Brake disc identification code recognition device based on machine vision

Abstract

The utility model discloses a brake disc identification code recognition device based on machine vision, relates to a passenger train part identification code recognition device, and solves the problems of low efficiency and easy misjudgment and missed judgment of manual recognition of brake disc manufacturing identification. A rotary motor is fixed on a piston rod of a lifting cylinder; the clamping air cylinder is fixed to a rotating shaft of the rotary motor, a piston rod of the clamping air cylinder can drive the clamp to clamp the brake disc, the lifting air cylinder can drive the brake disc to move in the axial direction of the lifting air cylinder, and the rotary motor can drive the brake disc to rotate in the circumferential direction of the lifting air cylinder. The image collecting assembly is used for collecting circumferential image information of the brake disc and uploading the circumferential image information to the upper computer, and the identification code of the identification area on the brake disc is recognized through the upper computer. The beneficial effects are that the identification speed is fast; the problems that in the prior art, time and labor are consumed in manual checking, and the identification accuracy is greatly influenced by human factors are solved, and repeated manual labor is reduced.

Description

Technical Field

[0001] The utility model relates to an identification device for the identification code of railway passenger car parts. Background Art

[0002] With the rapid development of railway transportation, the safety identification of railway passenger car brake discs has become crucial; the manufacturing identification code of railway passenger car brake discs contains important product information, such as manufacturer code, production date, product number or furnace number, etc.; the existing manual identification method has low efficiency, is prone to misjudgment and missed judgment, and is difficult to meet the high-quality and high-efficiency requirements of railway production. Content of the Utility Model

[0003] The purpose of the utility model is to solve the problems of low efficiency of manually identifying the manufacturing identification of brake discs and being prone to misjudgment and missed judgment, and a brake disc identification code identification device based on machine vision is proposed.

[0004] A brake disc identification code identification device based on machine vision according to the utility model includes a lifting cylinder, a rotary motor, a clamping cylinder, a fixture and an image acquisition component;

[0005] The rotary motor is fixed on the piston rod of the lifting cylinder;

[0006] The clamping cylinder is fixed on the rotating shaft of the rotary motor, and the piston rod of the clamping cylinder can drive the fixture to clamp the brake disc, so that the lifting cylinder can drive the brake disc to move axially along it, and the rotary motor can drive the brake disc to rotate circumferentially;

[0007] The image acquisition component is used to acquire the circumferential image information of the brake disc and upload the circumferential image information to the host computer, and the identification code of the identification area on the brake disc is identified through the host computer.

[0008] Further, the identification code identification device further includes a frame;

[0009] The frame is of a "U" - shaped structure, and the open end of the "U" - shaped structure is fixed on the ground;

[0010] The top end of the lifting cylinder is fixed on the frame;

[0011] The image acquisition component is fixed on the side wall of the frame.

[0012] Further, the identification code identification device further includes a sliding plate, two slide rails and a cross beam;

[0013] The cross beam is fixed in the middle of the frame;

[0014] The upper ends of the two slide rails are respectively fixed to the top of the frame, and the lower ends of the two slide rails are respectively fixed to the cross beam. Moreover, the two slide rails are respectively located on both sides of the lifting cylinder; meanwhile, both slide rails are parallel to the piston rod of the lifting cylinder.

[0015] The sliding plate is fixed to the bottom end of the piston rod of the lifting cylinder; and the sliding plate slides along the length direction of the two slide rails; the rotary motor is fixed to the sliding plate.

[0016] Furthermore, the identification code recognition device further includes a cylinder bottom plate.

[0017] The cylinder bottom plate is fixed to the rotating shaft of the rotary motor.

[0018] The clamping cylinder is fixed to the cylinder bottom plate.

[0019] Furthermore, the identification code recognition device further includes a light source assembly.

[0020] The light source assembly is fixed to the side wall of the frame, and the light source assembly is used to irradiate the brake disc.

[0021] Furthermore, the rotation angle of the rotating shaft of the rotary motor is: 360 degrees to 400 degrees.

[0022] Furthermore, the image acquisition component adopts a line array camera.

[0023] The utility model has the following beneficial effects compared with the prior art:

[0024] For an identification device for brake disc identification codes based on machine vision according to the present invention, a grasping component is formed by a clamping cylinder and a fixture. The grasping component is used to accurately grasp the brake disc after the brake disc arrives in place; the grasped brake disc is lifted to the image recognition position by a lifting cylinder; under the action of the rotary motor, an encoder is used to control the image acquisition component to scan the brake disc, obtaining a complete circumferential image of the brake disc, which takes a total of 10 s and has a fast recognition speed; it solves the problems in the prior art that manual inspection is time-consuming and laborious and the recognition accuracy is greatly affected by human factors, and reduces the manual repetitive labor.

[0025] The present invention relates to the field of identification of identification codes for railway passenger car parts. Brief Description of the Drawings

[0026] Figure 1 It is a schematic diagram of the overall structure of an identification device for brake disc identification codes based on machine vision described in the first specific embodiment;

[0027] Figure 2 It is a physical diagram of the brake disc in the first specific embodiment.

[0028] Among them, 1 is a frame; 2 is a skateboard; 3 is a lifting cylinder; 4 is a slewing motor; 5 is a clamping cylinder; 6 is a cylinder base plate; 7 is a fixture; 8 is a light source assembly; 9 is an image acquisition assembly; 10 is a brake disc; 11 is a slide rail; 12 is a cross beam; 13 is an identification area. Detailed implementation manner

[0029] Detailed implementation manner 1. Combining Figures 1 to 2 To illustrate this implementation manner, for an identification device for brake disc identification codes based on machine vision described in this implementation manner, the identification device for identification codes includes a lifting cylinder 3, a slewing motor 4, a clamping cylinder 5, a fixture 7, and an image acquisition assembly 9;

[0030] The slewing motor 4 is fixed on the piston rod of the lifting cylinder 3;

[0031] The clamping cylinder 5 is fixed on the rotating shaft of the slewing motor 4, and the piston rod of the clamping cylinder 5 can drive the fixture 7 to clamp the brake disc 10, so that the lifting cylinder 3 can drive the brake disc 10 to move along its axial direction, and the slewing motor 4 can drive the brake disc 10 to rotate along its circumferential direction;

[0032] The image acquisition assembly 9 is used to collect the circumferential image information of the brake disc 10, and upload the circumferential image information to the host computer, and the identification code of the identification area 13 on the brake disc 10 is identified through the host computer.

[0033] In this implementation manner, the image acquisition assembly 9 is controlled by an encoder; the clamping cylinder 5 and the fixture 7 form a grasping assembly, and the grasping assembly is used to accurately grasp and lift the brake disc 10 to the image detection position after the brake disc 10 reaches directly below the fixture 7; the fixture 7 is composed of two grippers, and the distance between the two grippers is controlled by the clamping cylinder 5. When the two grippers of the fixture 7 approach, the two grippers extend into the central hole of the brake disc 10. During the process of the piston rod of the clamping cylinder 5 extending, the distance between the two grippers of the fixture 7 gradually becomes larger, so that the two grippers are in close contact with the inner wall of the central hole of the brake disc 10, and then the grasping assembly completes the radial grasping of the brake disc 10; the rotating shaft of the slewing motor 4 drives the grasping assembly and the brake disc 10 to rotate, so that the image acquisition assembly 9 collects the complete circumferential image information of the brake disc 10;

[0034] The working principle of the identification device for identification codes is as follows:

[0035] When the piston rod of the clamping cylinder 5 moves telescopically, it drives the fixture 7 to clamp or loosen the brake disc 10. After the fixture 7 clamps the brake disc, the lifting cylinder 3 drives the rotary motor 4 to move up and down, and then raises the brake disc 10 clamped by the fixture 7 to the image detection position. After the brake disc 10 is raised to the image detection position, the rotary reduction motor 4 operates to drive the clamped and raised brake disc 10 to rotate according to the rotation angle, and the image acquisition component 9 acquires the complete circumferential image of the brake disc 10 through the rotation of the brake disc 10.

[0036] The host computer first intercepts along the width direction of the circumferential image information obtained by the image acquisition component 9 to obtain a strip image with the entire identification area 13, and then equally divides the strip image. The image with the identification area 13 is selected from the equally divided images as the pre-annotated image; the pre-annotated image is subjected to positioning annotation and OCR character annotation to obtain the image to be recognized; the obtained image to be recognized is input into the character recognition algorithm model, and the identification code is output.

[0037] The host computer intercepts the width of the circumferential image information obtained by the image acquisition component 9 and equally divides the image after width interception to remove the parts of the pictures that do not contain the identification code, thereby improving the recognition speed of the picture information by the host computer and the recognition accuracy. In step four, the character recognition algorithm model uses a convolutional neural network CNN, which can better process high-dimensional data, extract important features in the image through local connection, weight sharing, translational invariance and sparse connection design, and trains the model with a large amount of labeled data to ensure the accuracy of recognition. The recognition method uses width interception and equally divides the image after width interception to accurately obtain the ROI area, remove the pictures without the identification area, and perform enhancement and noise reduction processing on the image, reducing the training and inference time, accelerating the efficiency, reducing misjudgment, improving the accuracy, and the recognition accuracy is greater than 95%, meeting the production requirements of industrial enterprises; the purpose of performing positioning annotation on the pre-annotated image in step three is to prevent the order of the identification codes contained in the equally divided pictures from being confused, and thus improve the recognition accuracy through positioning annotation. The specific formula for equally dividing the strip image is:

[0038]

[0039] where s is the interception length; i is the interception times.

[0040] If the width of the strip image is \(w\) and the total width of the circumferential image information is \(n\), then \(n>w\); and the width \(w\) of the strip image is greater than the width of the identification area 13. At the same time, the closer the width \(w\) of the strip image is to the width of the identification area 13, the faster the host computer processes the image to be recognized in the subsequent process; the intercepted length \(s\) is less than the length of the identification area 13, but must be greater than the width of a single character; among them, the 11 characters contained in the identification code are arranged side by side at equal intervals, and the length of the identification area 13 is equal to the sum of the widths of 11 characters and the intervals between 10 adjacent characters.

[0041] After forming the pre-labeled image, first perform contrast adjustment, denoising, and geometric transformation on the pre-labeled image in sequence, and then perform positioning annotation and OCR character annotation on the pre-labeled image; improve the quality of the pre-labeled image and increase the diversity of data through contrast adjustment, denoising, and geometric transformation.

[0042] The character recognition algorithm model is obtained through training with a neural network model. The specific training method is as follows:

[0043] Step 1: Collect the circumferential image information of multiple brake discs 10 to obtain a data set;

[0044] Step 2: Intercept the data set obtained in Step 1 with equal width and equal step length, and select the images with brake disc manufacturing identification from the intercepted images to obtain a preprocessed data set;

[0045] Step 3: Preprocess the preprocessed data set obtained in Step 2 to obtain a pre-labeled data set; the preprocessing refers to performing contrast adjustment, denoising, and geometric transformation on the preprocessed data set in sequence;

[0046] Step 4: Perform positioning annotation and OCR character annotation on the pre-labeled data set obtained in Step 3 to obtain a pre-training data set;

[0047] Step 5: Input the pre-training data set obtained in Step 4 into the neural network model for training, optimize the parameters of the neural network model, and form a character recognition algorithm model.

[0048] Training and optimizing the parameters of the neural network model with a large amount of labeled data in the pre-training data set ensures the accuracy of the character recognition algorithm model.

[0049] Specific Embodiment 2: This embodiment further limits the brake disc identification code recognition device described in Specific Embodiment 1. In this embodiment, the identification code recognition device further includes a frame 1;

[0050] The frame 1 is a "U" - shaped structure, and the open end of the "U" - shaped structure is fixed on the ground;

[0051] The top end of the lifting cylinder 3 is fixed to the frame 1;

[0052] The image acquisition component 9 is fixed to the side wall of the frame 1.

[0053] In this embodiment, the frame 1 is installed on the automatic conveying line of the brake disc 10; the frame 1 is used to fix the lifting cylinder 3 and the image acquisition component 9.

[0054] Specific Embodiment Three: This embodiment further limits the brake disc identification code recognition device based on machine vision described in Specific Embodiment Two. In this embodiment, the identification code recognition device further includes a sliding plate 2, two sliding rails 11 and a cross beam 12;

[0055] The cross beam 12 is fixed to the middle of the frame 1;

[0056] The upper ends of the two sliding rails 11 are respectively fixed to the top of the frame 1, the lower ends of the two sliding rails 11 are respectively fixed to the cross beam 12, and the two sliding rails 11 are respectively located on both sides of the lifting cylinder 3; at the same time, both sliding rails 11 are parallel to the piston rod of the lifting cylinder 3;

[0057] The sliding plate 2 is fixed to the bottom end of the piston rod of the lifting cylinder 3; and the sliding plate 2 slides along the length direction of the two sliding rails 11; the rotary motor 4 is fixed to the sliding plate 2.

[0058] In this embodiment, through the arrangement of the sliding plate 2 and the sliding rails 11, the lifting cylinder 3 will not rotate independently during the telescopic process, avoiding the rotation angle of the rotary motor 4 being out of control, so that the image acquisition component 9 can acquire a complete circumferential image of the brake disc 10; and through the arrangement of the sliding plate 2, the structural stability of the entire recognition device can be strengthened, and the possibility of damage to the lifting cylinder 3 during the lifting process can be reduced.

[0059] Specific Embodiment Four: This embodiment further limits the brake disc identification code recognition device based on machine vision described in Specific Embodiment One or Two. In this embodiment, the identification code recognition device further includes a cylinder bottom plate 6;

[0060] The cylinder bottom plate 6 is fixed to the rotating shaft of the rotary motor 4;

[0061] The clamping cylinder 5 is fixed to the cylinder bottom plate 6.

[0062] In this embodiment, by adding the cylinder bottom plate 6, it is convenient to fix the outer wall of the clamping cylinder 5 to the rotating shaft of the rotary motor 4, and the cylinder bottom plate 6 can increase the firmness of the outer wall of the clamping cylinder 5.

[0063] Specific Embodiment V. This embodiment further limits a brake disc identification code recognition device based on machine vision described in Specific Embodiment I. In this embodiment, the identification code recognition device further includes a light source assembly 8;

[0064] The light source assembly 8 is fixed on the side wall of the frame 1, and the light source assembly 8 is used to irradiate the brake disc 10.

[0065] In this embodiment, the light source assembly 8 is arranged close to the image acquisition assembly 9. By adding the light source assembly 8, it is used to provide sufficient illumination for the image acquisition assembly 9 to ensure that the manufacturing marks of the brake disc 10 are clearly visible.

[0066] Specific Embodiment VI. This embodiment further limits a brake disc identification code recognition device based on machine vision described in Specific Embodiment I. In this embodiment, the rotation angle of the rotating shaft of the rotary motor 4 is: 360 degrees to 400 degrees.

[0067] In this embodiment, after multiple identification verifications, when the rotation angle of the rotating shaft of the rotary motor 4 is between 360 degrees and 400 degrees, on the basis of ensuring the acquisition of complete circumferential image information of the brake disc 10, the acquisition time is the shortest; too large a rotation angle wastes the acquisition time and affects the recognition efficiency.

[0068] Specific Embodiment VII. This embodiment further limits a brake disc identification code recognition device based on machine vision described in Specific Embodiment I. In this embodiment, the image acquisition assembly 9 uses a line array camera.

[0069] In this embodiment, the image acquisition assembly 9 acquires images in the circumferential direction of the brake disc 10. When the image acquisition assembly 9 is a line array camera, the acquisition field of view is in the shape of a slender strip, which can meet the rotation angle of the rotating shaft of the rotary motor 4, and the line array camera can meet the extremely large field of view of the recognition device and the circumferential image information after acquisition has extremely high precision.

[0070] The above is only the specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, and all should be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims.

Claims

1. An identification device for brake disc identification codes based on machine vision, characterized in that, The identification code recognition device includes a lifting cylinder (3), a rotary motor (4), a clamping cylinder (5), a fixture (7) and an image acquisition component (9); The rotary motor (4) is fixed on the piston rod of the lifting cylinder (3); The clamping cylinder (5) is fixed on the rotating shaft of the rotary motor (4), and the piston rod of the clamping cylinder (5) can drive the fixture (7) to clamp the brake disc (10), so that the lifting cylinder (3) can drive the brake disc (10) to move along its axial direction, and the rotary motor (4) can drive the brake disc (10) to rotate along its circumferential direction; The image acquisition component (9) is used to acquire the circumferential image information of the brake disc (10) and upload the circumferential image information to the host computer, and the identification code of the identification area (13) on the brake disc (10) is recognized by the host computer.

2. The identification device for the brake disc identification code based on machine vision according to claim 1, characterized in that, The identification code recognition device further includes a frame (1); The frame (1) is of a "U" - shaped structure, and the open end of the "U" - shaped structure is fixed on the ground; The top end of the lifting cylinder (3) is fixed on the frame (1); The image acquisition component (9) is fixed on the side wall of the frame (1).

3. The identification device for the brake disc identification code based on machine vision according to claim 2, characterized in that The identification code recognition device further includes a slide plate (2), two slide rails (11) and a cross beam (12); The cross beam (12) is fixed in the middle of the frame (1); The upper ends of the two slide rails (11) are respectively fixed on the top of the frame (1), the lower ends of the two slide rails (11) are respectively fixed on the cross beam (12), and the two slide rails (11) are respectively located on both sides of the lifting cylinder (3); at the same time, both slide rails (11) are parallel to the piston rod of the lifting cylinder (3); The slide plate (2) is fixed at the bottom end of the piston rod of the lifting cylinder (3); and the slide plate (2) slides along the length direction of the two slide rails (11); The rotary motor (4) is fixed on the slide plate (2).

4. The identification device for the brake disc identification code based on machine vision according to claim 1, characterized in that The identification code recognition device further includes a cylinder bottom plate (6); The cylinder bottom plate (6) is fixed on the rotating shaft of the rotary motor (4); The clamping cylinder (5) is fixed on the cylinder bottom plate (6).

5. The identification device for the brake disc identification code based on machine vision according to claim 1, wherein, The identification code recognition device further includes a light source component (8); The light source component (8) is fixed on the side wall of the frame (1), and the light source component (8) is used to irradiate the brake disc (10).

6. The identification device for the brake disc identification code based on machine vision according to claim 1, characterized in that The rotation angle of the rotating shaft of the rotary motor (4) is: 360 degrees to 400 degrees.

7. The identification device for the brake disc identification code based on machine vision according to claim 1, wherein The image acquisition component (9) uses a line - array camera.

Images