A roundness intelligent detection method and system for automobile pipes

Abstract

The present application relates to the field of automobile pipe fittings, and more particularly, to a roundness intelligent detection method and system for automobile pipe fittings.The method comprises the following steps: obtaining the positions of preset key points of the automobile pipe fitting, obtaining the distances between each key point and the camera position respectively to calculate the perpendicularity of the pipe fitting, and taking the image corresponding to the maximum value of the perpendicularity of the pipe fitting as the pipe fitting cross-sectional view; inputting the pipe fitting cross-sectional view into a preset extraction network to obtain a pipe fitting region map, obtaining the lower edge and the upper edge, obtaining the pixel point set between the upper edge and the lower edge as a first set, obtaining the pixel point set between the inner edge and the outer edge in the pipe fitting region map as a second set, and calculating the pipe fitting roundness evaluation according to the first set, the second set and the pipe fitting perpendicularity to complete the roundness detection.The technical scheme of the present application can improve the accuracy of the pipe fitting roundness detection result.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of automobile pipe fittings. More particularly, the present application relates to a roundness intelligent detection method and system for automobile pipe fittings. BACKGROUND

[0002] In modern automobile manufacturing industry, pipe fittings, as an important part of automobile manufacturing, are widely used in automobile braking, fuel, cooling, air conditioning and other systems. The performance of pipe fittings directly affects the safety, reliability and working performance of automobiles. Therefore, the quality control of automobile pipe fittings is of great significance, especially in the manufacturing process of pipe fittings, the detection of roundness is one of the key factors to determine the quality of pipe fittings.

[0003] Roundness refers to the degree to which the shape of an object approaches a perfect circle. For pipe fittings, the accuracy of roundness directly affects the installation effect and working performance of pipe fittings in automobile systems. For example, in the fuel pipeline, if the roundness of the pipe fitting does not meet the requirements, it may cause poor sealing, excessive vibration, and even leakage; in the braking system, the deviation of roundness may affect the braking efficiency, and in severe cases, it may cause brake failure, endangering driving safety. Therefore, ensuring that the roundness of pipe fittings meets the standards is a key step to ensure the quality and safety of automobiles.

[0004] However, in actual production, the cross section of the pipe fitting may be uneven. This irregularity in geometry can cause deviations in the detection process, making it difficult to accurately measure and analyze the pipe fitting. In addition, the uneven cross section of the pipe fitting can cause uneven contact between the lens and the pipe fitting, resulting in inaccurate roundness detection results. SUMMARY

[0005] To solve the problem of inaccurate roundness detection results, the present application provides solutions in the following aspects.

[0006] In a first aspect, the present application discloses a roundness intelligent detection method for automobile pipe fittings, comprising: obtaining the positions of the preset key points of the automobile pipe fitting, obtaining the distances between each key point and the camera position respectively to calculate the perpendicularity of the pipe fitting, and taking the shooting image corresponding to the maximum value of the perpendicularity of the pipe fitting as the pipe fitting cross section image; inputting the pipe fitting cross section image into a preset extraction network to obtain a pipe fitting region image, taking the incircle edge of the inner edge in the pipe fitting region image as the lower edge, taking the excircle edge of the outer edge in the pipe fitting region image as the upper edge, obtaining the pixel point set between the upper edge and the lower edge as a first set, obtaining the pixel point set between the inner edge and the outer edge in the pipe fitting region image as a second set, calculating the pipe fitting roundness evaluation according to the first set, the second set and the pipe fitting perpendicularity, and completing the roundness detection.

[0007] By obtaining the distance between each preset key point and the camera position and calculating the perpendicularity of the pipe, it can be ensured that the selected image best reflects the true geometric shape of the pipe, reducing errors caused by inclination or position deviation. After obtaining the pipe cross-section image, by extracting the key points of the inner and outer edges and defining the upper and lower edges respectively, the actual contour of the pipe can be accurately defined, thereby obtaining more accurate area information.

[0008] Preferably, the positions of the preset key points include: the preset key points at 0 degrees, 90 degrees, 180 degrees and 270 degrees on the automobile pipe, and the midpoints between the inner edge and the outer edge.

[0009] Preferably, the perpendicularity of the pipe satisfies the relationship:

[0010] , represents the perpendicularity of the pipe, represents the distance between the key point on the left side in the horizontal direction and the camera position, represents the distance between the key point on the right side in the horizontal direction and the camera position, represents the distance between the key point on the upper side in the vertical direction and the camera position, represents the distance between the key point on the lower side in the vertical direction and the camera position, represents the standard ratio of the key point in the horizontal direction, represents the standard ratio of the key point in the vertical direction, represents the horizontal direction, represents the vertical direction.

[0011] Considering the position difference of the key points in the horizontal and vertical directions, by comparing the deviation between the actual key points and the standard ratio, the inclination degree of the pipe can be effectively judged. Through this calculation, it can be ensured that the control of the perpendicularity is more precise during the production and processing of the pipe, thereby improving the quality consistency and precision of the pipe.

[0012] Preferably, the step of taking the photographed image corresponding to the maximum value of the pipe perpendicularity as the pipe cross-section image comprises: changing the inclination degree of the automobile pipe to obtain a plurality of pipe perpendicularities to obtain the maximum value of the key perpendicularity; each calculated key perpendicularity corresponds to a photographed image, and the photographed image corresponding to the maximum value of the pipe perpendicularity is taken as the pipe cross-section image.

[0013] By accurately capturing the state of the pipe at different angles, shape deviations caused by improper shooting angles or other factors are avoided, so that the final cross-section image has higher representativeness and accuracy.

[0014] Preferably, the preset extraction network is an FCN network or a U-Net network.

[0015] Preferably, the training of the preset extraction network comprises: constructing a training set by taking the pipe section images corresponding to the several automobile pipes obtained in the history as input and taking the true values of the pipe region images of the automobile pipes as labels for training the preset extraction network, wherein the loss function of the preset extraction network uses cross-entropy loss.

[0016] By fully utilizing the real pipe data and labels, the network can learn the accurate features of the pipe region, and the accuracy of pipe detection is effectively improved.

[0017] Preferably, the training of the preset extraction network further comprises: for the same automobile pipe, obtaining several pipe perpendicularities by changing the inclination of the automobile pipe, setting a perpendicularity threshold, and taking the shooting images corresponding to the pipe perpendicularities greater than the perpendicularity threshold as a set of section images of the automobile pipe; constructing a training set by taking the set of section images of the automobile pipe in the history as input and taking the true values of the pipe region set in the set of section images as labels for training the preset extraction network, wherein the loss function of the preset extraction network uses cross-entropy loss.

[0018] The network can learn different features of the pipe under multiple angles and perpendicularities, and ensure that the network can extract accurate pipe region information from image samples. By using these diversified images and their true labels as a training set and combining a cross-entropy loss function for optimization, the network can more accurately identify the pipe region, thereby improving the accuracy and robustness of section image extraction.

[0019] Preferably, the pipe roundness evaluation satisfies the relationship:

[0020] , represents the pipe roundness evaluation, represents the pipe perpendicularity, represents the first set, represents the second set.

[0021] Preferably, the completion of the roundness detection comprises: in response to the pipe roundness evaluation being not less than a preset roundness threshold, the roundness of the automobile pipe meets the requirements; and in response to the pipe roundness evaluation being less than the preset roundness threshold, generating and sending an alarm signal.

[0022] In a second aspect, the present application discloses a roundness intelligent detection system for automobile pipes, comprising: a processor and a memory, the memory stores computer program instructions, when the computer program instructions are executed by the processor, any one of the roundness intelligent detection methods for automobile pipes is realized.

[0023] The beneficial effects of the present application are:

[0024] The application calculates the perpendicularity of the pipe by obtaining the positions of the key points of the pipe and calculating the perpendicularity, automatically extracts the region map of the pipe by combining the image extraction network, and then accurately identifies the inner and outer edges of the pipe to calculate the roundness of the pipe. The geometry of the automobile pipe can be comprehensively and automatically detected, avoiding the limitations of manual detection and improving the detection efficiency and accuracy. In particular, by taking the photographed image corresponding to the maximum value of the perpendicularity of the pipe as the pipe cross-section image for processing, the error caused by the change in the angle of the pipe can be effectively excluded, further improving the reliability of the roundness detection.

[0025] In addition, the application sets a roundness threshold and an alarm mechanism to issue an alarm in time when detecting unqualified pipes, ensuring the stability and consistency of the pipe quality, and helping to accurately control the quality and optimize the process in automobile manufacturing. BRIEF DESCRIPTION OF DRAWINGS

[0026] Figure 1 is a flowchart of an intelligent roundness detection method for automobile pipes according to an embodiment of the application. DETAILED DESCRIPTION

[0027] The technical solutions in the embodiments of the application will be described clearly and completely below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are part of the embodiments of the application, rather than all the embodiments of the application.

[0028] The specific embodiments of the application will be described in detail below with reference to the drawings.

[0029] Reference Figure 1 An intelligent roundness detection method for automobile pipes includes steps S1-S2, which will be described in detail below.

[0030] S1: Obtain the positions of the preset key points of the automobile pipe, and obtain the distance between each key point and the camera position to calculate the perpendicularity of the pipe, and take the photographed image corresponding to the maximum value of the perpendicularity of the pipe as the pipe cross-section image.

[0031] It should be noted that the degree of perpendicularity of the automobile pipe to the shooting lens directly affects the accuracy of the subsequent roundness detection. When the pipe wall is not perpendicular to the shooting lens, the pipe cross-section in the collected image will be distorted in shape. If the original pipe cross-section is circular, it will be deformed into an ellipse due to the deviation of the angle. Since the roundness detection usually relies on the cross-sectional shape in the image to calculate the difference in roundness, the elliptical error in the image will directly cause the deviation of the roundness calculation result.

[0032] In one embodiment, four key points are preset on the cross-section of the pipe, located at 0 degrees, 90 degrees, 180 degrees and 270 degrees on the cross-section of the pipe. The positions of these key points are set at the midpoint between the inner edge and the outer edge of the pipe, i.e. the midpoint of the radius from the inside to the outside of the pipe cross-section. Through this setting, comprehensive measurement of the pipe in different angular directions can be ensured, ensuring accurate evaluation of the roundness.

[0033] The distance between each key point and the camera position is obtained by laser ranging, and the pipe perpendicularity is calculated. The pipe perpendicularity satisfies the relationship:

[0034] , represents the pipe perpendicularity, represents the distance between the left key point in the horizontal direction and the camera position, represents the distance between the right key point in the horizontal direction and the camera position, represents the distance between the upper key point in the vertical direction and the camera position, represents the distance between the lower key point in the vertical direction and the camera position, represents the standard ratio of the horizontal direction key points, represents the standard ratio of the vertical direction key points, represents the horizontal direction, represents the vertical direction.

[0035] It needs to be explained that the value range of the pipe perpendicularity is between negative infinity and 1. The closer the value is to 1, the more vertical the pipe cross-section is, and the closer the photographed pipe cross-section image is to the actual situation. When the value is negative, it means that the pipe is too large.

[0036] The distances between the key points of the pipe and the camera position are obtained by laser ranging technology, and the perpendicularity of the pipe is calculated based on these distances, providing an efficient and accurate method for quality detection of the pipe. This method can directly quantify the perpendicularity deviation of the pipe, and by comparing the difference between the distance ratio of the horizontal and vertical direction key points and the standard ratio, the perpendicularity state of the pipe in space can be intuitively reflected.

[0037] By changing the inclination of the automobile pipe, a number of pipe perpendicularities are obtained to obtain the maximum value of the key perpendicularity. Each calculated key perpendicularity corresponds to a photographed image, and the photographed image corresponding to the maximum value of the pipe perpendicularity is taken as the pipe cross-section image.

[0038] S2: input the pipe section image into a preset extraction network to obtain a pipe region image, take an incircle edge of an inner edge in the pipe region image as a lower edge, take an excircle edge of an outer edge in the pipe region image as an upper edge, obtain a pixel point set between the upper edge and the lower edge as a first set, obtain a pixel point set between the inner edge and the outer edge in the pipe region image as a second set, and calculate a pipe roundness evaluation according to the first set, the second set and a pipe perpendicularity, so as to complete the roundness detection.

[0039] It should be noted that in the actual detection process, the pipe perpendicularity is often not equal to 1 directly, because in the processing site, the precision of the clamp is limited. Although modern production equipment and clamp systems can provide high precision, due to the possibility of small errors in the processing process, the clamp is difficult to completely reach the ideal vertical state when fixing the pipe.

[0040] In one embodiment, the pipe section image is input into a preset extraction network to obtain a pipe region image, wherein the preset extraction network is an FCN network or a U-Net network. A training set is constructed by taking a plurality of pipe section images corresponding to automobile pipes obtained in the history as input and taking the true value of the pipe region image of the automobile pipe as a label, which is used to train the preset extraction network, wherein the loss function of the preset extraction network uses cross-entropy loss.

[0041] The network can automatically learn the features in the pipe section image and accurately identify the pipe region, thereby significantly improving the efficiency and accuracy of pipe detection. By using the cross-entropy loss function, the network can better optimize the classification performance, so that the detection result is closer to the true value, and the possibility of misjudgment and omission is reduced.

[0042] The excircle edge of the outer edge in the pipe region image is taken as the upper edge, the pixel point set between the upper edge and the lower edge is taken as the first set, and the pixel point set between the inner edge and the outer edge in the pipe region image is taken as the second set. It should be explained that the outer edge and the inner edge of the pipe are inconsistent, that is, the outer edge is a square and the inner edge is a circle. At this time, due to the existence of the inclination angle of the pipe, the outer edge in the pipe section image is a rectangle and the inner edge is an ellipse, so the incircle edge of the inner edge, that is, the lower edge, and the excircle edge of the outer edge, that is, the upper edge, need to be obtained.

[0043] The pipe roundness evaluation is calculated, and the pipe roundness evaluation satisfies the relationship:

[0044] , represents the pipe roundness evaluation, represents the pipe perpendicularity, represents the first set, represents the second set.

[0045] In response to the pipe roundness evaluation being not less than the preset roundness threshold, the roundness of the automobile pipe meets the requirements; and in response to the pipe roundness evaluation being less than the preset roundness threshold, an alarm signal is generated and sent.

[0046] In another embodiment, the training of the preset extraction network further comprises: for the same automobile pipe, by changing the inclination degree of the automobile pipe, obtaining a plurality of pipe perpendicularities, setting a perpendicularity threshold, and taking the photographed image corresponding to the pipe perpendicularity greater than the perpendicularity threshold as the cross-sectional image set of the automobile pipe; and constructing a training set by taking the cross-sectional image set of the automobile pipe in the history as input and the real value of the pipe region set in the cross-sectional image set as label, for training the preset extraction network, wherein the loss function of the preset extraction network uses cross-entropy loss.

[0047] By training using images of the same pipe taken at multiple angles close to vertical, the extraction network can more comprehensively learn the geometric features of the pipe and the subtle changes under different angles. Since the projection and morphology of the pipe under different angles may be different, a single angle image may not be able to fully capture all the features of the pipe. Therefore, by training with multi-angle images, the network can extract more rich and accurate pipe region information under different perspectives.

[0048] The system comprises a processor and a memory, and the memory stores computer program instructions which, when executed by the processor, implement the roundness intelligent detection method for automobile pipes according to the first aspect of the application.

[0049] The system further comprises a communication bus and a communication interface and other components well known to those skilled in the art, the settings and functions of which are known in the art, and thus will not be described here.

[0050] It should be noted that, for those skilled in the art, without departing from the concept of the present application, a number of modifications and improvements can be made, which are all within the scope of the present application. Therefore, the protection scope of the present application patent should be subject to the appended claims.

Claims

1. An intelligent roundness detection method for automobile pipe fittings, characterized in that: include: Get the position of the preset key points of the automobile pipe fittings, and get the distance between each key point and the camera position to calculate the verticality of the pipe fittings, which satisfies the relationship: , Indicates the verticality of the pipe fittings. Indicates the distance between the left key point and the camera position in the horizontal direction. Indicates the distance between the right key point and the camera position in the horizontal direction. Indicates the distance between the upper key point and the camera position in the vertical direction. Indicates the distance between the vertical lower key point and the camera position. Indicates the standard ratio of key points in the horizontal direction, Indicates the standard ratio of key points in the vertical direction, Indicates the horizontal direction, Indicates vertical direction; The captured image corresponding to the maximum value of the verticality of the pipe fitting is used as a cross-sectional view of the pipe fitting, including: obtaining a plurality of verticalities of the pipe fitting by changing the tilt degree of the automobile pipe fitting to obtain the maximum value of the verticality of the pipe fitting; Each calculated pipe verticality corresponds to a captured image. The closer the pipe verticality value is to 1, the more vertical the pipe cross section is. The captured image corresponding to the maximum value of the pipe verticality is used as the pipe cross section. The pipe cross-section diagram is input into the preset extraction network to obtain the pipe area diagram. Due to the inconsistency between the outer edge and the inner edge of the pipe, the inscribed edge of the inner edge in the pipe area diagram is used as the lower edge, and the circumscribed edge of the outer edge in the pipe area diagram is used as the upper edge. The set of pixel points between the upper and lower edges is obtained as the first set, and the set of pixel points between the inner and outer edges in the pipe area diagram is obtained as the second set. The roundness evaluation of the pipe is calculated based on the first set, the second set and the verticality of the pipe to complete the roundness detection.

2. The intelligent roundness detection method for automobile pipes according to claim 1, characterized in that: The positions of the preset key points include: The preset key points are at 0 degrees, 90 degrees, 180 degrees and 270 degrees on the automotive tube, and are located at the midpoint between the inner edge and the outer edge.

3. The intelligent roundness detection method for automobile pipes according to claim 1, characterized in that: The preset extraction network is an FCN network or a U-Net network.

4. The intelligent roundness detection method for automobile pipes according to claim 1, characterized in that: The training of the preset extraction network includes: A training set is constructed by taking the cross-sectional images of several automobile pipes obtained in history as input and the real values ​​of the pipe area images of automobile pipes as labels for training the preset extraction network. The loss function of the preset extraction network uses the cross-entropy loss.

5. The intelligent roundness detection method for automobile pipes according to claim 1, characterized in that: The training of the preset extraction network further includes: For the same automobile pipe fitting, the verticality of several pipe fittings is obtained by changing the tilt degree of the pipe fitting, and a verticality threshold is set. The captured images corresponding to the verticality of the pipe fitting greater than the verticality threshold are used as the cross-sectional atlas of the automobile pipe fitting; A training set is constructed by taking the cross-sectional atlas of historical automobile pipe fittings as input and the true values ​​of the pipe fitting area set in the cross-sectional atlas as labels for training the preset extraction network, wherein the loss function of the preset extraction network uses the cross entropy loss.

6. The intelligent roundness detection method for automobile pipes according to claim 1, characterized in that: The roundness evaluation of the pipe fitting satisfies the relationship: , Indicates the roundness evaluation of pipe fittings, Indicates the verticality of the pipe fittings. represents the first set, Represents the second set.

7. The intelligent roundness detection method for automobile pipes according to claim 1, characterized in that: The roundness detection is completed including: In response to the pipe fitting roundness evaluation being not less than a preset roundness threshold, the roundness of the automobile pipe fitting meets the requirement; In response to the roundness evaluation of the pipe being less than a preset roundness threshold, an alarm signal is generated and sent.

8. An intelligent roundness detection system for automobile pipe fittings, characterized in that: include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, an intelligent roundness detection method for automobile pipes according to any one of claims 1 to 7 is implemented.

Images