Automobile lining inner surface defect detection method and system based on image processing

Abstract

The invention relates to the technical field of image processing application, in particular to an automobile lining inner surface defect detection method and system based on image processing. The method comprises the following steps: firstly, obtaining N automobile lining pipe samples produced in the same batch; secondly, longitudinally cutting the automobile lining pipe samples, and collecting an inner surface image corresponding to each cut automobile lining pipe sample; thirdly, the inner surface image is input into a lining pipe image segmentation model, and a chamfering area image and a non-chamfering area image are obtained; then, respectively carrying out defect detection on the chamfered area image and the non-chamfered area image, and obtaining an inner surface defect detection result corresponding to each automobile lining pipe sample; and finally, according to an inner surface defect detection result, obtaining an inner surface defect detection qualification rate corresponding to the automobile lining pipe sample. By means of the method, various inner surface defects of the automobile lining pipe sample can be synthesized, and the accuracy of inner surface defect detection can be improved.

Description

Technical Field

[0001] The invention relates to the technical field of image processing applications, and in particular to an image processing-based method and system for detecting inner surface defects of an automobile bushing. Background Art

[0002] Automobile bushings are a type of component used between various mechanical connection parts of automobiles, usually made of rubber, polyurethane, metal or composite materials. Its main function is to reduce friction between moving parts, reduce shock, vibration and carry load. Automobile bushings consist of inner liner tubes, outer liner tubes and rubber tubes. The inner liner tube is usually the core part of the bushing, providing the necessary mechanical support for the automobile bushing, enhancing the structural strength of the bushing, and preventing the bushing from deformation or damage due to external forces.

[0003] The inner surface of the automobile bushing includes the inner surface of the inner liner tube, which is the surface inside the bushing that contacts the automobile metal parts (such as control arms, bearings, and suspension arms, etc.). During the production process, the inner surface of the inner liner tube will be affected by factors such as temperature changes and external force impacts, and may have different forms of damage or defects, such as cracks, pores, and surface scratches. These defects may cause the performance of the inner liner tube to deteriorate, thereby affecting the quality of the automobile bushing, so it is necessary to perform inner surface defect detection on the inner liner tube of the automobile bushing.

[0004] During the processing of the inner liner pipe, the inner surface of the pipe mouth will be chamfered to form a chamfered area. The existing inner surface defect detection of the inner liner pipe in the automobile bushing is mostly based on simple image recognition. Although this method can realize the inner surface defect detection, it does not detect the chamfered area and non-chamfered area of ​​the inner surface of the inner liner pipe separately, which is easy to cause false detection of the inner surface defect detection of the automobile bushing, thereby reducing the accuracy of the automobile bushing quality inspection.

[0005] Therefore, a method and system for detecting inner surface defects of automobile bushing based on image processing are proposed. Summary of the invention

[0006] The purpose of the present invention is to provide an automobile bushing inner surface defect detection method and system based on image processing. First, obtain N automobile bushing inner liner tube samples produced in the same batch; second, cut the automobile bushing inner liner tube samples longitudinally, and collect the inner surface image corresponding to each automobile bushing inner liner tube sample after cutting; third, input the inner surface image into the inner liner tube image segmentation model to obtain the chamfered area image and the non-chamfered area image; then, perform defect detection on the chamfered area image and the non-chamfered area image respectively, and obtain the inner surface defect detection result corresponding to each automobile bushing inner liner tube sample; finally, according to the inner surface defect detection result, obtain the inner surface defect detection qualified rate corresponding to the automobile bushing inner liner tube sample.

[0007] To achieve the above object, the present invention provides the following technical solutions: A method for detecting inner surface defects of automobile bushings based on image processing, comprising: Obtain N samples of automobile bushing liner tubes produced in the same batch; Cutting the automobile bushing inner lining tube sample longitudinally, and collecting the inner surface image corresponding to each automobile bushing inner lining tube sample after cutting; Inputting the inner surface image into the inner liner tube image segmentation model to obtain the chamfered area image and the non-chamfered area image; The inner liner tube image segmentation model includes an inner liner tube image input layer, an inner liner tube image feature extraction layer, an inner liner tube image feature fusion layer and an inner liner tube image segmentation layer; the inner liner tube image feature extraction layer is used to extract multiple inner liner tube feature maps of the inner liner tube image; the inner liner tube image feature fusion layer is used to perform feature fusion on multiple inner liner tube feature maps to obtain inner surface color features and inner surface texture features; The inner liner image segmentation layer is used to segment the inner surface image according to the inner surface color features and the inner surface texture features to obtain the chamfered area image and the non-chamfered area image; Perform defect detection on the chamfered area image and the non-chamfered area image respectively to obtain a first inner surface defect detection value and a second inner surface defect detection value; According to the first inner surface defect detection value and the second inner surface defect detection value, obtaining the inner surface defect detection result corresponding to each of the automobile bushing inner liner tube samples; According to the inner surface defect detection result corresponding to each of the automobile bushing inner lining tube samples, the inner surface defect detection pass rate corresponding to the automobile bushing inner lining tube samples is obtained.

[0008] Further, performing defect detection on the chamfer area image to obtain a first inner surface defect detection value includes: Inputting the chamfer area image into a chamfer area defect detection model to obtain a first defect feature of the chamfer area; the first defect feature includes the number of defects, the defect type of each defect, and the defect area; According to the chamfered area image, obtaining an average width and a uniform width value of the chamfered area; A first defect characteristic value is obtained based on the first defect characteristic; and the first inner surface defect detection value is obtained based on the first defect characteristic value, the average width and the width uniform value.

[0009] Furthermore, the calculation formula of the first inner surface defect detection value is: ; in, It is represented as the first inner surface defect detection value corresponding to the i-th automobile bushing inner liner tube sample; the value of i ranges from 1 to N; , and Respectively expressed as The average width, uniform width and first defect characteristic value of the corresponding chamfered area; Expressed as a set standard width; , and They are respectively expressed as the first width coefficient, the second width coefficient and the defect characteristic coefficient; Expressed as an exponential function with a natural constant as base.

[0010] Furthermore, the process of obtaining the second inner surface defect detection value includes: inputting the non-chamfered area image into the non-chamfered area defect detection model to obtain the second defect feature of the non-chamfered area; the second defect feature includes the number of defects, the type of each defect and the defect area; and obtaining the second inner surface defect detection value based on the second defect feature.

[0011] Furthermore, the process of obtaining the inner surface defect detection result corresponding to each of the automobile bushing inner liner tube samples includes: According to the first inner surface defect detection value and the second inner surface defect detection value, the inner surface comprehensive defect detection value corresponding to the automobile bushing inner liner tube sample is obtained, and the calculation formula is: ; in, It is represented by the comprehensive defect detection value of the inner surface corresponding to the i-th automobile bushing inner liner tube sample; the value of i ranges from 1 to N; and They are respectively represented as the first inner surface defect detection value and the second inner surface defect detection value corresponding to the i-th automobile bushing inner liner tube sample; and They are respectively expressed as the defect weight of the chamfered area and the defect weight of the non-chamfered area; When the first inner surface defect detection value is less than the set first defect threshold, the second inner surface defect detection value is less than the set second defect threshold, and the inner surface comprehensive defect detection value is less than the set comprehensive defect threshold, the inner surface defect detection result corresponding to the automobile bushing inner liner tube sample is that the inner surface defect detection is qualified; otherwise, it indicates that the inner surface defect detection is unqualified.

[0012] An automobile bushing inner surface defect detection system based on image processing, comprising: Automobile bushing inner liner tube sample acquisition unit: used to obtain N automobile bushing inner liner tube samples produced in the same batch; Inner surface image acquisition unit: used for longitudinally cutting the automobile bushing inner lining tube sample and collecting the inner surface image corresponding to each automobile bushing inner lining tube sample after cutting; Inner surface image segmentation unit: used for inputting the inner surface image into the inner liner tube image segmentation model to obtain the chamfered area image and the non-chamfered area image; Defect detection unit: used to perform defect detection on the chamfered area image and the non-chamfered area image respectively, and obtain a first inner surface defect detection value and a second inner surface defect detection value; An inner surface defect detection result acquisition unit is used to acquire an inner surface defect detection result corresponding to each of the automobile bushing inner liner tube samples according to the first inner surface defect detection value and the second inner surface defect detection value; Inner surface defect detection qualified rate acquisition unit: used to obtain the inner surface defect detection qualified rate corresponding to the automobile bushing inner lining tube sample according to the inner surface defect detection result corresponding to each of the automobile bushing inner lining tube samples.

[0013] Furthermore, the inner tube image segmentation model includes an inner tube image input layer, an inner tube image feature extraction layer, an inner tube image feature fusion layer and an inner tube image segmentation layer; the inner tube image feature extraction layer is used to extract multiple inner tube feature maps of the inner tube image; the inner tube image feature fusion layer is used to perform feature fusion on multiple inner tube feature maps to obtain inner surface color features and inner surface texture features; The inner liner tube image segmentation layer is used to segment the inner surface image according to the inner surface color features and the inner surface texture features to obtain the chamfered area image and the non-chamfered area image.

[0014] Further, performing defect detection on the chamfer area image to obtain a first inner surface defect detection value includes: Inputting the chamfer area image into a chamfer area defect detection model to obtain a first defect feature of the chamfer area; the first defect feature includes the number of defects, the defect type of each defect, and the defect area; According to the chamfered area image, obtaining an average width and a uniform width value of the chamfered area; A first defect characteristic value is obtained based on the first defect characteristic; and the first inner surface defect detection value is obtained based on the first defect characteristic value, the average width and the width uniform value.

[0015] Furthermore, the calculation formula of the first inner surface defect detection value is: ; in, It is represented as the first inner surface defect detection value corresponding to the i-th automobile bushing inner liner tube sample; the value of i ranges from 1 to N; , and Respectively expressed as The average width, uniform width and first defect characteristic value of the corresponding chamfered area; Expressed as a set standard width; , and They are respectively expressed as the first width coefficient, the second width coefficient and the defect characteristic coefficient; Expressed as an exponential function with a natural constant as base.

[0016] Furthermore, the process of obtaining the inner surface defect detection result corresponding to each of the automobile bushing inner liner tube samples includes: According to the first inner surface defect detection value and the second inner surface defect detection value, the inner surface comprehensive defect detection value corresponding to the automobile bushing inner liner tube sample is obtained, and the calculation formula is: ; in, It is represented by the comprehensive defect detection value of the inner surface corresponding to the i-th automobile bushing inner liner tube sample; the value of i ranges from 1 to N; and They are respectively represented as the first inner surface defect detection value and the second inner surface defect detection value corresponding to the i-th automobile bushing inner liner tube sample; and They are respectively expressed as the defect weight of the chamfered area and the defect weight of the non-chamfered area; When the first inner surface defect detection value is less than the set first defect threshold, the second inner surface defect detection value is less than the set second defect threshold, and the inner surface comprehensive defect detection value is less than the set comprehensive defect threshold, the inner surface defect detection result corresponding to the automobile bushing inner liner tube sample is that the inner surface defect detection is qualified; otherwise, it indicates that the inner surface defect detection is unqualified.

[0017] Compared with the prior art, the present invention has the following beneficial effects: 1. The present invention uses the inner liner tube image segmentation model to perform image segmentation on the inner surface of the automobile bushing inner liner tube, and extracts the inner surface color features and inner surface texture features of the inner surface image to segment the inner surface image into a chamfered area image and a non-chamfered area image. This method can accurately segment the inner surface image based on the various features of the inner surface image, thereby improving the accuracy of inner surface defect detection.

[0018] 2. The present invention performs defect detection on the chamfered area of ​​the inner surface of the liner pipe to obtain a first inner surface defect detection value. The first inner surface defect detection value can not only measure defects such as cracks and foreign matter in the chamfered area, but also measure whether the width of the chamfered area is compliant. This method can comprehensively consider various defects in the chamfered area and accurately obtain the first inner surface defect detection value corresponding to the chamfered area, thereby improving the accuracy of inner surface defect detection.

[0019] 3. There are differences in texture and color between the chamfered area and the non-chamfered area of ​​the liner tube. The present invention performs defect detection on the chamfered area and the non-chamfered area of ​​the liner tube respectively, which can effectively identify and distinguish the defects of the two. Corresponding detection methods are set for different areas to obtain the inner surface defect detection values ​​of different areas; and the inner surface defect detection results of the automobile bushing inner liner tube samples are judged by the inner surface defect detection values ​​of different areas. This method can comprehensively detect various inner surface defects of automobile bushing inner liner tube samples and improve the accuracy of inner surface defect detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 A schematic flow chart of a method for detecting inner surface defects of an automobile bushing based on image processing according to the present invention; Figure 2 It is a structural schematic diagram of an automobile bushing inner surface defect detection system based on image processing of the present invention; Figure 3 A schematic diagram of a sample of an automotive bushing inner liner tube provided by an embodiment of the present invention; Figure 4 A flow chart for obtaining the first inner surface defect detection value provided in this embodiment.

[0021] In the figure: 1. Chamfered area; 2. Non-chamfered area. DETAILED DESCRIPTION

[0022] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0023] The present invention provides a method for detecting inner surface defects of automobile bushings based on image processing. The method is applied to an inner surface defect detection system of automobile bushings based on image processing. For specific method flow chart and system structure chart, refer to Figure 1 and Figure 2 .

[0024] Embodiment 1 Reference Figure 1 S10 in the figure is applied to a sample acquisition unit of an automobile bushing inner surface defect detection system based on image processing.

[0025] Furthermore, this embodiment conducts sampling inspection on the same batch of automobile bushing inner liner tubes produced by Factory A, and a total of 260 automobile bushing inner liner tube samples are drawn. Figure 3 The inner surface of the automobile bushing inner liner tube sample includes a chamfered area 1 and a non-chamfered area 2. The specific standard size information of the automobile bushing inner liner tube sample is shown in Table 1.

[0026] Table 1 Standard size information table of automotive bushing liner pipe samples

[0027] Reference Figure 1 S20 in the figure is applied to an inner surface image acquisition unit of an automobile bushing inner surface defect detection system based on image processing.

[0028] Furthermore, in this embodiment, longitudinal cutting is performed on each of the automobile bushing inner lining tube samples; the longitudinal cutting is performed along the length direction of the automobile bushing inner lining tube sample; the longitudinal cutting method can be selected from water jet cutting or laser cutting. In this embodiment, the automobile bushing inner lining tube sample is cut by water jet cutting, and the inner surface image corresponding to each automobile bushing inner lining tube sample after cutting is collected.

[0029] Reference Figure 1 S30 in the embodiment is applied to an inner surface image segmentation unit of an inner surface defect detection system for an automobile bushing based on image processing.

[0030] Furthermore, the inner surface image is input into the inner liner image segmentation model to obtain the chamfered area image and the non-chamfered area image; the inner liner image segmentation model includes an inner liner image input layer, an inner liner image feature extraction layer, an inner liner image feature fusion layer and an inner liner image segmentation layer; The inner liner tube image feature extraction layer is used to extract multiple inner liner tube feature maps of the inner liner tube image; the inner liner tube image feature extraction layer uses the inner liner tube SegNet to extract the inner liner tube feature map; the inner liner tube SegNet is constructed by adding a CBAM attention mechanism after each convolution layer of the SegNet encoder; The inner liner image feature fusion layer is used to perform feature fusion on multiple inner liner feature maps to obtain inner surface color features and inner surface texture features; the inner liner image feature fusion layer uses pyramid pooling to perform feature map fusion; The inner liner tube image segmentation layer is used to segment the inner surface image according to the inner surface color features and the inner surface texture features to obtain the chamfered area image and the non-chamfered area image.

[0031] In order to verify the effectiveness of the liner tube image segmentation model provided in this embodiment, the inner surface images corresponding to 260 automobile bushing liner tube samples were segmented four times by the liner tube image segmentation model, and the segmentation accuracy was calculated; at the same time, the other two models were used to segment the surface images four times, and the segmentation accuracy was calculated; the specific data refer to Table 2; the calculation formula of the segmentation accuracy is: ; in, Expressed as the segmentation accuracy corresponding to the p-th model; It is represented by the total number of pixels of the inner surface image corresponding to the qth automobile bushing inner liner tube sample; It is represented by the number of pixels that are correctly segmented by the p-th model for the inner surface image corresponding to the q-th automobile bushing liner tube sample.

[0032] Table 2 Segmentation accuracy verification table

[0033] Model 1 in Table 2 is the lined tube image segmentation model provided in this embodiment; Model 2 is based on Model 1, but does not add the CBAM attention mechanism after SegNet; Model 3 uses maximum pooling for feature map fusion based on Model 2; Model 4 is the SegNet model; It can be seen from Table 2 that the lined tube image segmentation model provided in this embodiment can improve the accuracy of lined tube inner surface image segmentation.

[0034] In this step of the present embodiment, the inner surface of the inner liner tube of the automobile bushing is segmented by the inner liner tube image segmentation model, and the inner surface image is segmented into the chamfered area image and the non-chamfered area image by extracting the inner surface color features and the inner surface texture features of the inner surface image. This method can realize the accurate segmentation of the inner surface image based on the various features of the inner surface image, thereby improving the accuracy of the inner surface defect detection.

[0035] Reference Figure 1 S40 in the figure is applied to a defect detection unit of an automobile bushing inner surface defect detection system based on image processing.

[0036] Further, defect detection is performed on the chamfered area image and the non-chamfered area image respectively to obtain a first inner surface defect detection value and a second inner surface defect detection value; referring to Figure 4 , performing defect detection on the chamfer area image to obtain a first inner surface defect detection value includes: The chamfer area image is input into the chamfer area defect detection model to obtain the first defect feature of the chamfer area; the first defect feature includes the number of defects, the defect type and the defect area of ​​each defect; the defect type includes but is not limited to scratch defects and foreign body defects; in this embodiment, ResNet is selected as the chamfer area defect detection model to perform defect detection on the chamfer area image; According to the chamfered area image, obtaining an average width and a uniform width value of the chamfered area; A first defect characteristic value is obtained based on the first defect characteristic; and the first inner surface defect detection value is obtained based on the first defect characteristic value, the average width and the width uniform value.

[0037] Furthermore, the process of obtaining the average width value includes: M points are randomly selected in the chamfer area, and the chamfer area widths corresponding to the M points are obtained through the chamfer area image; the variance of the chamfer area widths corresponding to the M points is obtained as the average value of the width of the chamfer area corresponding to the liner pipe sample.

[0038] Furthermore, the calculation formula of the first defect characteristic value is: ; in, It is represented as the first defect characteristic value of the chamfer area corresponding to the i-th automobile bushing inner liner tube sample; It is represented by the number of defects in the first defect feature corresponding to the i-th automobile bushing inner liner tube sample; and They are respectively represented as the type identification value and area of ​​the j-th defect; the type identification value corresponding to each defect type is preset; and They are expressed as defect quantity coefficient and defect shape coefficient respectively; and The default value is 0.5, and can be changed according to actual application.

[0039] This step of the present embodiment performs defect detection on the chamfered area on the inner surface of the liner pipe, obtains the width characteristics and defect characteristics of the chamfered area, and obtains the first defect characteristic value based on the defect characteristics of the chamfered area; this method can combine the various defect characteristics of the chamfered area, accurately obtain the first defect characteristic value corresponding to the chamfered area, and provide a data basis for the subsequent calculation of the first inner surface defect detection value.

[0040] Furthermore, the calculation formula of the first inner surface defect detection value is: ; in, It is represented as the first inner surface defect detection value corresponding to the i-th automobile bushing inner liner tube sample; the value of i ranges from 1 to N; , and Respectively expressed as The average width, uniform width and first defect characteristic value of the corresponding chamfered area; Expressed as a set standard width; , and They are respectively expressed as the first width coefficient, the second width coefficient and the defect characteristic coefficient; , and The default is , and can also be changed according to actual applications; Expressed as an exponential function with a natural constant as base.

[0041] In this embodiment, this step obtains the first inner surface defect detection value, which can not only measure defects such as cracks and foreign matter in the chamfered area, but also measure whether the width of the chamfered area is compliant. This method can comprehensively consider various defects in the chamfered area and accurately obtain the first inner surface defect detection value corresponding to the chamfered area, thereby improving the accuracy of inner surface defect detection.

[0042] Furthermore, the process of obtaining the second inner surface defect detection value includes: inputting the non-chamfered area image into the non-chamfered area defect detection model to obtain the second defect feature of the non-chamfered area; this embodiment selects DenseNet as the non-chamfered area defect detection model to perform defect detection on the non-chamfered area image; the second defect feature includes the number of defects, the type of each defect and the defect area; nonlinear regression is performed based on the second defect feature to obtain the second inner surface defect detection value; the nonlinear regression model can select a neural network model or a support vector machine regression model; this embodiment selects a neural network model for nonlinear regression.

[0043] This step of the present embodiment performs defect detection on the non-chamfered area on the inner surface of the liner pipe to obtain a second inner surface defect detection value. This method can comprehensively analyze various defects in the non-chamfered area and accurately obtain the second inner surface defect detection value corresponding to the non-chamfered area, thereby improving the accuracy of inner surface defect detection.

[0044] Reference Figure 1 S50 in the embodiment is applied to an inner surface defect detection result acquisition unit of an inner surface defect detection system for an automobile bushing based on image processing.

[0045] Furthermore, the process of obtaining the inner surface defect detection result corresponding to each of the automobile bushing inner liner tube samples includes: According to the first inner surface defect detection value and the second inner surface defect detection value, the inner surface comprehensive defect detection value corresponding to the automobile bushing inner liner tube sample is obtained, and the calculation formula is: ; in, It is represented by the comprehensive defect detection value of the inner surface corresponding to the i-th automobile bushing inner liner tube sample; the value of i ranges from 1 to N; and They are respectively represented as the first inner surface defect detection value and the second inner surface defect detection value corresponding to the i-th automobile bushing inner liner tube sample; and They are respectively expressed as the defect weight of the chamfered area and the defect weight of the non-chamfered area; and It is set based on experience; When the first inner surface defect detection value is less than the set first defect threshold, the second inner surface defect detection value is less than the set second defect threshold, and the inner surface comprehensive defect detection value is less than the set comprehensive defect threshold, the inner surface defect detection result corresponding to the automobile bushing inner liner tube sample is that the inner surface defect detection is qualified; otherwise, it indicates that the inner surface defect detection is unqualified; the first defect threshold, the second defect threshold and the comprehensive defect threshold are obtained based on experience.

[0046] There are differences in texture and color between the chamfered area and the non-chamfered area of ​​the liner tube. This step of the present embodiment performs defect detection on the chamfered area and the non-chamfered area of ​​the liner tube respectively, which can effectively identify and distinguish the defects of the two. Corresponding detection methods are set for different areas to obtain the inner surface defect detection values ​​of different areas; and the inner surface defect detection results of the automobile bushing inner liner tube samples are judged by the inner surface defect detection values ​​of different areas. This method can comprehensively detect various inner surface defects of automobile bushing inner liner tube samples and improve the accuracy of inner surface defect detection.

[0047] Reference Figure 1 S60 in the embodiment is applied to an inner surface defect detection pass rate acquisition unit of an inner surface defect detection system for an automobile bushing based on image processing.

[0048] Furthermore, the number of qualified inner surface defect detection samples of automobile bushing inner liner tubes produced in the same batch obtained by the method of this embodiment is 243, and the number of unqualified is 17, and the qualified rate of inner surface defect detection is 93.46%.

[0049] Embodiment 2 In Example 1, the method of the present invention is combined with a system to realize the inner surface defect detection of the inner lining tube of the automobile bushing; in Example 2 of the present application, the method of the present invention will be explained again.

[0050] S10, obtaining N automobile bushing liner tube samples produced in the same batch.

[0051] Furthermore, this embodiment conducts sampling inspection on the automobile bushing inner liner tubes produced in the same batch by Factory B, and a total of 230 automobile bushing inner liner tube samples are taken. The specific standard size information of the automobile bushing inner liner tube samples is shown in Table 3.

[0052] Table 3 Standard size information table of automotive bushing liner pipe samples

[0053] S20, cutting the automobile bushing inner lining tube sample longitudinally, and collecting the inner surface image corresponding to each automobile bushing inner lining tube sample after cutting.

[0054] Furthermore, in this embodiment, each of the automobile bushing inner lining tube samples is cut longitudinally; the longitudinal cutting is performed according to the diameter of the automobile bushing inner lining tube sample; the longitudinal cutting method can be selected from water jet cutting or laser cutting. In this embodiment, the automobile bushing inner lining tube sample is cut by water jet cutting, and the inner surface image corresponding to each automobile bushing inner lining tube sample after cutting is collected.

[0055] S30, inputting the inner surface image into the inner liner tube image segmentation model to obtain the chamfered area image and the non-chamfered area image.

[0056] Furthermore, the inner surface image is input into the inner liner image segmentation model to obtain the chamfered area image and the non-chamfered area image; the inner liner image segmentation model includes an inner liner image input layer, an inner liner image feature extraction layer, an inner liner image feature fusion layer and an inner liner image segmentation layer; The inner liner tube image feature extraction layer is used to extract multiple inner liner tube feature maps of the inner liner tube image; the inner liner tube image feature extraction layer uses the inner liner tube SegNet to extract the inner liner tube feature map; the inner liner tube SegNet is constructed by adding a CBAM attention mechanism after each convolution layer of the SegNet encoder; The inner liner image feature fusion layer is used to perform feature fusion on multiple inner liner feature maps to obtain inner surface color features and inner surface texture features; the inner liner image feature fusion layer uses pyramid pooling to perform feature map fusion; The inner liner tube image segmentation layer is used to segment the inner surface image according to the inner surface color features and the inner surface texture features to obtain the chamfered area image and the non-chamfered area image.

[0057] S40, performing defect detection on the chamfered area image and the non-chamfered area image respectively to obtain a first inner surface defect detection value and a second inner surface defect detection value.

[0058] Further, performing defect detection on the chamfer area image to obtain a first inner surface defect detection value includes: Inputting the chamfer area image into the chamfer area defect detection model to obtain a first defect feature of the chamfer area; the first defect feature includes the number of defects, the defect type and defect area of ​​each defect; the defect type includes but is not limited to scratch defects and foreign body defects; According to the chamfered area image, obtaining an average width and a uniform width value of the chamfered area; A first defect characteristic value is obtained based on the first defect characteristic; and the first inner surface defect detection value is obtained based on the first defect characteristic value, the average width and the width uniform value.

[0059] Furthermore, the process of obtaining the average width value includes: M points are randomly selected in the chamfer area, and the chamfer area widths corresponding to the M points are obtained through the chamfer area image; the variance of the chamfer area widths corresponding to the M points is obtained as the average value of the width of the chamfer area corresponding to the liner pipe sample.

[0060] Furthermore, the calculation formula of the first defect characteristic value is: ; in, It is represented as the first defect characteristic value of the chamfer area corresponding to the i-th automobile bushing inner liner tube sample; It is represented by the number of defects in the first defect feature corresponding to the i-th automobile bushing inner liner tube sample; and They are respectively represented as the type identification value and area of ​​the j-th defect; the type identification value corresponding to each defect type is preset; and They are expressed as defect quantity coefficient and defect shape coefficient respectively; and The default value is 0.5, and can be changed according to actual application.

[0061] Furthermore, the calculation formula of the first inner surface defect detection value is: ; in, It is represented as the first inner surface defect detection value corresponding to the i-th automobile bushing inner liner tube sample; the value of i ranges from 1 to N; , and Respectively expressed as The average width, uniform width and first defect characteristic value of the corresponding chamfered area; Expressed as a set standard width; , and They are respectively expressed as the first width coefficient, the second width coefficient and the defect characteristic coefficient; , and The default is , and can also be changed according to actual applications; Expressed as an exponential function with a natural constant as base.

[0062] Furthermore, the process of obtaining the second inner surface defect detection value includes: inputting the non-chamfered area image into the non-chamfered area defect detection model to obtain the second defect feature of the non-chamfered area; the second defect feature includes the number of defects, the type of each defect and the defect area; performing nonlinear regression based on the second defect feature to obtain the second inner surface defect detection value; the nonlinear regression model can select a neural network model or a support vector machine regression model; this embodiment selects a neural network model for nonlinear regression.

[0063] S50. Obtain the inner surface defect detection result corresponding to each of the automobile bushing inner liner tube samples based on the first inner surface defect detection value and the second inner surface defect detection value.

[0064] Furthermore, the process of obtaining the inner surface defect detection result corresponding to each of the automobile bushing inner liner tube samples includes: According to the first inner surface defect detection value and the second inner surface defect detection value, the inner surface comprehensive defect detection value corresponding to the automobile bushing inner liner tube sample is obtained, and the calculation formula is: ; in, It is represented by the comprehensive defect detection value of the inner surface corresponding to the i-th automobile bushing inner liner tube sample; the value of i ranges from 1 to N; and They are respectively represented as the first inner surface defect detection value and the second inner surface defect detection value corresponding to the i-th automobile bushing inner liner tube sample; and They are respectively expressed as the defect weight of the chamfered area and the defect weight of the non-chamfered area; When the first inner surface defect detection value is less than the set first defect threshold, the second inner surface defect detection value is less than the set second defect threshold, and the inner surface comprehensive defect detection value is less than the set comprehensive defect threshold, the inner surface defect detection result corresponding to the automobile bushing inner liner tube sample is that the inner surface defect detection is qualified; otherwise, it indicates that the inner surface defect detection is unqualified.

[0065] S60. Obtain the inner surface defect detection pass rate corresponding to each of the automobile bushing inner lining tube samples according to the inner surface defect detection result corresponding to each of the automobile bushing inner lining tube samples.

[0066] Furthermore, the method of this embodiment obtains 214 qualified inner surface defect samples of the same batch of automobile bushing inner liner tubes, 16 unqualified samples, and the inner surface defect detection pass rate is 93.04%.

[0067] In order to verify the effectiveness of the inner surface defect detection method provided in this embodiment, this embodiment also obtained the inner surface images of 600 automobile bushing liner tube samples; and adopted four different schemes for inner surface defect detection; based on the inner surface defect detection results, the precision, recall rate and F1-score corresponding to each scheme were obtained. For specific detection results, refer to Table 4.

[0068] Table 4 Inner surface defect detection verification table

[0069] In Table 4, Scheme 1 is the inner surface defect detection method provided by this embodiment; Scheme 2 is to segment the inner surface image using a threshold segmentation method based on Scheme 1; Scheme 3 is to not segment the inner surface image based on Scheme 1; Scheme 4 is to perform defect detection based only on the image of the non-chamfered area based on Scheme 1. It can be seen from Table 4 that the method provided by this embodiment can improve the accuracy of inner surface defect detection of the inner lining tube of the automobile bushing.

[0070] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for detecting inner surface defects of automobile bushings based on image processing, characterized in that: include: Obtain N samples of automobile bushing liner tubes produced in the same batch; Cutting the automobile bushing inner lining tube sample longitudinally, and collecting the inner surface image corresponding to each automobile bushing inner lining tube sample after cutting; Inputting the inner surface image into the inner liner tube image segmentation model to obtain the chamfered area image and the non-chamfered area image; The inner liner tube image segmentation model includes an inner liner tube image input layer, an inner liner tube image feature extraction layer, an inner liner tube image feature fusion layer and an inner liner tube image segmentation layer; The inner liner tube image feature extraction layer is used to extract multiple inner liner tube feature maps of the inner liner tube image; The inner liner tube image feature fusion layer is used to perform feature fusion on multiple inner liner tube feature images to obtain inner surface color features and inner surface texture features; The inner liner image segmentation layer is used to segment the inner surface image according to the inner surface color features and the inner surface texture features to obtain the chamfered area image and the non-chamfered area image; Perform defect detection on the chamfered area image and the non-chamfered area image respectively to obtain a first inner surface defect detection value and a second inner surface defect detection value; According to the first inner surface defect detection value and the second inner surface defect detection value, obtaining the inner surface defect detection result corresponding to each of the automobile bushing inner liner tube samples; According to the inner surface defect detection result corresponding to each of the automobile bushing inner lining tube samples, the inner surface defect detection pass rate corresponding to the automobile bushing inner lining tube samples is obtained.

2. The method for detecting inner surface defects of automobile bushings based on image processing according to claim 1 is characterized in that: Performing defect detection on the chamfer area image to obtain a first inner surface defect detection value includes: Inputting the chamfer area image into a chamfer area defect detection model to obtain a first defect feature of the chamfer area; the first defect feature includes the number of defects, the defect type of each defect, and the defect area; According to the chamfered area image, obtaining an average width and a uniform width value of the chamfered area; A first defect characteristic value is obtained based on the first defect characteristic; and the first inner surface defect detection value is obtained based on the first defect characteristic value, the average width and the width uniform value.

3. The method for detecting inner surface defects of automobile bushings based on image processing according to claim 2 is characterized in that: The calculation formula of the first inner surface defect detection value is: ； in, It is represented as the first inner surface defect detection value corresponding to the i-th automobile bushing inner liner tube sample; the value of i ranges from 1 to N; , and Respectively expressed as The average width, uniform width and first defect characteristic value of the corresponding chamfered area; Expressed as a set standard width; , and They are respectively expressed as the first width coefficient, the second width coefficient and the defect characteristic coefficient; Expressed as an exponential function with a natural constant as base.

4. The method for detecting inner surface defects of automobile bushings based on image processing according to claim 1 is characterized in that: The process of obtaining the second inner surface defect detection value includes: inputting the non-chamfered area image into the non-chamfered area defect detection model to obtain the second defect feature of the non-chamfered area; the second defect feature includes the number of defects, the type of each defect and the defect area; and obtaining the second inner surface defect detection value based on the second defect feature.

5. The method for detecting inner surface defects of automobile bushings based on image processing according to claim 1 is characterized in that: The process of obtaining the inner surface defect detection result corresponding to each automobile bushing inner liner tube sample includes: According to the first inner surface defect detection value and the second inner surface defect detection value, the inner surface comprehensive defect detection value corresponding to the automobile bushing inner liner tube sample is obtained, and the calculation formula is: ； in, It is represented by the comprehensive defect detection value of the inner surface corresponding to the i-th automobile bushing inner liner tube sample; the value of i ranges from 1 to N; and They are respectively represented as the first inner surface defect detection value and the second inner surface defect detection value corresponding to the i-th automobile bushing inner liner tube sample; and They are respectively expressed as the defect weight of the chamfered area and the defect weight of the non-chamfered area; When the first inner surface defect detection value is less than the set first defect threshold, the second inner surface defect detection value is less than the set second defect threshold, and the inner surface comprehensive defect detection value is less than the set comprehensive defect threshold, the inner surface defect detection result corresponding to the automobile bushing inner liner tube sample is that the inner surface defect detection is qualified; otherwise, it indicates that the inner surface defect detection is unqualified.

6. An automobile bushing inner surface defect detection system based on image processing, characterized in that: include: Automobile bushing inner liner tube sample acquisition unit: used to obtain N automobile bushing inner liner tube samples produced in the same batch; Inner surface image acquisition unit: used for longitudinally cutting the automobile bushing inner lining tube sample and collecting the inner surface image corresponding to each automobile bushing inner lining tube sample after cutting; Inner surface image segmentation unit: used for inputting the inner surface image into the inner liner tube image segmentation model to obtain the chamfered area image and the non-chamfered area image; Defect detection unit: used to perform defect detection on the chamfered area image and the non-chamfered area image respectively, and obtain a first inner surface defect detection value and a second inner surface defect detection value; An inner surface defect detection result acquisition unit is used to acquire an inner surface defect detection result corresponding to each of the automobile bushing inner liner tube samples according to the first inner surface defect detection value and the second inner surface defect detection value; Inner surface defect detection qualified rate acquisition unit: used to obtain the inner surface defect detection qualified rate corresponding to the automobile bushing inner lining tube sample according to the inner surface defect detection result corresponding to each of the automobile bushing inner lining tube samples.

7. The automobile bushing inner surface defect detection system based on image processing according to claim 6 is characterized in that: The inner liner tube image segmentation model includes an inner liner tube image input layer, an inner liner tube image feature extraction layer, an inner liner tube image feature fusion layer and an inner liner tube image segmentation layer; the inner liner tube image feature extraction layer is used to extract multiple inner liner tube feature maps of the inner liner tube image; the inner liner tube image feature fusion layer is used to perform feature fusion on multiple inner liner tube feature maps to obtain inner surface color features and inner surface texture features; The inner liner tube image segmentation layer is used to segment the inner surface image according to the inner surface color features and the inner surface texture features to obtain the chamfered area image and the non-chamfered area image.

8. The automobile bushing inner surface defect detection system based on image processing according to claim 6 is characterized in that: Performing defect detection on the chamfer area image to obtain a first inner surface defect detection value includes: inputting the chamfer area image into a chamfer area defect detection model to obtain a first defect feature of the chamfer area; the first defect feature includes the number of defects, the defect type and the defect area of ​​each defect; based on the chamfer area image, obtaining an average width and a uniform width value of the chamfer area; obtaining a first defect feature value based on the first defect feature; and obtaining the first inner surface defect detection value based on the first defect feature value, the average width and the uniform width value.

9. The automobile bushing inner surface defect detection system based on image processing according to claim 6 is characterized in that: The calculation formula of the first inner surface defect detection value is: ； in, It is represented as the first inner surface defect detection value corresponding to the i-th automobile bushing inner liner tube sample; the value of i ranges from 1 to N; , and Respectively expressed as The average width, uniform width and first defect characteristic value of the corresponding chamfered area; Expressed as a set standard width; , and They are respectively expressed as the first width coefficient, the second width coefficient and the defect characteristic coefficient; Expressed as an exponential function with a natural constant as base.

10. The automobile bushing inner surface defect detection system based on image processing according to claim 6 is characterized in that: The process of obtaining the inner surface defect detection result corresponding to each automobile bushing inner liner tube sample includes: According to the first inner surface defect detection value and the second inner surface defect detection value, the inner surface comprehensive defect detection value corresponding to the automobile bushing inner liner tube sample is obtained, and the calculation formula is: ； in, It is represented by the comprehensive defect detection value of the inner surface corresponding to the i-th automobile bushing inner liner tube sample; the value of i ranges from 1 to N; and They are respectively represented as the first inner surface defect detection value and the second inner surface defect detection value corresponding to the i-th automobile bushing inner liner tube sample; and They are respectively expressed as the defect weight of the chamfered area and the defect weight of the non-chamfered area; When the first inner surface defect detection value is less than the set first defect threshold, the second inner surface defect detection value is less than the set second defect threshold, and the inner surface comprehensive defect detection value is less than the set comprehensive defect threshold, the inner surface defect detection result corresponding to the automobile bushing inner liner tube sample is that the inner surface defect detection is qualified; otherwise, it indicates that the inner surface defect detection is unqualified.

Images