Bearing end face defect detection method based on defect identification

Abstract

The invention relates to the technical field of flaw recognition, in particular to a bearing end face defect detection method based on flaw recognition, which comprises the following steps: acquiring a gray level image corresponding to a surface image of a bearing end face by using an optical means; converting the bearing pixel points in the target area into a Hough circle space to obtain a connected domain corresponding to the bearing ring; the method comprises the following steps: converting bearing pixel points into a three-dimensional Hough space to form a three-dimensional curve; obtaining the standard precision difference between the bearing rings and a standard circle corresponding to the circle center, taking the standard precision difference corresponding to any bearing ring as a reference plane, obtaining the distance between the circle center determined by every two adjacent bearing pixel points and the reference plane, and further obtaining the deviation degree of each bearing ring; and judging whether the end face of the bearing is abnormal or not by taking the product of the deviation degrees of the four bearing rings as the overall abnormal degree of the bearing. Fine defects on the edge of the bearing can be detected by identifying the defects through an optical means, and the analysis result is more accurate.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of defect recognition, and particularly relates to a bearing end face defect detection method based on defect recognition. BACKGROUND

[0002] The main function of a bearing is to support the rotating body, reduce the friction coefficient in the movement process, and ensure the rotation accuracy. In many rotating scenes, the quality of the bearing is highly required, and the quality of the bearing to some extent represents the efficiency and speed of the entire transmission system, so the quality of the bearing needs to be strictly controlled.

[0003] At present, the quality inspection of the bearing is relatively complex. First, the machine is used to rotate at high speed to judge whether there is noise. The machine can only detect one by one, otherwise the rotation sound of the abnormal bearing will affect the rotation sound of the normal bearing. Moreover, when the abnormality is detected, further disassembly and processing are needed, and the detection process is very complicated. Another method is manual detection. For any bearing, one hand holds the outer frame of the bearing, and the other hand or finger holds the inner frame of the bearing. Both hands are squeezed towards the middle, and then the friction and rotation of the bearing are felt to determine whether there is a defect or other abnormality. However, this method is too dependent on the feeling of the detection personnel, and is easy to misjudge the quality of the bearing. SUMMARY

[0004] In order to solve the problem of inaccurate bearing quality detection, the present application provides a bearing end face defect detection method based on defect recognition, and the technical scheme is as follows: An embodiment of the present application provides a bearing end face defect detection method based on defect recognition, which comprises the following steps: Collecting the surface image of the bearing end face and obtaining the corresponding gray image, segmenting the bearing area in the gray image as a target area, and the target area includes four bearing rings; Identifying the gap area in the target area and removing it to obtain the bearing pixel points in the target area; converting the bearing pixel points into a Hough circle space to obtain the corresponding connected domain of each bearing ring; converting the bearing pixel points to a three-dimensional Hough space to form a three-dimensional curve; Obtaining the standard precision difference between each bearing ring and the corresponding standard circle, taking the standard precision difference corresponding to any one bearing ring as a reference plane, obtaining the distance between the center of the circle determined by each two adjacent bearing pixel points and the reference plane, and then obtaining the deviation degree of each bearing ring; Taking the product of the deviation degrees of the four bearing rings as the overall abnormality degree of the bearing, when the overall abnormality degree is greater than a preset threshold, the bearing end face has an abnormality.

[0005] Preferably, the target area is obtained by: By performing semantic segmentation on the gray-scale image, the bearing area and the background area are distinguished, and the identified bearing area is taken as the target area.

[0006] Preferably, the gap area in the identified target area is removed, and the method further comprises the following steps: The center point of the bearing ring with the smallest radius is identified, and the position of the bearing end face is adjusted according to the coordinate information of the center point, so that the center point coordinate corresponds to the image center.

[0007] Preferably, the standard precision difference is obtained by: In the three-dimensional Hough circle space, the cumulative coordinate plane is generated by voting, the pixel point with the highest votes is taken as the current center point, and the current center point is taken as the center of the circle with the radius of the radius of the standard bearing, and the difference between the current center point and the center point of the bearing ring is taken as the standard precision difference.

[0008] Preferably, the deviation degree is obtained by: For each bearing ring, the average value of the distances between the center of the circle determined by all adjacent pixel points on the bearing ring and the reference plane is obtained, and the average value of the difference between each distance and the average value is taken as the deviation degree.

[0009] The embodiments of the present application have at least the following beneficial effects: On the basis of digitizing the bearing end face image, the image of the bearing end face is collected by optical means, and the defect identification of the bearing end face is further realized by feature analysis on the gray-scale image of the bearing end face. Through the abnormal specification detection of the bearing, a plurality of bearings can be detected at one time, and the operation is simple and convenient. At the same time, the optical means for bearing end face defect identification operation can detect the fine defects on the bearing edge, and the analysis result is more accurate. BRIEF DESCRIPTION OF DRAWINGS

[0010] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiment or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings according to these drawings without creative labor.

[0011] Figure 1 A step flow chart of a bearing end face defect detection method based on defect identification provided by an embodiment of the present application. DETAILED DESCRIPTION

[0012] In order to further illustrate the technical means and effects taken by the present application to achieve the predetermined inventive objectives, the specific implementation, structure, features and effects of a bearing end face defect detection method based on flaw identification according to the present application are described in detail as follows in combination with the drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0013] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0014] The specific scheme of the bearing end face defect detection method based on flaw identification provided by the present application is specifically described below in combination with the drawings.

[0015] Please refer to Figure 1 which shows the step flowchart of a bearing end face defect detection method based on flaw identification provided by one embodiment of the present application, which includes the following steps: Step S001, the surface image of the bearing end face is collected and the corresponding gray image is obtained, and the bearing area is segmented in the gray image as the target area, and the target area includes four bearing rings.

[0016] After the bearing is assembled, it needs to be detected for its specifications to see if its end face specifications meet the accuracy requirements. The bearing is placed on the conveyor belt, and the surface image of the bearing end face is collected by the camera from above.

[0017] The corresponding gray image is obtained by performing gray scale on the surface image, and the bearing area and the background area are distinguished by performing semantic segmentation on the gray image, and the recognized bearing area is taken as the target area.

[0018] The DNN semantic segmentation method is used to recognize and segment the target area in the image.

[0019] The data set used is the bearing end face image data set under the high-definition camera.

[0020] The pixels that need to be segmented are divided into 2 categories, that is, the label annotation process of the training set is: single-channel semantic label, the pixels at the corresponding position are annotated as 0 if they belong to the background class, and as 1 if they belong to the bearing surface.

[0021] The task of the network is classification, so the loss function used is the cross-entropy loss function.

[0022] Finally, the image with the segmented target area is binarized.

[0023] Step S002, the gap region in the target region is identified and removed to obtain the bearing pixel points in the target region; the bearing pixel points are converted into a Hough circle space to obtain a connected domain corresponding to each bearing ring; the bearing pixel points are converted into a three-dimensional Hough space to form a three-dimensional curve.

[0024] The bearing end face is a concentric circle under normal standards, and after the center of the bearing is determined, the Hough circle conversion is performed on the two bearing rings on the bearing end face. If the current bearing is a standard specification bearing, the two bearing rings should be four standard distributed planes in the Hough circle space, or in the error range, the two bearing rings are points on the four planes divided in the Hough space, and the fluctuation range is within the error allowable range.

[0025] The center point of the bearing ring with the smallest radius is identified, and the position of the bearing end face is adjusted according to the coordinate information of the center point, so that the center point coordinates correspond to the image center.

[0026] The pixel points of the bearing ring with the smallest radius are used as the research object to determine the center point of the current bearing end face, and the detected center point is denoted as The position of the bearing end face is adjusted according to the coordinate position information, so that the center point coordinates and the straight line of the camera collection system in the vertical direction are on the same straight line, or the deviation position of the two is within the allowable range of accuracy.

[0027] If there is a gap on the bearing cover on the bearing end face, it will affect the judgment of the abnormal degree, so it needs to be analyzed. There will be a corresponding gap on the bearing cover, which will affect the judgment of the abnormal degree of the bearing end face at the corresponding position. Therefore, a neural network is used to identify it first. Because of the uncertainty of the position, it cannot be guaranteed that the gap is located in a fixed position during bearing detection. Therefore, a simpler neural network is used to identify it. The identification process is as follows: The DNN semantic segmentation method is used to identify the gap region in the image. The data set used is the bearing end face image data set under the high-definition camera. The pixels to be segmented are divided into two categories, i.e. the training set corresponding label annotation process is: single channel semantic label, the pixel points in the corresponding position are labeled as 0, and the pixel points in the bearing ring gap are labeled as 1. The task of the network is classification, so the loss function used is cross-entropy loss function.

[0028] After the identification of the gap region is completed, the pixel points of the gap region are labeled, and then the corresponding pixel points in the corresponding position of the corresponding bearing ring are removed or directly set to 255 in the binary image to avoid affecting the subsequent abnormal judgment of the bearing ring.

[0029] The upper edge pixel points of the bearing ring are converted into a Hough circle space, the corresponding radius information of the center point under the corresponding standard condition is converted into the Hough space, the point cluster in the Hough space is divided into corresponding planes, and the abnormal degree is judged.

[0030] Since the center point of the end face of the bearing ring is determined by the smallest bearing ring, the Hough space circle conversion is not performed on the first bearing ring, and the set of edge image points on each ring is obtained from the second bearing ring , wherein the upper index indicates that the bearing ring is connected from the inside to the outside of the bearing ring, and the lower index indicates a pixel point in the current bearing ring.

[0031] For a pixel point on a bearing ring, the two-dimensional coordinates of each pixel point are converted into three-dimensional information in a three-dimensional Hough space, and the conversion process is: , wherein, are the coordinate axes in the Hough circle space, is processed as a parameter.

[0032] is the coordinate information of the pixel point in the connected domain in the two-dimensional rectangular coordinate system, and all the circles passing through the point correspond to a three-dimensional rectangular coordinate system a-b-r. All parameters are drawn to obtain a three-dimensional curve.

[0033] Step S003, obtain the standard precision difference between each bearing ring and the corresponding standard circle center, take the standard precision difference corresponding to any one bearing ring as the reference plane, obtain the distance between the circle center determined by each two adjacent bearing pixel points and the reference plane, and then obtain the deviation degree of each bearing ring.

[0034] In the Hough circle three-dimensional space, the cumulative coordinate plane is generated by voting, the pixel point with the highest votes is taken as the current center point, the current center point is taken as the center of the circle, and the radius of the standard bearing is taken as the radius. The difference between the current center point and the bearing ring center point is compared as the standard precision difference.

[0035] For all pixel points on the bearing ring, it is considered that each pixel point may be a point on a potential circle, so in the three-dimensional space of the Hough circle, the cumulative coordinate plane is generated by voting, and the pixel point with the highest votes is taken as the center point of the current bearing ring end face. However, the corresponding bearing center point has been determined at this time, so the concentric circle with the current center pixel point as the center point and the standard distance as the radius is obtained to analyze the pixel points on the current bearing ring. The standard precision differences of the three bearing rings relative to the center point are respectively , , , .

[0036] Since the bearing center point is determined at the beginning, and the bearing specification information is known in the production process, the radius information corresponding to the current bearing end face is known, so the bearing center point is determined by the first bearing ring, and the bearing center point is determined by the first bearing ring. , , , Four planes in the Hough circle space, first for the pixel points on the first bearing ring, the bearing center point is determined by the first bearing ring, so the pixel points on the first bearing ring, the center of the circle determined by the adjacent pixel points, most of them are on the plane with a height of , but because the Hough circle space uses a voting mechanism to determine the center of the circle, some of the pixel points determined by the center of the circle are not on the plane with a height of , and the corresponding plane equation in the Hough circle space is recorded as: Therefore, the distance between the center of the circle determined by each two adjacent pixel points and the plane is: , which is calculated by the distance formula between two points.

[0037] For each bearing ring, get the average distance between the center of the circle determined by all adjacent pixel points on the bearing ring and the reference plane, and take the average of the difference between each distance and the average as the deviation degree.

[0038] Calculate the abnormality degree of the first bearing ring to judge the deviation degree of the first bearing ring: In the formula, represents the average distance between the center of the circle determined by the adjacent pixel points on the current bearing ring and the plane with a height of , , The distance between the center of the circle determined by the pixel points on the current bearing ring and the plane with a height of , when is larger, it means that the confusion degree of the pixel points on the current first bearing ring that do not participate in the center point determination is larger, that is, there are more edge abnormal points on the current first bearing ring, and there may be defects or irregularities on the bearing ring.

[0039] Similarly, take as the reference plane in the Hough circle space, respectively calculate the deviation degree of the bearing ring relative to the center point under the corresponding radius .

[0040] In the bearing specification detection, the fixed point center point is researched, so that the bearing ring can run in the form of concentric circles in use, and some different step problems in running are reduced.

[0041] In step S004, the product of the deviation degrees of the four bearing rings is taken as the overall abnormality degree of the bearing, and when the overall abnormality degree is greater than a preset threshold, the bearing end face has an abnormality.

[0042] After the abnormality degree of each bearing ring relative to the bearing center point is calculated, the overall abnormality degree of the bearing is calculated, and the overall abnormality degree of the bearing is The calculation is When the overall abnormality degree YC is greater than a preset threshold 0.8, it indicates that there is a bearing ring with a high abnormality degree in the current bearing, that is, the current bearing ring has a large abnormality, and the current bearing ring needs to be disassembled and replaced. After replacement, the assembled bearing ring is detected again to determine whether it meets the bearing precision requirement.

[0043] To sum up, the embodiment of the present application collects the surface image of the bearing end face and obtains the corresponding gray image, divides out the bearing area in the gray image as the target area, the target area includes four bearing rings; the gap area in the target area is identified and removed to obtain the bearing pixel points in the target area; the bearing pixel points are converted to the Hough circle space to obtain the corresponding connected domain of each bearing ring; by converting the bearing pixel points to the three-dimensional Hough space, a three-dimensional curve is formed; the standard precision difference between each bearing ring and the corresponding standard circle is obtained, and the standard precision difference corresponding to any one bearing ring is taken as the reference plane, the distance between the center of the circle determined by each two adjacent bearing pixel points and the reference plane is obtained, and then the deviation degree of each bearing ring is obtained; the product of the deviation degrees of the four bearing rings is taken as the overall abnormality degree of the bearing, and when the overall abnormality degree is greater than a preset threshold, the bearing end face has an abnormality. The embodiment of the present application can detect the subtle defects on the bearing edge, and the analysis result is more accurate.

[0044] It should be noted that: the above-mentioned embodiment of the present application is only for description, and does not represent the advantages and disadvantages of the embodiment. And the above describes a specific embodiment of the present application. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or may be advantageous.

[0045] Each embodiment in the present specification is described in a progressive manner, and the same or similar parts of each embodiment can be referred to each other, and each embodiment mainly describes the difference from other embodiments.

[0046] The above-described embodiments are only used to illustrate the technical solutions of the present application, but not limit the present application; the technical solutions described in the foregoing embodiments are modified, or some technical features are replaced equivalently, and the essence of the corresponding technical solutions does not deviate from the scope of the technical solutions of the embodiments of the present application, which should be included in the protection scope of the present application.

Claims

1. A method for detecting bearing end face defects based on defect identification, characterized in that, The method includes the following steps: Acquire a surface image of the bearing end face and obtain the corresponding grayscale image. Segment the bearing region in the grayscale image as the target region. The target region includes four bearing rings. The missing areas within the target area are identified and removed to obtain the bearing pixels within the target area; the bearing pixels are converted into Hough circle space to obtain the connected components corresponding to each bearing circle; by converting the bearing pixels into three-dimensional Hough space, a three-dimensional curve is formed. Obtain the standard accuracy difference between each bearing ring and the standard circle corresponding to its center. Using the standard accuracy difference of any bearing ring as a reference plane, obtain the distance between the center of each two adjacent bearing pixels and the reference plane, and thus obtain the degree of deviation of each bearing ring. The overall abnormality of the bearing is determined by the product of the deviations of the four bearing rings. When the overall abnormality exceeds a preset threshold, an abnormality is found on the bearing end face.

2. The bearing end face defect detection method based on defect identification according to claim 1, characterized in that, The method for obtaining the target region is as follows: By performing semantic segmentation on the grayscale image, the bearing region and the background region are distinguished, and the identified bearing region is used as the target region.

3. The bearing end face defect detection method based on defect identification according to claim 1, characterized in that, Before removing the gaps within the identified target area, the following steps are also included: Identify the center point of the bearing ring with the smallest radius, and adjust the position of the bearing end face according to the coordinate information of the center point so that the coordinates of the center point correspond to the center of the image.

4. The bearing end face defect detection method based on defect identification according to claim 1, characterized in that, The method for obtaining the standard accuracy difference is as follows: In the three-dimensional space of the Hough circle, a cumulative coordinate plane is generated through voting. The pixel with the highest number of votes is taken as the current center point. A circle is drawn with the current center point as the center and the radius of the standard bearing as the radius. The difference between the current center point and the center point of the bearing ring is used as the standard accuracy difference.

5. The bearing end face defect detection method based on defect identification according to claim 1, characterized in that, The method for obtaining the degree of deviation is as follows: For each bearing ring, the average distance between the center of the circle determined by all adjacent pixels on the bearing ring and the reference plane is obtained, and the average difference between each distance and the average value is used as the deviation degree.

Images