Vision-based Detection Method and Detection System for Pillow Spring Drop of Railway Vehicle Bogie

Abstract

The present invention provides a vision-based detection method and detection system for the dropping of the bolster spring of a railway vehicle bogie, which relates to the technical field of bogies. The present invention is composed of an image acquisition module, an industrial computer, a vision system module, a communication module, a PLC controller, etc. The present invention locates and divides different regions of an image with bolster spring features through the vision system module, and detects and identifies whether the pose state of the bolster spring is correct, facilitating the subsequent correct installation of the bolster spring. The present invention utilizes an object detection algorithm based on deep learning to perform region-based recognition on an image with bolster spring features, timely detecting the bolster springs with installation failures, reducing safety risks, and improving the installation efficiency of the bolster springs.

Description

Technical Field

[0001] The present invention relates to the technical field of bogies, and particularly to a method and a detection system for detecting the dropping of bolster springs of railway vehicle bogies based on vision. Background Art

[0002] Railway freight cars, as key transportation equipment for the railway to improve transportation efficiency, have been running on the railway under long-term load. To ensure the good operation state of the vehicles, the railway system has set up departments such as vehicle repair factories, vehicle depots, depot repair shops, and train inspection stations for regular maintenance and daily repair of freight cars. As a key component of railway freight cars, the overhaul of bogies is particularly important.

[0003] Currently, when overhauling the vibration damping device of a bogie, to reduce labor intensity and improve work efficiency, a bolster spring and inclined wedge installation robot and a bogie vibration damping device installation and disassembly system are usually adopted. By adding an end manipulator to a six-axis robot, the installation and disassembly of the inclined wedge and bolster spring on the bogie are realized. During the installation process of the bolster spring, the transfer manipulator in the bogie vibration damping device installation and disassembly system assists in clamping the bolster spring. After clamping the bolster spring from the tray on the conveyor belt, its posture is adjusted and placed on the transfer platform to facilitate the subsequent installation by the installation manipulator. However, due to problems such as water droplets, paint, and reflection on the surface of the bolster spring, problems such as deviation in the installation position of the bolster spring, tipping, and dropping of the bolster spring may occur during the installation and disassembly process of the bolster spring, and timely correction is required. However, in the existing technology, there is a lack of an intelligent recognition and positioning method for the rotation angle of the bolster spring pose and a method for judging whether the installation is correct. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to propose a method and a detection system for detecting the dropping of bolster springs of railway vehicle bogies based on vision to solve the above problems.

[0005] Based on the above purpose, the present invention provides a method for detecting the dropping of bolster springs of railway vehicle bogies based on vision, which is carried out according to the following steps:

[0006] S1. Obtain an image with bolster spring features.

[0007] S2. Use a deep learning object detection algorithm to perform bolster spring detection and side frame base detection on the image with bolster spring features respectively to obtain a bolster spring target bounding box and a side frame base bounding box.

[0008] S3. Divide the area of the image with bolster spring features into a first area and a second area; the first area includes all side frame base bounding boxes where no bolster spring has been installed, and the second area is the correct installation position of the newly installed bolster spring.

[0009] S4. Determine the tipping of the bolster spring based on the first region and the target bounding box of the bolster spring. If tipping exists, send an error signal; otherwise, send a correct signal.

[0010] S5. Determine the dropping of the bolster spring based on the second region and the target bounding box of the bolster spring. If dropping exists, send an error signal; otherwise, send a correct signal.

[0011] Furthermore, the image with the characteristics of the bolster spring is collected by an industrial camera, and the focal length of the industrial camera is set to 6 mm.

[0012] Furthermore, the specific steps of S4 are as follows: Perform target detection on the first region. If there is a target bounding box of the bolster spring in the first region, it is determined that tipping exists, an error signal is sent, and the installation of the bolster spring is interrupted; otherwise, a correct signal is sent and the installation of the bolster spring continues.

[0013] Furthermore, the specific steps of S5 are as follows: Perform target detection on the second region. If there is a target bounding box of the bolster spring in the second region, a correct signal is sent and the installation of the bolster spring continues; otherwise, it is determined that tipping exists, an error signal is sent, and the installation of the bolster spring is interrupted.

[0014] The vision-based bolster spring dropping detection system for railway vehicle bogies includes: an image acquisition module, an industrial control computer, a vision system module, a communication module, and a PLC controller; the image acquisition module includes an industrial camera and a light source, and the image acquisition module is used to collect images with the characteristics of the bolster spring; the industrial control computer is provided with a vision system module, and the vision system module is used to determine the tipping of the bolster spring and the dropping of the bolster spring based on the images with the characteristics of the bolster spring; the communication module is used to realize the interaction between the industrial control computer and the PLC controller.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: The present invention locates and divides different regions of the image with the characteristics of the bolster spring through the vision system module, detects and identifies whether the pose state of the bolster spring is correct, which is convenient for the subsequent correct installation of the bolster spring. The present invention uses a target detection algorithm based on deep learning to perform region-based recognition on the image with the characteristics of the bolster spring, timely detects the bolster spring with installation failure, reduces the safety risk, and improves the installation efficiency of the bolster spring. Description of the Drawings

[0016] Figure 1 Schematic diagram of the vision-based bolster spring dropping detection system for railway vehicle bogies provided by an embodiment of the present invention;

[0017] Figure 2 Flowchart of the vision-based bolster spring dropping detection method provided by an embodiment of the present invention;

[0018] Figure 3Schematic diagram of an image with bolster spring features for the vision-based detection method of bolster spring drop in bogies of rail vehicles provided by an embodiment of the present invention;

[0019] Figure 4 Schematic diagram of the bounding box of the bolster spring target for the vision-based detection method of bolster spring drop in bogies of rail vehicles provided by an embodiment of the present invention;

[0020] Figure 5 Bounding box of the side frame base for the vision-based detection method of bolster spring drop in bogies of rail vehicles provided by an embodiment of the present invention;

[0021] Figure 6 Schematic diagram of the first area for the vision-based detection method of bolster spring drop in bogies of rail vehicles provided by an embodiment of the present invention Figure 1 ;

[0022] Figure 7 Schematic diagram of the first area for the vision-based detection method of bolster spring drop in bogies of rail vehicles provided by an embodiment of the present invention Figure 2 ;

[0023] Figure 8 Schematic diagram of the second area for the vision-based detection method of bolster spring drop in bogies of rail vehicles provided by an embodiment of the present invention. Detailed implementation manners

[0024] To make the objectives, technical solutions and advantages of the present invention clearer and more understandable, the present invention will be further described in detail below with reference to specific embodiments.

[0025] As Figure 1 shown, a vision-based detection system for bolster spring drop in bogies of rail vehicles proposed by the present invention is composed of an image acquisition module, an industrial control computer, a vision system module, a communication module, a PLC controller, etc. The image acquisition module is composed of an industrial camera and a light source, and the image acquisition module is used to acquire images with bolster spring features. A vision system module is set in the industrial control computer, and the vision system module is used to judge the tipping of the bolster spring and the dropping of the bolster spring according to the images with bolster spring features to obtain the detection result. The communication module is used to realize the interaction between the industrial control computer and the PLC controller. Through the communication module, the industrial control computer transmits the detection result to the PLC controller, and the PLC controller then issues an instruction to the industrial control computer through the communication module according to the action process.

[0026] As Figure 2 shown, a vision-based detection method for bolster spring drop in bogies of rail vehicles proposed by the present invention is carried out according to the following steps:

[0027] S1. As Figure 3As shown, an image with bolster spring features is acquired, and the image with bolster spring features is captured by an industrial camera. In this embodiment, the focal length of the industrial camera is set to 6 mm.

[0028] S2. Use a deep learning object detection algorithm to perform bolster spring detection and side frame base detection on the image with bolster spring features respectively, obtaining a bolster spring target bounding box and a side frame base bounding box, as Figure 4 and Figure 5 shown.

[0029] S3. Divide the area of the image with bolster spring features, dividing it into a first area and a second area. The first area is the area where the bolster spring may be toppled, and the first area includes all side frame base bounding boxes where the bolster spring is not installed, as Figure 6 and Figure 7 shown. The second area is the area where the bolster spring may fall off, and the second area is the correct installation position of the newly installed bolster spring, as Figure 8 shown.

[0030] S4. Perform bolster spring toppling judgment based on the first area and the bolster spring target bounding box. Specifically: perform object detection on the first area. If there is a bolster spring target bounding box in the first area, it is judged that toppling exists, an error signal is sent and the bolster spring installation is interrupted. Otherwise, a correct signal is sent and the bolster spring installation continues.

[0031] S5. Perform bolster spring falling-off judgment based on the second area and the bolster spring target bounding box. Specifically: perform object detection on the second area. If there is a bolster spring target bounding box in the second area, a correct signal is sent and the bolster spring installation continues. Otherwise, it is judged that toppling exists, an error signal is sent and the bolster spring installation is interrupted.

[0032] According to the bolster spring assembly sequence, every time a bolster spring is installed, the position of this bolster spring is detected according to steps S1 to S5. After all bolster springs are installed, the process ends.

[0033] Among them, if the outer surface of the bolster spring is clean, the pose is accurate, and the detection environment light is completely unchanged, the bolster spring toppling and falling-off judgments in this method can be directly obtained by the opencv template matching algorithm. However, often the rotation directions of the installation threads of the bolster springs on the production line are different, there are dust and oil stains on the surface of the bolster springs, and the environmental light varies greatly at different times, making it difficult to detect the bolster springs. Therefore, the bolster spring toppling and falling-off judgment method in the present invention has more advantages in positioning accuracy compared with alternative solutions.

[0034] The present invention locates and divides different regions of an image with pillow spring features through a vision system module, detects and identifies whether the pose state of the pillow spring is correct, and facilitates the correct installation of the subsequent pillow spring. The present invention utilizes an object detection algorithm based on deep learning to perform region-based identification on an image with pillow spring features, timely detects pillow springs with installation failures, reduces safety risks, and improves the installation efficiency of pillow springs.

[0035] Embodiments of the present invention are intended to cover all such substitutions, modifications, and variations that fall within the broad scope of the appended claims. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A vision-based detection method for the dropping of the bolster spring of a railway vehicle bogie, characterized in that, Proceed as follows: S1. Obtain an image with bolster spring features; S2. Use a deep learning object detection algorithm to perform bolster spring detection and side frame base detection on the image with bolster spring features respectively, and obtain the bolster spring target bounding box and the side frame base bounding box; S3. Divide the image with bolster spring features into a first area and a second area; the first area includes all side frame base bounding boxes where no bolster spring is installed, and the second area is the correct installation position of the newly installed bolster spring; S4. Judge whether the bolster spring is toppled according to the first area and the bolster spring target bounding box. If toppling exists, send an error signal; otherwise, send a correct signal; S5. Judge whether the bolster spring has fallen according to the second area and the bolster spring target bounding box. If falling exists, send an error signal; otherwise, send a correct signal; The specific steps of S4 are: perform object detection on the first area. If there is a bolster spring target bounding box in the first area, judge that toppling exists, send an error signal and interrupt the installation of the bolster spring; otherwise, send a correct signal and continue the installation of the bolster spring; The specific steps of S5 are: perform object detection on the second area. If there is a bolster spring target bounding box in the second area, send a correct signal and continue the installation of the bolster spring; otherwise, judge that toppling exists, send an error signal and interrupt the installation of the bolster spring.

2. The method for detecting the dropping of the bolster spring of the railway vehicle bogie based on vision according to claim 1, wherein, The image with bolster spring features is collected by an industrial camera, and the focal length of the industrial camera is set to 6 mm.

3. The vision-based bogie bolster spring drop detection system for the vision-based rail vehicle bogie bolster spring drop detection method according to claim 1 or 2, comprising: An image acquisition module, an industrial control computer, a vision system module, a communication module and a PLC controller; characterized in that The image acquisition module includes an industrial camera and a light source, and the image acquisition module is used to collect an image with bolster spring features; a vision system module is set in the industrial control computer, and the vision system module is used to judge whether the bolster spring is toppled and whether the bolster spring has fallen according to the image with bolster spring features; the communication module is used to realize the interaction between the industrial control computer and the PLC controller.

Images