the rectangle. Represents the number of pixels along the height of the rectangle. and All are positive integers greater than 1;

[0057] Tooth height ratio The expression is as follows:

[0058]

[0059] Tooth width ratio The expression is as follows:

[0060] .

[0061] The beneficial effects of this invention are as follows:

[0062] In existing technologies, the collection of gear contact marks is completed when the gear is stationary after running-in, and is greatly affected by the exposure quality.

[0063] Using the method of this application, real-time dynamic online acquisition of contact imprint feature information can be achieved. By introducing HSV color space adaptive segmentation and multi-scale morphological processing algorithms, the accuracy of feature information recognition under poor exposure conditions (overexposure, underexposure) can be further improved. Attached Figure Description

[0064] Figure 1 This is a flowchart of a vision-based method for collecting and recognizing gear contact marks;

[0065] Figure 2 This is the architecture of the gear contact imprint acquisition and recognition system according to an embodiment of the present invention;

[0066] Figure 3 These are three tooth surface images under normal exposure according to an embodiment of the present invention;

[0067] Figure 4 These are three images of tooth surface features extracted under normal exposure according to an embodiment of the present invention;

[0068] Figure 5 These are three overexposed images of the tooth surface according to an embodiment of the present invention;

[0069] Figure 6 These are three overexposed tooth surface feature extraction images from embodiments of the present invention. Detailed Implementation

[0070] The present invention will now be described in further detail.

[0071] This invention provides a vision-based method for collecting and recognizing gear contact marks, the flowchart of which is shown below. Figure 1 As shown, it includes the following steps:

[0072] S10. Real-time acquisition of gear meshing video via industrial camera;

[0073] Real-time acquisition of gear meshing video, recording a dynamic image sequence of the gear pair meshing process, as detailed below:

[0074] A gear contact imprint acquisition and recognition system is constructed, and the system architecture is as follows: Figure 2 As shown. The motor drives the gear at its output end to rotate, and this gear meshes with the gear ring on the slewing bearing. Through mechanical transmission, the slewing bearing rotates synchronously, thus achieving gear meshing. The light source includes a light source controller and a bar light source, used to provide supplemental lighting to the gear meshing area. The industrial camera acquires an image of the spur gear meshing area of ​​the antenna mount at a distance L from the meshing area (in practical applications, the distance L can be flexibly adjusted according to specific imaging requirements, lens focal length, and target size, as long as a clear and complete contact imprint image is obtained during gear meshing). The image is then converted from analog to digital (A / D) by an image sensor, and the converted data stream is sent to the host computer via a gigabit network cable. The host computer receives the gear meshing video captured by the industrial camera; the host computer software processes the data stream received by the host computer.

[0075] In this embodiment, the parameters of the gear pair are shown in Table 1.

[0076] Table 1. Basic Geometric Parameters of Gear Pairs

[0077]

[0078] Before the gear pair is run-in, apply red lead to the tooth surface of the driven gear and rotate it at a speed of 2 r / min.

[0079] The video of gear meshing is captured by an industrial camera, converted by an image sensor, and the converted data stream is sent to the host computer.

[0080] S20. Extract frames from the video to obtain irregular acquisition images of the gear surface, and convert the irregular acquisition images into rectangular images;

[0081] When the motor drives the gear to rotate at a constant speed and the encoder outputs angle pulses, the actual rotation angle and the theoretical angle will have a phase deviation due to speed fluctuations and transmission backlash. If frames are directly extracted according to a fixed angle increment, tooth surface misalignment, ghosting, or missed sampling will occur. By adopting the phase compensation method, the encoder angle phase error can be corrected in real time. The uniform angle increment is used as the frame extraction trigger reference, and only the image of the key meshing area of ​​the gear is retained, while invalid background frames are discarded.

[0082] The method used to extract frames from the video to obtain irregular images of the gear surface is as follows:

[0083] The phase compensation method is used to extract frames based on angle increments to obtain gear surface images within key areas. The specific steps are as follows:

[0084] Step 201: Initially, the cumulative angle is 0°, and the first trigger angle is set to the position of the first selected tooth surface;

[0085] Step 202: For each frame processed, the cumulative angle is increased by a rotation angle per frame. When the cumulative angle is greater than the initial trigger angle, it is determined that the current tooth surface has reached the frame extraction position, and frame extraction is triggered.

[0086] Step 203: After triggering frame skipping, calculate the overshoot by which the actual angle exceeds the target angle:

[0087]

[0088] Where Os represents overshoot; Ca represents cumulative angle; and Nta represents frame-triggered angle.

[0089] Step 204: Calculate the compensation amount C using modulo operation:

[0090]

[0091] By continuously updating the next frame-drawing trigger angle (Nta) and the current cumulative angle (Ca), the baseline for subsequent frame-drawing is ensured. The calculation formulas for the frame-drawing trigger angle (Nta) and the cumulative angle (Ca) are as follows:

[0092]

[0093] Where Dpt represents the central angle corresponding to a single tooth face.

[0094] The formula for calculating the rotation angle Dps per frame is as follows:

[0095]

[0096] Where n is the rotational speed, in revolutions per minute; Fr is the video frame rate, in Hertz.

[0097] The formula for calculating the central angle Dpt corresponding to a single tooth surface is as follows:

[0098]

[0099] Where Nt is the number of teeth on the gear.

[0100] The perspective transformation algorithm is used to convert irregularly acquired images into rectangular images and store them for convenient subsequent image preprocessing.

[0101] S30. Preprocess the rectangular image to filter out noise that interferes with recognition;

[0102] During the preprocessing of the rectangular image, noise elements that are identified are removed by contrast enhancement and Gaussian blurring algorithms.

[0103] S40. Perform color space conversion on the preprocessed rectangular image, converting the RGB (red, green, blue) color space to the HSV (hue, saturation, brightness) color space;

[0104] Because industrial environments often involve flickering light sources, reflections, shadows, and uneven brightness on workpiece surfaces, the brightness values ​​of the same color can vary greatly under RGB. After separating brightness using HSV, the target color can be stably identified using only H, S, and V.

[0105] The specific method for color space conversion is as follows:

[0106] S401. Convert the RGB values ​​of the rectangular image to a range of 0 to 1;

[0107] S402. Calculate the HSV value of the rectangular image using the following formula:

[0108]

[0109] in, This represents the calculated lightness component; This represents the calculated saturation component; This represents the calculated hue component; These represent the component values ​​of the red, green, and blue channels in the input RGB color model, respectively. This represents the function that takes the maximum value. This represents the function that takes the minimum value; when the calculated value is... If the value is less than 0, add 360 to the value.

[0110] S403. Convert the corresponding values ​​of the rectangular image to a range of 0 to 255, and perform dynamic analysis. The channel histogram adaptively determines the red threshold range.

[0111] S50. Threshold segmentation is performed on the rectangular image in the HSV color space to separate the contact imprint area from the rectangular image;

[0112] The specific steps for color thresholding are as follows:

[0113] S501. Define the red threshold range: Considering the color change caused by the thickness of the red lead powder applied to the tooth surface, and taking into account the HSV values ​​of rust red, pure red, and adjacent colors such as rust yellow, the predefined red threshold range is as follows:

[0114]

[0115] in and Indicates two ranges of values. express The range of values ​​with lower values ​​is [0, 10]. express The range of values ​​for the higher value is [170, 179]. Represents rust red; It represents pure red.

[0116] S502. Merge the rust-red mask, then merge the pure red area mask, and finally merge the two masks into a single red area mask.

[0117] S60. Perform morphological processing on the contact imprint area, remove isolated noise points by opening operation, and fill holes by closing operation.

[0118] S70. Extract the edge features of the contact imprint area;

[0119] When extracting edge features of the contact imprint area, Canny edge detection is used, including Gaussian filtering to smooth the image and suppress noise; calculating the gradient magnitude and direction of each pixel in the filtered image; traversing the gradient magnitude matrix and retaining only the pixels with local maximum values ​​in the gradient direction to refine the edges; dual threshold detection and edge connection, using a high threshold and a low threshold to distinguish strong edge pixels, weak edge pixels and non-edge pixels, and confirming weak edges connected to strong edges as true edges through edge connection.

[0120] When using dual threshold monitoring, the specific rules are as follows:

[0121] Pixels above the high threshold (150) are marked as strong edges;

[0122] Pixels below the low threshold (50) are marked as non-edge;

[0123] Pixels in between are marked as weak edges if they are connected to strong edges, otherwise they are considered noise removal.

[0124] S80. Calculate contact imprint information, which includes contact imprint area, tooth width ratio, and tooth height ratio.

[0125] The contact imprint is binarized to obtain a A digital image (two-dimensional pixel matrix), in which, and All are positive integers greater than 1. This indicates the height of the image, which is the number of rows of pixels in the image. This indicates the width of the image, which is the number of columns of pixels in the image.

[0126] Will The digital image undergoes morphological processing (opening operations remove isolated noise points, and closing operations fill holes). The sum of the pixels in the processed digital image is then identified as the contact imprint area. The expression is:

[0127]

[0128] This indicates the location of the processed digital image at coordinates ( , The pixel value at () and It is a digital image pixel coordinate index. Indicates the column number, from arrive ; Indicates the line number, from arrive .

[0129] The length of the major axis of the circumscribed ellipse of the meshing region minor axis length Establish a rectangular region that encloses the entire tooth surface. The pixel size of this rectangular region in the image is [missing information]. ,in Represents the number of pixels along the width of the rectangle. Represents the number of pixels along the height of the rectangle. and All are positive integers greater than 1.

[0130] Tooth height ratio The expression is as follows:

[0131]

[0132] Tooth width ratio The expression is as follows:

[0133]

[0134] In this embodiment, following the above implementation steps, three normally exposed tooth surface images ( Figure 3 The images are collected and analyzed, and the extracted images are as follows: Figure 4 As shown in Table 2, the characteristic parameters of the gear contact imprint extracted using the method of this application are compared with those measured manually.

[0135] Table 2 Comparison of Feature Parameters Extracted by the Method of this Application and the Manual Measurement Method (Under Normal Exposure Conditions)

[0136]

[0137] The area ratio represents the proportion of the contact imprint area to the theoretically available contact area.

[0138] It is evident that, under normal exposure conditions, the feature information recognition accuracy using the method of this application is high, and the error ratio of each feature parameter is less than 10% compared with that of manually measured parameters.

[0139] Then examine the three overexposed tooth surface images ( Figure 5 The images are collected and analyzed, and the extracted images are as follows: Figure 6 As shown in Table 3, the feature parameters extracted under overexposure conditions and the parameters measured manually are compared.

[0140] Table 3 Comparison of Feature Parameters Extracted by the Method of this Application and the Manual Measurement Method (Under Overexposure Conditions)

[0141]

[0142] It is evident that even under poor exposure conditions, the feature parameters extracted by the method of this application are not significantly different from those under normal exposure conditions. Compared with manually measured parameters, the error ratio of each feature parameter is less than 10%, and it still has high recognition efficiency and accuracy.

[0143] While the present invention has been disclosed above with reference to preferred embodiments, these embodiments are not intended to limit the invention. Any equivalent changes or modifications made without departing from the spirit and scope of the invention are also within the scope of protection of the invention. Therefore, the scope of protection of the present invention should be determined by the claims of this application.

Claims

1. A vision-based method for collecting and identifying gear contact imprints, characterized in that, Includes the following steps: Real-time video capture of gear meshing is obtained using an industrial camera; Frames are extracted from the video to obtain irregular images of the gear surface, and the irregular images are converted into rectangular images. The rectangular image is preprocessed to filter out noise that interferes with recognition; The preprocessed rectangular image is then converted to an HSV color space. The rectangular image in the HSV color space is thresholded to separate the contact imprint region from the rectangular image; The contact imprint area is subjected to morphological processing, which removes isolated noise points by opening operations and fills holes by closing operations. The edge features of the contact imprint area are extracted; Calculate contact imprint information, which includes contact imprint area, tooth width ratio, and tooth height ratio.

2. The vision-based gear contact imprint acquisition and recognition method according to claim 1, characterized in that, When extracting frames from the video to obtain irregular images of the gear surface, a phase compensation method based on angle increments is used for frame extraction, as detailed below: Initially, the cumulative angle is 0°, and the first trigger angle is set to the position of the first selected tooth surface; For each frame processed, the cumulative angle increases by one frame rotation angle. When the cumulative angle is greater than the initial trigger angle, it is determined that the current tooth surface has reached the frame extraction position, and frame extraction is triggered. After triggering frame skipping, calculate the overshoot Os by which the actual angle exceeds the target angle: Where Os represents overshoot; Ca represents cumulative angle; and Nta represents frame-triggered angle. The compensation amount C is calculated using modulo operations: By continuously updating the next frame-skipping trigger angle and the current cumulative angle, the calculation formulas for the frame-skipping trigger angle and the cumulative angle are as follows: Where Dpt represents the central angle corresponding to a single tooth face; The formula for calculating the rotation angle Dps per frame is as follows: in, Rotational speed, in revolutions per minute; This refers to the video frame rate, measured in Hz. The central angle corresponding to a single tooth surface The calculation formula is as follows: in This represents the number of teeth on the gear.

3. The vision-based gear contact imprint acquisition and recognition method according to claim 2, characterized in that, When converting the irregularly acquired image into a rectangular image, a perspective transformation algorithm is used.

4. The vision-based gear contact imprint acquisition and recognition method according to claim 1, characterized in that, When preprocessing the rectangular image to filter out noise that interferes with recognition, contrast enhancement and Gaussian blur algorithms are used.

5. The vision-based gear contact imprint acquisition and recognition method according to claim 1, characterized in that, When converting the color space of the preprocessed rectangular image from RGB to HSV color space, the specific steps are as follows: Convert the RGB values ​​of a rectangular image to a range of 0 to 1; The specific formula for calculating the HSV value of a rectangular image is as follows: in, This represents the calculated lightness component; This represents the calculated saturation component; This represents the calculated hue component; These represent the component values ​​of the red, green, and blue channels in the input RGB color model, respectively. This represents the function that takes the maximum value. This represents the function that takes the minimum value; when the calculated value is... If the value is less than 0, add 360 to the value; Convert the values ​​corresponding to the rectangular image to a range of 0 to 255 to determine the red threshold range.

6. The vision-based gear contact imprint acquisition and recognition method according to claim 1, characterized in that, When performing threshold segmentation on the rectangular image in the HSV color space to separate the contact imprint region from the rectangular image, the specific steps are as follows: The red threshold range is defined as follows: in, and Indicates two ranges of values. The value range of is [0, 10]; The value range of is [170, 179]; Represents rust red; Represents pure red; Merge the rust-red mask, then merge the pure red area mask, and finally merge the two masks into a single red area mask.

7. The vision-based gear contact imprint acquisition and recognition method according to claim 1, characterized in that, The edge features of the contact imprint area are extracted as follows: When extracting edge features of the contact imprint area, Canny edge detection is used, and the dual-threshold detection rules are as follows: Pixels exceeding a high threshold are marked as strong edges; Pixels below a low threshold are marked as non-edges; Pixels in between are marked as weak edges if they are connected to strong edges, otherwise they are considered noise removal.

8. The vision-based gear contact imprint acquisition and recognition method according to claim 7, characterized in that, The high threshold is 150; the low threshold is 50.

9. The vision-based gear contact imprint acquisition and recognition method according to claim 1, characterized in that, Calculate contact imprint information, which includes contact imprint area, tooth width ratio, and tooth height ratio; The contact imprint is binarized to obtain a A digital image (two-dimensional pixel matrix), in which, and All are positive integers greater than 1. This indicates the height of the image, which is the number of rows of pixels in the image. This indicates the width of the image, which is the number of columns of pixels in the image. The contact imprint is binarized to obtain a digital images, in which, and All are positive integers greater than 1. This indicates the height of the image, which is the number of rows of pixels in the image. This indicates the width of the image, which is the number of columns of pixels in the image. Will The digital image undergoes morphological processing, and the sum of the pixels in the processed digital image is identified as the contact imprint area. The expression is: in, This indicates the location of the processed digital image at coordinates ( , The pixel value at () and It is a digital image pixel coordinate index. Indicates the column number, from arrive ; Indicates the line number, from arrive ; The length of the major axis of the circumscribed ellipse of the meshing region minor axis length Establish a rectangular region that encloses the entire tooth surface. The pixel size of this rectangular region in the image is [missing information]. ,in Represents the number of pixels along the width of the rectangle. Represents the number of pixels along the height of the rectangle. and All are positive integers greater than 1; Tooth height ratio The expression is as follows: Tooth width ratio The expression is as follows: 。

Images