A four-wheel track robot rail fastener identification and positioning method based on deep learning depth camera

Abstract

The application provides a deep learning depth camera-based rail spike identification and positioning method, which captures the rail spike and the surrounding environment of the rail spike by a depth camera and inputs the captured image into a computer, and the computer identifies and analyzes the content captured by the camera. The four-wheel track robot is driven at a speed V0 to reach an X0 point, so that the rail spike enters the shooting area of the depth camera, the rail spike is identified and preliminarily positioned, according to the preliminary positioning information, the four-wheel track robot is driven at a slightly slower speed V1 to reach an X1 point near the work point, the rail spike is identified again and accurately positioned, according to the accurate positioning information, the four-wheel track robot is driven at a slower speed V to reach the work point X, and then the work is started. The two-step positioning method of preliminary positioning and accurate positioning combined with the speed control mode of the four-wheel track robot provides accurate rail spike position information for the four-wheel track robot and improves the work precision of the four-wheel track robot.

Description

Technical Field

[0001] This invention relates to a method for identifying and locating road studs. Background Technology

[0002] Today, my country's railway development ranks among the world's leading positions, bringing us much convenience while also posing numerous hidden dangers. Regarding railway tracks, frequent train operations place enormous loads on them, causing track spikes to deform, break, and fall off, creating significant safety hazards. Therefore, routine track maintenance is crucial. Traditionally, workers inspected each track individually, a method that not only consumes enormous social resources but also poses significant safety risks to the inspectors. With the development and increasingly widespread application of artificial intelligence and machine vision technologies, research into methods for detecting, identifying, and locating railway track spikes is particularly important. Summary of the Invention

[0003] When a four-wheeled rail robot operates on railway tracks, it needs to accurately identify and locate rail spikes to determine the work point. In order to provide the four-wheeled rail robot with more accurate rail spike position information, improve the working efficiency of the rail robot, and reduce maintenance costs, this invention proposes a rail spike identification and positioning method for a four-wheeled rail robot based on a deep learning depth camera.

[0004] To address the aforementioned problems, this invention provides the following technical solution: a method for identifying and locating road spikes on a four-wheeled track robot based on a deep learning depth camera, comprising the following steps:

[0005] Step 1: The four-wheeled track robot moves along the rails at a speed of V0 from the initial point O to the work point X.

[0006] Step 2: Capture the environment around the railway tracks using a depth camera and input it into a computer. Then, the computer uses a deep learning algorithm to identify the track spikes from the content captured by the camera.

[0007] Step 3: When the four-wheeled track robot arrives at point X0, the road spike appears in the depth camera's field of view. The deep learning algorithm is then used to identify and initially locate the road spike, and the collected road spike position information is fed back to the computer. After receiving the initial positioning information, the computer issues a drive command to drive the four-wheeled track robot to continue moving towards the work point X at a slightly slower speed V1. Point X0 is located between the initial point O and the work point X.

[0008] Step 4: When the four-wheeled track robot moves to point X1, the depth camera accurately identifies and positions the rail spikes and feeds back the accurate positioning information of the spikes to the computer. After receiving the accurate positioning information, the computer issues another drive command to drive the four-wheeled track robot to move slowly to the work point X at a slower speed V. Point X1 is located between point X0 and work point X.

[0009] Step 5: After the four-wheeled track robot arrives at the work point X, it determines the position of the road spike based on the deep learning algorithm, reduces the speed of the four-wheeled track robot to 0, and then begins the operation.

[0010] Step Six: After the four-wheeled rail robot completes its task, repeat steps one through five to perform the next set of rail spike operations.

[0011] Where V0 > V1 > V > 0.

[0012] In steps three and four, the four-wheeled track robot moves forward, and the depth camera begins to capture the environment around the rails and identify and detect the rail spikes. Once the camera initially captures a spike, it identifies, detects, and locates it. The data is then fed back to the computer, which issues a drive command to move the four-wheeled track robot towards the work point X. During this process, the robot's speed decelerates from V0 to V1, then from V1 back to V, and finally to 0. This combination of initial and precise spike identification improves the accuracy of spike recognition.

[0013] In step three, deep learning algorithms are used to identify and detect targets, while depth images captured by a depth camera are used to determine the distance between the camera and the object.

[0014] The road spike identification and positioning method for four-wheeled track robots based on deep learning depth cameras provided by this invention has the following advantages compared with the prior art: This invention uses deep learning algorithms to identify and detect road spikes, and uses depth cameras to locate road spikes. By combining the two-step positioning method of preliminary positioning and precise positioning with the speed control of the four-wheeled track robot, it provides accurate road spike position information for the four-wheeled track robot and improves the operation accuracy of the four-wheeled track robot. Attached Figure Description

[0015] Figure 1 This is a schematic diagram illustrating the target recognition and localization operation of the depth camera according to the present invention.

[0016] Figure 1 In the process, a depth camera is mounted on a four-wheeled track robot. The track robot moves from the initial point to the work point X at a speed of V0. When the four-wheeled track robot reaches point X0, its speed decreases to V1. It continues to move, and when it reaches point X1, its speed decreases to V. When the four-wheeled track robot reaches the work point X, its speed decreases to 0, and the operation begins. Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The following description of at least one exemplary embodiment is merely illustrative and is in no way intended to limit the present invention or its application or use. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0018] Techniques, methods, and equipment known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and equipment should be considered part of the specification.

[0019] This invention provides a method for identifying and locating road spikes on a four-wheeled track robot based on a deep learning depth camera, comprising the following steps:

[0020] Step 1: The four-wheeled track robot moves along the rails at a speed of V0 from the initial point O to the work point X;

[0021] Step 2: Capture the environment around the railway tracks using a depth camera and input it into the computer. Then, the computer uses a deep learning algorithm to identify the track spikes from the content captured by the camera.

[0022] Step 3: When the four-wheeled track robot arrives at point X0, the road spike appears in the field of view of the depth camera. The deep learning algorithm is used to identify and initially locate the road spike, and the collected road spike position information is fed back to the computer. After receiving the initial positioning information, the computer issues a drive command to drive the four-wheeled track robot to continue moving towards the work point X at a slightly slower speed V1.

[0023] Step 4: When the four-wheeled track robot moves to X1, the depth camera accurately identifies and locates the rail spikes and feeds back the accurate positioning information of the spikes to the computer. After receiving the accurate positioning information, the computer issues another drive command to drive the four-wheeled track robot to move slowly to the work point X at a slower speed V.

[0024] Step 5: After the four-wheeled track robot arrives at the work point X, it determines the position of the road spike based on the deep learning algorithm, reduces the speed of the four-wheeled track robot to 0, and then begins the operation.

[0025] Step Six: After the four-wheeled rail robot completes its task, repeat steps one through five to perform the next set of rail spike operations.

[0026] In steps three and four of this method, the four-wheeled track robot moves forward, and the depth camera begins to capture the environment around the rails and identify and detect rail spikes. Once the camera initially captures a spike, it identifies, detects, and locates it. The data is then fed back to the computer, which issues a drive command to move the four-wheeled track robot towards the work point X. During this process, the robot's speed decelerates from V0 to V1, then from V1 back to V, and finally to 0. Combining the initial and precise identification of the spikes improves the accuracy of spike recognition. This method provides the four-wheeled track robot with precise spike location information based on deep learning algorithms, improving robot operating efficiency and saving costs.

[0027] In step three of this method, a deep learning algorithm is used to identify and detect the target, while the distance between the camera and the object is determined using a depth image captured by a depth camera.

[0028] In this method, the target distance information acquired by the depth camera is fed back to the computer, and the computer issues corresponding instructions to make the four-wheeled track robot move towards the work point X until the four-wheeled track robot arrives at the work point X and completes the work.

[0029] In steps three and five of this method, after the four-wheeled track robot arrives at point X0 and point X1, the computer obtains the preliminary positioning information and precise positioning information of the track spikes, respectively. Then, the computer issues drive commands to change the movement speed of the four-wheeled track robot.

[0030] For brevity, the specific artificial intelligence algorithms described in the above embodiments are not detailed. Currently, object detection algorithms mainly include YOLO, SSD, R-CNN, Fast R-CNN, and Faster R-CNN. This method is primarily based on the YOLO algorithm. The idea behind this method is to use the entire image as input and directly regress the bounding box position and its category at the output layer. Compared to other methods, YOLO's advantage lies in its fast detection speed, as it directly predicts road spikes based on image information acquired by the camera, greatly improving work efficiency. This application does not improve image recognition technology; existing image recognition technologies are sufficient.

[0031] The above description is merely an exemplary embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for identifying and locating road spikes on a four-wheeled track robot based on a deep learning depth camera, characterized in that, Includes the following steps: Step 1: The four-wheeled track robot moves along the rails at a speed of V0 from the initial point O to the work point X; Step 2: Capture the environment around the railway tracks using a depth camera and input it into the computer. The computer then uses a deep learning algorithm to identify the track spikes from the content captured by the camera. Step 3: When the four-wheeled track robot arrives at point X0, the road spike appears in the depth camera's field of view. The deep learning algorithm is then used to identify and initially locate the road spike, and the collected road spike position information is fed back to the computer. After receiving the initial positioning information, the computer issues a drive command to drive the four-wheeled track robot to continue moving towards the work point X at a slightly slower speed V1. Point X0 is located between the initial point O and the work point X. Step 4: When the four-wheeled track robot moves to point X1, the depth camera accurately identifies and locates the rail spikes and feeds back the accurate positioning information of the spikes to the computer. After receiving the accurate positioning information, the computer issues another drive command to drive the four-wheeled track robot to move slowly to the work point X at a slower speed V. Point X1 is located between point X0 and work point X. Step 5: After the four-wheeled track robot arrives at the work point X, it determines the position of the road spikes based on a deep learning algorithm, reduces the speed of the four-wheeled track robot to 0, and then begins the operation; Step Six: After the four-wheeled track robot completes its work, repeat steps one through five to perform the next set of track spike operations.

2. The method for identifying and locating road spikes on a four-wheeled track robot based on a deep learning depth camera according to claim 1, characterized in that: In steps three and four, the four-wheeled track robot moves forward, and the depth camera begins to capture the environment around the rails and identify and detect the rail spikes. After the camera initially captures the spikes, it identifies, detects, and locates them. Then, the data is fed back to the computer, which issues a drive command to drive the four-wheeled track robot to move towards the work point X. During this process, the speed of the four-wheeled track robot decelerates from V0 to V1, then from V1 to V, and finally decelerates to 0. By combining the initial and precise identification of the rail spikes, the recognition accuracy of the rail spikes is improved.

3. The method for identifying and locating road spikes on a four-wheeled track robot based on a deep learning depth camera according to claim 2, characterized in that: In step three, deep learning algorithms are used to identify and detect targets, while depth images captured by a depth camera are used to determine the distance between the camera and the object.

4. The method for identifying and locating road spikes on a four-wheeled track robot based on a deep learning depth camera according to claim 1, characterized in that: V0>V1>V>0.

Images