Online automatic detection system and method for surface defects of shaft parts

Abstract

The invention discloses an online automatic detection system and method for surface defects of shaft parts. According to the invention, a lighting mode of combining a coaxiality light source and an arc light source is provided, and the surface defects of the shaft parts are comprehensively and clearly identified by combining the rotation of the detected shaft with the linear movement of the visual detection module; comprising a base, a clamping linear guide rail, a live center, a detected shaft, a three-jaw chuck, a rotating motor, a detection linear guide rail and a visual detection module. According to the invention, automatic defect detection and feedback of the shaft parts are realized, and the method has the characteristics of high detection precision, high detection efficiency, wide detection range and quantitative recognition. According to the method, the surface defects of various shaft parts can be comprehensively and clearly identified, the defect detection of dynamic images is realized by adopting a deep learning method, the defect detection problem of the complex surfaces of the shaft parts is solved, the defects of the surfaces of stepped table surfaces, flat grooves, threads and the like of the shaft parts can be detected, and the method is suitable for online detection.

Description

Technical Field

[0001] The present invention belongs to the field of machine vision inspection, and relates to an on-line automatic inspection system and method for surface defects of shaft parts. Background Art

[0002] As the main core component of a rotating mechanism, the machining quality of shaft parts directly affects the assembly quality of the rotating mechanism. At present, with the improvement of production technology, the traditional manual visual inspection method for defects cannot keep up with the productivity of the existing technology, and there are problems such as large human observation errors, easy fatigue of long-term visual inspection workers, low efficiency, and high management costs. In view of the above problems, it is necessary to use an automatic means to detect the surface defects of machined shafts and achieve closed-loop feedback to ensure the quality of machined shafts.

[0003] Therefore, the present invention proposes an on-line automatic inspection system and method for surface defects of shaft parts. The system proposes a lighting method combining a coaxiality light source and an arc light source, and a detection method in which the rotation of the shaft to be detected is combined with the linear movement of the shaft defect detection module, so as to comprehensively and clearly identify the surface defects of shaft parts. The deep learning method is used to realize the defect detection of dynamic images, solve the problem of defect detection on the complex surface of shaft parts, can detect the defects on the surfaces such as the stepped table surface, flat groove, and thread of shaft parts, avoid the situation of processing defective products, can greatly improve the processing quality, and can adapt to on-line inspection. Summary of the Invention

[0004] The purpose of the present invention is to provide an on-line automatic inspection system and method for surface defects of shaft parts to solve the problem of on-line inspection of surface defects of shaft parts in view of the deficiencies of the prior art. The present invention adopts the following solutions.

[0005] The technical solutions adopted by the present invention to solve its technical problems are as follows:

[0006] An on-line automatic inspection system for surface defects of shaft parts proposes a lighting method combining a coaxiality light source and an arc light source, and the rotation of the shaft to be detected is combined with the linear movement of the vision inspection module to comprehensively and clearly identify the surface defects of shaft parts; it includes: a base (S0), a clamping linear guide (S1), a live center (S2), a shaft to be detected (S3), a three-jaw chuck (S4), a rotating motor (S5), a detection linear guide (S6), and a vision inspection module (S7);

[0007] The rotating motor (S5) is rigidly fixed on the base (S0). A three-jaw chuck (S4) is installed on the top of the rotating motor (S5). One end of the shaft to be detected (S3) is clamped by the three-jaw chuck (S4), and the other end of the shaft to be detected (S3) is fixed by being supported by a live center (S2). The live center (S2) is rigidly fixed on the clamping linear guide (S1) and can move axially along with the clamping linear guide (S1) to facilitate clamping shafts to be detected with different lengths. The clamping linear guide (S1) is rigidly fixed on the base (S0); as Figure 1 shown, wherein the detection linear guide (S6) is rigidly fixed on the base (S0). A vision detection module (S7) is rigidly fixed on the slider of the detection linear guide (S6). The vision detection module (S7) can move up and down horizontally along with the detection linear guide (S6) to detect surface defects of the shaft at different positions.

[0008] Furthermore, the vision detection module (S7) includes: an imaging camera (S8), a rotary stage (S9), an imaging adjustment block I (S10), a coaxial light source (S11), an arc light source (S12), an imaging adjustment block III (S13), a detection module mounting plate (S14), an imaging adjustment block II (S15), and a detection module base plate (S16). The detection module base plate (S16) is fixed on the slider of the detection linear guide (S6) through the detection module mounting plate (S14). The detection module base plate (S16) can drive the entire vision detection module (S7) to move up and down horizontally along with the detection linear guide (S6) to detect surface defects of the shafts to be detected at different positions; wherein the imaging camera (S8) is fixed on the detection module base plate (S16) through the rotary stage (S9) and the imaging adjustment block I (S10), and the imaging camera (S8) can be adjusted along with the rotary stage (S9) and the imaging adjustment block I (S10); wherein the coaxial light source (S11) is fixed on the detection module base plate (S16) through the imaging adjustment block II (S15), and the coaxial light source (S11) can perform two-dimensional adjustments of front-back translation and pitching rotation along with the imaging adjustment block II (S15). Two arc light sources (S12) are symmetrically fixed on the detection module base plate (S16) through the imaging adjustment block III (S13), and the arc light source (S12) can perform two-dimensional adjustments of translation and rotation along with the imaging adjustment block III (S13) to clearly capture images of different shafts.

[0009] Furthermore, specifically, the shooting angle of the imaging camera (S8) is adjusted up and down by the rotary stage (S9), and the position of the imaging camera (S8) can be adjusted front-back and up-down by adjusting the imaging adjustment block I (S10) according to actual requirements to completely capture images of different shafts.

[0010] Furthermore, the system uses a combined lighting method with a coaxial light source (S11) and two arc light sources (S12) to achieve defect detection and imaging of the on-axis thread and the shaft light surface. The optical path formed by the imaging camera (S8), the coaxial light source (S11), and the two arc light sources (S12) is as follows: The three outgoing light beams generated by the coaxial light source (S11) and the two arc light sources (S12) are reflected by the surface of the shaft to be detected (S3) and then enter the imaging camera (S8) to generate an image.

[0011] Furthermore, when performing surface defect detection on the shaft to be detected (S3), it is necessary to rotate and detect simultaneously. The shaft to be detected (S3) rotates under the clamping of the three-jaw chuck (S4) following the rotation motor (S5). After the detection module (S7) captures an image of the shaft to be detected (S3) rotating one week, it performs a linear translation upward, and continues to capture an image of the shaft to be detected (S3) rotating one week until the shaft to be detected (S3) can be completely detected.

[0012] The beneficial effects of the present invention are as follows:

[0013] The present invention proposes an on-line automatic detection system and method for surface defects of shaft parts, which realizes automatic defect detection and feedback of shaft parts, and has the characteristics of high detection accuracy, high detection efficiency, wide detection range, and quantitative recognition. The system proposes a lighting method combining a coaxiality light source and an arc light source, and a detection method of combining the rotation of the shaft to be detected with the linear movement of the shaft defect detection module, which comprehensively and clearly identifies the surface defects of shaft parts, solves the problem of defect detection on the complex surface of shaft parts, can detect the defects on the surfaces such as the stepped table surface, flat groove, and thread of shaft parts, avoids the situation of processing defective products, can greatly improve the processing quality, and can adapt to on-line detection.

[0014] The system structure of the present invention is a part of the entire large system and is the part for surface defect detection of shaft parts on the production line. There are also series of system structures for automatic detection subsequently. Description of the Drawings

[0015] Figure 1 It is a diagram of the on-line detection system for surface defects of shaft parts;

[0016] Figure 2 It is a diagram of the clamping device of the shaft to be detected;

[0017] Figure 3 It is a structure diagram of the detection module;

[0018] Figure 4 It is a structure diagram of the shaft to be detected;

[0019] Figure 5 It is an optical path diagram of the detection module;

[0020] Figure 6 It is an imaging image;

[0021] Figure 7 is the detection result image;

[0022] Detailed implementation description

[0023] The present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0024] As Figure 1 shown, an on-line automatic detection system for surface defects of shaft parts includes: a base (S0), a clamping linear guide (S1), a movable center (S2), a shaft to be detected (S3), a three-jaw chuck (S4), a rotating motor (S5), a detection linear guide (S6), and a vision detection module (S7).

[0025] As Figure 2 shown, wherein the rotating motor (S5) is rigidly fixed on the base (S0), a three-jaw chuck (S4) is installed on the top of the rotating motor (S5), one end of the shaft to be detected (S3) is clamped by the three-jaw chuck (S4), the other end of the shaft to be detected (S3) is fixed by being supported by the movable center (S2), the movable center (S2) is rigidly fixed on the clamping linear guide (S1) and can move axially along with the clamping linear guide (S1) to facilitate clamping of shafts to be detected with different lengths, and the clamping linear guide (S1) is rigidly fixed on the base (S0). As Figure 1 shown, wherein the detection linear guide (S6) is rigidly fixed on the base (S0), a vision detection module (S7) is rigidly fixed on the sliding block of the detection linear guide (S6), and the vision detection module (S7) can move up and down horizontally along with the detection linear guide (S6) to detect surface defects of the shaft at different positions.

[0026] As Figure 3 shown, the vision detection module (S7) includes: an imaging camera (S8), a rotating stage (S9), an imaging adjustment block I (S10), a coaxial light source (S11), an arc light source (S12), an imaging adjustment block III (S13), a detection module mounting plate (S14), an imaging adjustment block II (S15), and a detection module bottom plate (S16).

[0027] Among them, the bottom plate (S16) of the detection module is fixed on the slider of the detection linear guide (S6) through the installation plate (S14) of the detection module. The bottom plate (S16) of the detection module can drive the overall vision detection module (S7) to move up and down linearly along with the detection linear guide (S6) to detect the surface defects of the detected shaft at different positions. Among them, the imaging camera (S8) is fixed on the bottom plate (S16) of the detection module through the rotary stage (S9) and the imaging adjustment block I (S10). The imaging camera (S8) can be adjusted along with the rotary stage (S9) and the imaging adjustment block I (S10). Specifically, the shooting angle of the imaging camera (S8) is adjusted up and down through the rotary stage (S9), and the position of the imaging camera (S8) is further adjusted back and forth, up and down by adjusting the imaging adjustment block I (S10) according to actual needs to completely capture the images of different shafts. Among them, the coaxial light source (S11) is fixed on the bottom plate (S16) of the detection module through the imaging adjustment block II (S15). The coaxial light source (S11) can be adjusted two-dimensionally in the front-back translation and pitch rotation along with the imaging adjustment block II (S15). The two arc-shaped light sources (S12) are symmetrically fixed on the bottom plate (S16) of the detection module through the imaging adjustment block III (S13). The arc-shaped light source (S12) can be adjusted two-dimensionally in the translation and rotation along with the imaging adjustment block III (S13) to clearly capture the images of different shafts.

[0028] The present invention uses a combined lighting of a coaxial light source (S11) and two arc-shaped light sources (S12) to achieve defect detection and imaging of the thread on the shaft and the shaft light surface. The optical paths formed by the imaging camera (S8), the coaxial light source (S11), and the two arc-shaped light sources (S12) are as Figure 5 shown. The three outgoing light beams generated by the coaxial light source (S11) and the two arc-shaped light sources (S12) are reflected by the surface of the detected shaft (S3) and then enter the imaging camera (S8) to generate an image.

[0029] When the present invention performs surface defect detection on the detected shaft (S3), it needs to be detected while rotating. The detected shaft (S3) rotates following the rotation motor (S5) under the clamping of the three-jaw chuck (S4). After the detection module (S7) captures the image of the detected shaft (S3) rotating one week, it makes a linear translation upward and continues to capture the image of the detected shaft (S3) rotating one week until the detected shaft (S3) can be completely detected.

[0030] As Figure 4 shown, the present invention also provides an on-line automatic detection method for surface defects of shaft parts. The detected shaft (S3) includes a shaft thread section (S17) and a shaft body section (S18). The present invention uses a deep learning method to achieve defect detection of the shaft thread section (S17) and the shaft body section (S18) of the dynamic image. The specific steps are as follows:

[0031] Step 1: Read the image of the shaft to be detected (S3), and obtain the image data of the shaft to be detected (S3) in real time from the camera or image acquisition device of the detection system;

[0032] Step 2: Image preprocessing, including operations such as denoising, brightness adjustment, and contrast enhancement, to improve the recognizability of defect features and extract the ROI regions of the shaft thread section and the shaft body section to reduce the calculation amount;

[0033] Step 3: Image region segmentation, using the Otsu threshold segmentation and Canny edge detection algorithms to separate the shaft thread section and the shaft body section, and locate the key feature regions of the shaft parts, such as threads, notches, etc., for subsequent targeted defect detection;

[0034] Step 4: Image annotation, annotate various types of defects, and divide the image dataset into a training set, a validation set, and a test set.

[0035] Step 5: Deep learning model training, select the ResNet50-Rev convolutional neural network, and train the model with the labeled training set, where the label indicates whether there are defects in the shaft parts and marks the type and location of the defects. Optimize the model parameters through repeated iterations until the performance of the model on the validation set reaches the expected goal;

[0036] Step 6: Defect detection, input the image read in real time into the trained deep learning model, and the model will output the detection results of the defects, including the presence or absence of defects, type, location, and possible severity.

[0037] Step 7: Feedback and optimization, according to the feedback in actual applications, retrain and optimize the deep learning model to improve the detection accuracy and robustness, and regularly update the training dataset to include more types of defect samples to enhance the generalization ability of the model. Embodiment

[0038] Using a camera with 2448×2048 pixels, paired with a lens with a focal length of 25 mm, imaging in a field of view of 150 mm×150 mm can be achieved, and the imaging resolution can reach 73μm for defect detection. Using the above imaging, imaging was performed on a shaft with a length of 100 mm, and the imaging image is as Figure 6 shown, Figure 6 which contains defects in the thread section and the shaft body section. Through the deep learning method, rapid detection of dynamic images was achieved, and the detection results are as Figure 7 shown.

[0039] Figure 7 The detected defect results are shown in the following table:

[0040]

[0041] Finally, it should be noted that the purpose of disclosing the embodiments is to help further understand the present invention. However, those skilled in the art can understand that various substitutions and modifications are possible without departing from the spirit and scope of the present invention and the appended claims. Therefore, the present invention should not be limited to the content disclosed in the embodiments, and the scope of protection claimed by the present invention shall be subject to the scope defined by the claims.

Claims

1. An on-line automatic inspection system for surface defects of shaft parts, characterized in that A lighting method combining a coaxial light source and an arc light source is proposed. The rotation of the inspected shaft is combined with the linear movement of the visual inspection module to comprehensively and clearly identify the surface defects of shaft parts; including: base (S0), clamping linear guide (S1), movable center (S2), inspected shaft (S3), three-jaw chuck (S4), rotating motor (S5), inspection linear guide (S6), and visual inspection module (S7); The rotating motor (S5) is rigidly fixed on the base (S0), and a three-jaw chuck (S4) is installed on the top of the rotating motor (S5). The three-jaw chuck (S4) clamps one end of the detected shaft (S3), and the other end of the detected shaft (S3) is supported and fixed by a movable top (S2). The movable top (S2) is rigidly fixed on the clamping linear guide (S1) and can move axially with the clamping linear guide (S1) to clamp the detected shafts of different lengths. The clamping linear guide (S1) is rigidly fixed on the base (S0); as shown in Figure 1, the detection linear guide (S6) is rigidly fixed on the base (S0), and a visual detection module (S7) is rigidly fixed on the sliding block of the detection linear guide (S6). The visual detection module (S7) can move up and down with the detection linear guide (S6) to detect shaft surface defects at different positions.

2. The on-line automatic detection system for surface defects of shaft parts according to claim 1, characterized in that The visual inspection module (S7) includes: an imaging camera (S8), a rotating slide (S9), an imaging adjustment block I (S10), a coaxial light source (S11), an arc light source (S12), and an imaging adjustment block III (S13), a detection module mounting plate (S14), an imaging adjustment block II (S15), and a detection module base plate (S16), wherein the detection module base plate (S16) is fixed on a slider of a detection linear guide rail (S6) through the detection module mounting plate (S14), and the detection module base plate (S16) can drive the overall visual detection module (S7) to move up and down along with the detection linear guide rail (S6) so as to detect surface defects of the detected shaft at different positions; wherein the imaging camera (S8) is fixed on the detection module base plate (S16) through a rotating slide (S9) and an imaging adjustment block I (S10), and the imaging camera (S8) can move along with the rotating slide (S9) and the imaging adjustment block I (S10) is adjusted; wherein the coaxial light source (S11) is fixed on the detection module bottom plate (S16) through the imaging adjustment block II (S15), and the coaxial light source (S11) can be adjusted in two dimensions by front and rear translation and pitch rotation along with the imaging adjustment block II (S15); two arc light sources (S12) are symmetrically fixed on the detection module bottom plate (S16) through the imaging adjustment block III (S13), and the arc light source (S12) can be adjusted in two dimensions by translation and rotation along with the imaging adjustment block III (S13), so as to clearly capture the images of different axes.

3. An on-line automatic inspection system for surface defects of shaft parts according to claim 2, characterized in that Specifically, the shooting angle of the imaging camera (S8) is adjusted up and down by rotating the slide (S9), and the imaging adjustment block I (S10) is adjusted according to actual needs to adjust the position of the imaging camera (S8) forward, backward, upward and downward, so as to completely capture images of different axes.

4. An on-line automatic inspection system for surface defects of shaft parts according to claim 2, characterized in that The system uses a combined lighting method with a coaxial light source (S11) and two arc light sources (S12) to achieve defect detection and imaging of the on-axis thread and the shaft light surface. The optical path formed by the imaging camera (S8), the coaxial light source (S11), and the two arc light sources (S12) is as follows: The three outgoing light beams generated by the coaxial light source (S11) and the two arc light sources (S12) are reflected by the surface of the shaft to be detected (S3) and then enter the imaging camera (S8) to generate an image.

5. The on-line automatic detection system for surface defects of shaft parts according to claim 2, characterized in that When performing surface defect detection on the shaft to be detected (S3), it is necessary to rotate and detect simultaneously. The shaft to be detected (S3) rotates under the clamping of the three-jaw chuck (S4) following the rotation motor (S5). After the detection module (S7) takes images of the shaft to be detected (S3) rotating one week, it makes a linear translation upward and continues to take images of the shaft to be detected (S3) rotating one week until the shaft to be detected (S3) can be completely detected.

6. The on-line automatic inspection system for surface defects of shaft parts according to claim 2, characterized in that The shaft to be detected (S3) includes a shaft thread section (S17) and a shaft body section (S18).

7. An on-line automatic detection method for surface defects of shaft parts, characterized in that The shaft to be detected (S3) includes a shaft thread section (S17) and a shaft body section (S18). This method uses deep learning to achieve defect detection of the shaft thread section (S17) and the shaft body section (S18) in dynamic images. The specific steps are as follows: Step 1: Read the image of the shaft to be detected (S3) and obtain the image data of the shaft to be detected (S3) in real time from the camera or image acquisition device of the detection system. Step 2: Image preprocessing, including operations such as denoising, brightness adjustment, and contrast enhancement, to improve the recognizability of defect features and extract the ROI regions of the shaft thread section and the shaft body section to reduce the calculation amount. Step 3: Image region segmentation, using the Otsu threshold segmentation and Canny edge detection algorithms to separate the shaft thread section and the shaft body section, and locate the key feature regions of the shaft parts, so as to detect defects specifically in the follow-up. The key feature regions include threads and notches. Step 4: Image annotation, annotate various types of defects, and divide the labeled image dataset into a training set, a validation set, and a test set. Step 5: Deep learning model training, select the ResNet50-Rev convolutional neural network, and train the model with the labeled training set. The label indicates whether there are defects in the shaft parts and marks the presence or absence, type, location, and possible severity of the defects. By repeatedly iterating to optimize the model parameters until the performance of the model on the validation set reaches the expected goal. Step 6: Defect detection, input the real-time read image into the trained deep learning model, and the model will output the detection results of the defects, including the presence or absence, type, location, and possible severity of the defects. Step 7: Feedback and optimization, according to the feedback in actual applications, retrain and optimize the deep learning model to improve the detection accuracy and robustness, and regularly update the training dataset to include more types of defect samples to enhance the generalization ability of the model.

Images