A system and method for measuring a profile

Abstract

The application relates to the technical field of detection, and particularly discloses a special-shaped hole contour line measurement system and method, wherein the method comprises the following contents: S1, obtaining a special-shaped hole contour line image; S2, inputting the special-shaped hole contour line image into a special-shaped hole contour line detection model, judging whether the special-shaped hole contour line is qualified according to a recognition result, if qualified, completing detection, if not qualified, jumping to S3; S3, enlarging the contour line in the special-shaped hole contour line image; S4, extracting a contour of the enlarged contour line; S5, performing contour enhancement on the special-shaped hole contour line image; S6, extracting a fingerprint of the special-shaped hole contour line image; S7, calculating the similarity between an actual contour line of the special-shaped hole and a preset theoretical contour line through fingerprint comparison; S8, judging whether the similarity belongs to a preset interval, if belonging to the preset interval, judging that the special-shaped hole contour line is qualified, if not belonging to the preset interval, judging that the special-shaped hole contour line is not qualified. The technical scheme of the application can more accurately and conveniently judge whether the special-shaped hole of a product is qualified.

Description

Technical Field

[0001] This invention relates to the field of detection technology, and in particular to a measurement system and method for irregular hole profiles. Background Technology

[0002] Many components involve various holes, but traditional regular holes, such as cylindrical ones, are prone to fatigue damage under long-term alternating stress. To extend the fatigue life of these components, designers create irregular hole shapes, such as flared or inward-facing holes, to ensure a more uniform stress distribution at the hole location, as seen in the irregular hole at the small end of an engine connecting rod. However, due to the spatial limitations of these irregular holes, it is difficult to accurately detect the crucial parameters of their shape.

[0003] Therefore, there is a need for a system and method for measuring the shape parameters of holes accurately and conveniently, so as to screen out non-conforming products with irregular hole profiles. Summary of the Invention

[0004] One of the objectives of this invention is to provide a method for measuring the profile of irregular holes, which can more accurately and conveniently determine whether the irregular holes of a product are qualified.

[0005] To solve the above-mentioned technical problems, this application provides the following technical solution: A method for measuring irregular hole profiles, characterized by comprising the following: S1. Obtain the irregular aperture profile image using the principle of laser triangular reflection; S2. Input the irregular hole line image into the pre-trained irregular hole line detection model. Determine whether the irregular hole line is qualified based on the recognition result of the irregular hole line detection model. If qualified, the detection is completed; if not qualified, jump to S3. S3. Use the Johannes algorithm to magnify the profile lines in the irregular hole profile line image; S4. Perform Gabor wavelet processing on the magnified profile to extract the contour; S5. Use a fixed threshold binarization algorithm to enhance the contour of the irregular hole line image; S6. The contour-enhanced irregular hole line image is processed by the surviving point extraction method to obtain the fingerprint of the irregular hole line image; S7. Calculate the similarity between the actual profile of the irregular hole and the preset theoretical profile by comparing fingerprints; S8. Determine whether the similarity falls within the preset range. If it falls within the preset range, the irregular hole shape line is deemed qualified. If it does not fall within the preset range, the irregular hole shape line is deemed unqualified.

[0006] The basic principles and beneficial effects of the scheme are as follows: This scheme utilizes the principle of laser triangulation to obtain an image of the irregular hole profile to be tested. A neural network module then directly identifies this image to quickly determine if the irregular hole profile is acceptable. If it is unacceptable, a further step is taken: contour enhancement is applied to the irregular hole profile image to extract its fingerprint. The fingerprint is then compared to calculate the similarity between the actual profile and the preset theoretical profile to determine acceptance again. In summary, compared to traditional measurement methods, this scheme is less restrictive, more convenient, and allows for a second assessment in cases of unacceptability, reducing the probability of misjudgment and improving measurement accuracy.

[0007] Furthermore, in step S1, the laser beam is magnified to form a laser line, and the laser line is projected onto the surface of the irregular hole to be tested, so that the diffuse reflection light of the laser line is projected onto the CCD photosensitive element for imaging, and processed to output the irregular hole profile image.

[0008] Furthermore, step S2 specifically includes: S201. Obtain the theoretical profile image and augmented data of irregular hole, and create an irregular hole profile image dataset; S202. Train an irregular aperture line detection model based on an irregular aperture line image dataset; S203. Input the irregular aperture line image output in step S1 into the trained irregular aperture line detection model for target recognition; S204. Determine whether the image confidence level of the irregular hole line image after target recognition is within the preset range. If it is, determine that the irregular hole line is qualified and complete the detection; otherwise, proceed to step S3.

[0009] Furthermore, step S3 specifically includes: S301. Extract the profile lines: Extract edge pixels using Canny edge detection; S302. Optimized Profile: Find the minimum energy term at each node, using the following formula:

[0010] In the formula: Pi refers to the position of the i-th node; the energy of a node is equal to the sum of the smoothness and the displacement term, as shown in the following formula: In the formula: For smoothness, the formula is as follows: In the formula: For smoothness, the formula is as follows:

[0011] In the formula, S∈r(i), r(i) is the curved region affected by Pi, and k(S) is the curvature at point S.

[0012] Furthermore, in step S4, the Gabor wavelet processing formula is as follows:

[0013] in, , Let kv be the direction of μ, and kv be the scale of kv,μ. Let Z(x,y) be the area of ​​the x and y axis, and kv = kmax / fv. f refers to the sampling step size in the frequency domain, which is taken as . μ,v represents different values ​​of the Gabor wavelet direction, and σ is the scaling factor of the Gaussian window.

[0014] Furthermore, in step S5, the fixed threshold is 127. If the grayscale pixel is greater than 127, it is defined as 1; if the grayscale pixel is less than 127, it is defined as 0.

[0015] Furthermore, step S6 specifically includes: S601. After Gaussian blurring the image of the irregular hole line after contour enhancement, delete the rows and columns until the image size becomes 32×32. S602. Starting from the first pixel in the upper left corner, traverse row by row towards the lower right corner. For each point with a gray value of 0, create a matrix with a length of 4 and a width of 4, with that point as the upper left corner. S603. After obtaining the above matrix, set the grayscale value of all pixels except the top left pixel to 1. S604. Then, starting from the first pixel in the upper left corner, the processed irregular aperture line image is traversed row by row towards the lower left corner. The points with a gray value of zero in the obtained image are fully connected and compared. The maximum of 63 distances obtained are the fingerprints of the irregular aperture line image.

[0016] Furthermore, step S7 specifically includes: The fingerprint of the irregular aperture profile image is fully mapped with the fingerprint of the pre-acquired theoretical profile image. Let Q be the quotient of each fingerprint ratio obtained by iterating through all fingerprint ratios.

[0017] Then, if the condition is satisfied once, N is satisfied and one is added; where h_to_be and w_to_be are the height and width of the line image to be tested, h_database and w_database are the height and width of the theoretical line image collected in advance, and N is satisfied as the number of quotients that satisfy the condition; Then perform similarity calculations, using the following formula:

[0018] Where N is the number of fingerprint ratios obtained from the above traversal.

[0019] Furthermore, in step S8, the preset interval is [0.95, 1].

[0020] The second objective of this invention is to provide a measurement system for irregularly shaped holes, including a laser generation module, an imaging module, and a processing module; The laser generating module is used to generate a laser beam and amplify it to form a laser line, which is then projected onto the surface of the irregular hole to be tested. The imaging module is used to image the diffuse reflected light from the surface of the irregular hole onto the photosensitive element, calculate and output the irregular hole profile image; the processing module is used to detect the irregular hole profile image of the small end of the engine connecting rod using steps S2-S8 of the above method. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating a method for measuring irregular hole profiles, as described in Example 1. Detailed Implementation

[0022] The following detailed description illustrates the specific implementation method: Example 1 This embodiment uses the irregular-shaped hole at the small end of the engine connecting rod as an example for illustration. Figure 1 As shown, this embodiment of a method for measuring irregular hole profile lines includes the following: S1. The laser beam is magnified to form a laser line, and the laser line is projected onto the surface of the irregular hole of the small end of the engine connecting rod to be tested. After being reflected by the surface of the irregular hole, the diffuse reflected light of the laser line passes through the optical system and is projected onto the CCD photosensitive element, so that the diffuse reflected light of the laser line is imaged on the photosensitive element. The distance from the photosensitive element to the surface of the irregular hole at the small end of the engine connecting rod, as well as the position information along the laser line, are then calculated. Finally, the above data are processed and converted to output an image of the irregular hole profile at the small end of the engine connecting rod.

[0023] S2. The image recognition method of the small end irregular hole profile of the engine connecting rod is used to identify the profile. In this embodiment, the structure of the convolutional neural network includes one input layer, M convolutional layers (Conv) and ReLU activation function layers, b pooling layers, k fully connected layers, and one output layer. M ranges from 2 to 5; N ranges from 1 to 100; b is 0 or 1; and k ranges from 0 to 2. The specific recognition includes: S201. First, obtain the theoretical profile image and augmented data of the irregular hole at the small end of the engine connecting rod, and create a dataset of irregular hole profile images at the small end of the engine connecting rod.

[0024] S202. Using TensorFlow as the deep learning framework, train an irregular hole line detection model based on the R_FCN_ResNet algorithm.

[0025] The R-FCN algorithm optimizes the incomplete weighting and insufficient position sensitivity issues based on Fully Convolutional Network (FCN) and Faster R-CNN. R-FCN uses FCN to extract target object features to generate feature maps. These feature maps are then convolved to generate independent Region Proposal Networks (RPNs) and position-sensitive score maps. The RPN generates Regions of Interest (ROIs), and the position-sensitive score network compares the feature maps generated by FCN with the ROIs output by the RPN to produce a score map and bounding box for the target region. Non-Maximum Suppression (NMS) is used to vote on the combined score of the specific class probability and position regression, thereby locating and classifying the target object. This embodiment combines the CNN-based feature extraction network ResNet with a CNN-based Region-based object detection algorithm to construct a deep learning model based on R-FCN_ResNet for detecting irregularly shaped holes in the small end of an engine connecting rod. The structure and principle of its R-FCN_ResNet algorithm are as follows: Conv4 of the fully convolutional feature extraction network (ResNet) is input into the region proposal network (RPN) to obtain regions of interest (RoIs) and region proposal boxes at different scales. Then, the RoIs and Conv5 of the ResNet network are input into the position-sensitive score map. Convolution and pooling are then used to obtain the score of a certain type of feature. Finally, a voting mechanism is used to provide classification information and localization regression information for the image of the irregular hole profile of the small end of the engine connecting rod.

[0026] S203. Input the image of the irregular hole profile of the small end of the connecting rod of the engine to be detected, which was output in step S1, into the exported irregular hole profile detection model based on the R-FCN_ResNet algorithm for target recognition.

[0027] If the image of the irregular hole profile of the small end of the engine connecting rod is viewed after identification, and its confidence level is within the preset range (in this embodiment, it is [0.99, 1]), then the irregular hole profile of the small end of the engine connecting rod is qualified. If its confidence level is within [0, 0.99], then it is re-verified through subsequent steps.

[0028] S3. The Johannes algorithm is used to magnify the profile lines in the image of the irregular bore at the small end of the engine connecting rod. The specific steps are as follows: S301. Extract the profile lines: Extract edge pixels using Canny edge detection; S302. Optimize the profile: Improve the smoothness of the profile by optimizing the control point positions, i.e., find the minimum energy term at each node, as shown in the following formula: (1) In the formula: Pi refers to the position of the i-th node.

[0029] The energy of a node is equal to the sum of its smoothness and displacement terms, as shown in the following formula: (2) In the formula: For smoothness, the formula is as follows: (3) In the formula, S∈r(i), r(i) is the curve region affected by Pi, and k(S) is the curvature at point S. The curve is sampled to calculate the integral. This is used to detect sharp features in the curve and exclude the surrounding areas of these features.

[0030] S4. Apply Gabor wavelet processing to the optimized and enlarged profile to extract the contour. The Gabor filter is a Gaussian kernel function that can be used to enhance image features such as peaks, valleys, ridges, and edges. The two-dimensional Gabor wavelet function expression is as follows: (4) in, , Let kv be the direction of μ, and kv be the scale of kv,μ. Let Z(x,y) be the area of ​​the x and y axis, and kv = kmax / fv. f refers to the sampling step size in the frequency domain, which is taken as . μ,v represents different values ​​of the Gabor wavelet direction, and σ is the scaling factor of the Gaussian window.

[0031] S5. After Gabor wavelet processing, the profile image is enhanced using a fixed threshold binarization algorithm. The fixed threshold is 127; if the grayscale pixel value is greater than 127, it is set to 1, and if the grayscale pixel value is less than 127, it is set to 0.

[0032] S6. The fingerprint of the profile image of the engine connecting rod small end irregular hole, which was enhanced in step S5, is obtained by processing the profile image using the surviving point extraction method. The specific steps are as follows: S601. After Gaussian blurring the image of the irregular hole profile of the small end of the engine connecting rod, delete the rows and columns until the image size becomes 32×32; that is, after Gaussian blurring the image of the irregular hole profile after contour enhancement, delete the rows and columns until the image size becomes 32×32.

[0033] S602. Starting from the first pixel in the upper left corner, traverse row by row towards the lower right corner. For each point with a gray value of 0, create a matrix with a length of w / 8 and a width of h / 8, with that point as the upper left corner. w and h are the width and height of the contour-enhanced image, respectively; that is, starting from the first pixel in the upper left corner, traverse row by row towards the lower right corner. For each point with a gray value of 0, create a matrix with a length of 4 and a width of 4, with that point as the upper left corner.

[0034] S603. After obtaining the above matrix, set the grayscale value of all pixels except the top left pixel to 1.

[0035] S604. Next, starting from the first pixel in the upper left corner, the processed engine connecting rod small end irregular hole profile image is traversed row by row towards the lower left corner. The points with a gray value of zero in the obtained image are compared by performing a fully connected mapping. The maximum of 63 distances obtained are the fingerprints of the engine connecting rod small end irregular hole profile image.

[0036] S7. Calculate the similarity between the actual profile of the irregular bore at the small end of the engine connecting rod and the pre-collected theoretical profile through fingerprint comparison. The specific steps are as follows: The fingerprint of the engine connecting rod small end profile image to be matched after the above steps is fully associatively mapped and compared with the fingerprint of the pre-acquired theoretical profile image. All fingerprint ratios are iterated through, and for each fingerprint, the quotient is calculated with all subsequent ratios. Let the quotient be Q. If: (6) Then, if the condition is satisfied once, N is incremented by one, where hdestination and wdestination are the height and width of the model image to be tested, respectively; hdatabase and wdatabase are the height and width of the theoretical model image obtained beforehand, respectively; and N is the number of quotients that satisfy the condition. Then, similarity calculations are performed. The formula is as follows: (7) Where N is the number of fingerprint ratios obtained from the above traversal.

[0037] S8. The similarity value ranges from [0, 1]. The similarity is defined as 1 in the preset interval. In this embodiment, the preset interval is [0.95, 1]. The similarity is 0 in the range of [0, 0.95). If the calculated output result is 0, an alarm is triggered; if the output result is 1, no alarm is triggered, that is, the profile of the small end of the engine connecting rod is qualified.

[0038] Based on a method for measuring the line shape of irregular holes, this embodiment also provides a system for measuring the line shape of irregular holes, including a laser generation module, an imaging module, and a processing module.

[0039] The laser generating module is used to generate a laser beam and amplify it to form a laser line, which is then projected onto the surface of the irregular hole at the small end of the engine connecting rod to be tested. In this embodiment, the laser generating module uses a laser.

[0040] The imaging module is used to image diffuse reflected light from the surface of the irregular hole onto the photosensitive element, calculate and output the image of the irregular hole profile of the small end of the engine connecting rod; in this embodiment, the imaging module uses an image sensor with a CCD photosensitive element.

[0041] The processing module is used to detect the profile image of the irregular hole at the small end of the engine connecting rod using the methods in steps S2-S8. In this embodiment, the processing module uses a computer.

[0042] Example 2 The difference between this embodiment and Embodiment 1 is that, in the method of this embodiment, the irregular hole profile detection model in step S203 is also used to mark the positions of the irregular hole profile images of the small end of the engine connecting rod where the profile is unqualified.

[0043] In step S4, a dataset of locations where the profile is defective is also established. Based on the dataset, the parts that are prone to shape parameter defects are statistically analyzed, and the magnified profile is divided into several measurement segments. Among them, the parts that are prone to shape parameter defects have corresponding measurement segments. There are overlapping parts between adjacent measurement segments. In this embodiment, when statistically analyzing the parts that are prone to shape parameter defects, the locations of the profile defects are superimposed on the theoretical profiles that have been collected in advance. When the number of superimposed parts exceeds a first threshold, the corresponding part is regarded as the part that is prone to shape parameter defects.

[0044] Determine if there is a corresponding measurement segment at the location of the current defective profile. If there is no corresponding measurement segment, perform Gabor wavelet processing on the entire optimized and enlarged profile to extract the contour. If there is a corresponding measurement segment, perform Gabor wavelet processing on the measurement segment to extract the contour. In step S7, it is determined whether the fingerprint extracted is a complete profile image or a measurement segment. If it is a complete profile image, the similarity between the actual profile of the engine connecting rod small end irregular hole and the theoretical profile collected in advance is calculated by fingerprint comparison. If it is a measurement segment, the similarity between the profile of the measurement segment of the engine connecting rod small end irregular hole and the corresponding measurement segment of the theoretical profile collected in advance is calculated by fingerprint comparison.

[0045] In step S8, if the similarity between the measured section profile of the small end irregular hole of the engine connecting rod and the corresponding measured section of the pre-acquired theoretical profile is within [0, 0.95], the engine connecting rod small end irregular hole profile is deemed unqualified. If the similarity is within [0.95, 1], proceed to step S4 to perform Gabor wavelet processing on the remaining measured sections and extract the contour. If the similarity between the remaining measured section profile and the corresponding remaining measured section of the pre-acquired theoretical profile is within [0, 0.95], the engine connecting rod small end irregular hole profile is deemed unqualified. If the similarity is within [0.95, 1], the engine connecting rod small end irregular hole profile is deemed qualified.

[0046] The solution in this embodiment can reduce the amount of calculation and increase the measurement speed while ensuring accuracy.

[0047] Example 3 The difference between this embodiment and embodiment two is that in the method of this embodiment, in step S8, the number of times the divided measurement segments are judged as qualified according to the measurement segment profile and the number of times the remaining measurement segments are judged as unqualified is also counted. If the number of times exceeds the second threshold, the corresponding measurement segment is marked as an abnormal measurement segment.

[0048] In step S4, it is also determined whether there is a corresponding measurement segment at the current defective position of the profile. If there is a corresponding measurement segment, it is determined whether the corresponding measurement segment is an abnormal measurement segment. If so, Gabor wavelet processing is performed on the entire optimized and enlarged profile to extract the contour; otherwise, Gabor wavelet processing is performed on the measurement segment to extract the contour.

[0049] The solution in this embodiment can further improve the measurement speed.

[0050] The above are merely embodiments of the present invention. The invention is not limited to the fields covered by these embodiments. Commonly known structures and characteristics in the solutions are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are able to access all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, under the guidance of this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A method for measuring irregular hole profiles, characterized in that, Includes the following: S1. Obtain the irregular aperture profile image using the principle of laser triangular reflection; S2. Input the irregular hole line image into the pre-trained irregular hole line detection model. Determine whether the irregular hole line is qualified based on the recognition result of the irregular hole line detection model. If qualified, the detection is completed; if not qualified, proceed to step S3. S3. Magnify the profile lines in the irregular aperture profile image using the Johannes algorithm; specifically including: S301. Extract the profile lines: Extract edge pixels using Canny edge detection; S302. Optimized Profile: Find the minimum energy term at each node, using the following formula: In the formula: P i The position of the i-th node is indicated; the energy of a node is equal to the sum of its smoothness and displacement terms, as shown in the following formula: In the formula: For smoothness, the formula is as follows: In the formula, , Is affected by P i The affected curve region It is the curvature at point S; S4. Perform Gabor wavelet processing on the magnified profile to extract the contour; S5. Use a fixed threshold binarization algorithm to enhance the contour of the irregular hole line image; S6. Process the contour-enhanced irregular aperture line image using the surviving point extraction method to obtain the fingerprint of the irregular aperture line image; specifically including: S601. After Gaussian blurring the image of the irregular hole line after contour enhancement, delete the rows and columns until the image size becomes 32×32. S602. Starting from the first pixel in the upper left corner, traverse row by row towards the lower right corner. For each point with a gray value of 0, create a matrix with a length of 4 and a width of 4, with that point as the upper left corner. S603. After obtaining the above matrix, set the grayscale value of all pixels except the top left pixel to 1. S604. Next, starting from the first pixel in the upper left corner of the processed irregular aperture line image, traverse row by row towards the lower left corner. Perform fully connected mapping between points with a grayscale value of zero in the obtained image and compare them. The maximum of 63 distances obtained are the fingerprints of the irregular aperture line image. S7. Calculate the similarity between the actual profile of the irregular hole and the preset theoretical profile by comparing fingerprints; S8. Determine whether the similarity falls within the preset range. If it falls within the preset range, the irregular hole shape line is deemed qualified. If it does not fall within the preset range, the irregular hole shape line is deemed unqualified.

2. The method for measuring irregular hole profiles according to claim 1, characterized in that: In step S1, the laser beam is magnified to form a laser line, and the laser line is projected onto the surface of the irregular hole to be tested, so that the diffuse reflection light of the laser line is projected onto the CCD photosensitive element for imaging, and the image of the irregular hole shape line is processed and output.

3. The method for measuring irregular hole profiles according to claim 1, characterized in that: Step S2 specifically includes: S201. Obtain the theoretical profile image and augmented data of irregular hole, and create an irregular hole profile image dataset; S202. Train an irregular aperture line detection model based on an irregular aperture line image dataset; S203. Input the irregular aperture line image output in step S1 into the trained irregular aperture line detection model for target recognition; S204. Determine whether the image confidence level of the irregular hole line image after target recognition is within the preset range. If it is, determine that the irregular hole line is qualified and complete the detection; otherwise, proceed to step S3.

4. The method for measuring irregular hole profiles according to claim 1, characterized in that: In step S4, the Gabor wavelet processing formula is as follows: in, , for direction, for The scale, , , f refers to the sampling step size in the frequency domain, which is taken as . μ and v represent different values ​​for the Gabor wavelet direction, and σ is the scaling factor of the Gaussian window.

5. The method for measuring irregular hole profiles according to claim 1, characterized in that: In step S5, the fixed threshold is 127. If the grayscale pixel is greater than 127, it is defined as 1; if the grayscale pixel is less than 127, it is defined as 0.

6. The method for measuring irregular hole profiles according to claim 1, characterized in that: Step S7 specifically includes: The fingerprint of the irregular aperture profile image is fully coupled and compared with the fingerprint of the pre-acquired theoretical profile image. All fingerprint ratios are iterated through, and for each fingerprint, the quotient is calculated with all subsequent ratios. Let the quotient be Q. If: Then the condition is satisfied once. Add one; among them and These are the height and width of the profile image to be tested, respectively. and These represent the height and width of the pre-acquired theoretical profile image, respectively. The number of quotients that satisfy the conditions; Then perform similarity calculations, using the following formula: in, This represents the number of all fingerprint ratios obtained through the above iteration.

7. The method for measuring irregular hole profiles according to claim 6, characterized in that: In step S8, the preset interval is: .

8. A measurement system for irregularly shaped holes, characterized in that, It includes a laser generation module, an imaging module, and a processing module; The laser generating module is used to generate a laser beam and amplify it to form a laser line, which is then projected onto the surface of the irregular hole to be tested. The imaging module is used to image diffusely reflected light from the surface of the irregular aperture onto the photosensitive element, calculate and output the irregular aperture profile image; The processing module is used to detect the profile image of the small end bore of the engine connecting rod using steps S2-S8 of any one of the methods of claims 1-7.

Images