Train signal state detection method based on deep learning and storage medium
By combining the deep learning-based SSD algorithm with RetinaNet-YOLO v3 in a two-step detection method, the real-time and accuracy issues of signal status detection during train operation are solved, achieving efficient signal status recognition.
CN116246245BActive Publication Date: 2025-12-16HEFEI UNIV OF TECH +1
Patent Information
- Application Number
- CN202211706150.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-12-29
- Publication Date
- 2025-12-16
- Estimated Expiration
- 2042-12-29
AI Technical Summary
Technical Problem
Existing technologies struggle to detect signals and their status accurately and in real time during train operation, especially when the detection accuracy is low for small light bulb targets.
Method used
The SSD algorithm of deep learning is used for signal machine target detection. The detection is performed in two steps by combining RetinaNet and YOLO v3 models. The VGG16 network structure and FPN feature pyramid network are used for signal machine state recognition.
Benefits of technology
It improves the real-time performance and accuracy of signal condition detection, meeting the testing requirements of engineering standards.
✦ Generated by Eureka AI based on patent content.
Smart Images
Figure CN116246245B_ABST
Abstract
The application discloses a state detection method of a train signal machine based on deep learning and a storage medium, and comprises the following steps: using a SSD algorithm of deep learning to detect a signal machine target; then, using a two-step detection method to detect the state of the signal machine, specifically, using relevant signal machine image data sets, putting the signal machine image data sets into a RetinaNet deep learning model for training, completing the detection of overall low-column signal machines and high-column signal machines, and then putting the detection results as data sources into a YOLO v3 deep learning network to detect the state of signal machine bulbs, and completing the detection of the state of the signal machine. The application adopts the SSD algorithm to detect the target of the signal machine, adopts the RetinaNet method combined with the YOLO v3, forms a two-step recognition signal machine state method, and improves the recognition efficiency and precision.
Need to check novelty before this filing date? Find Prior Art
Citation Information
Patent Citations
Traffic signal identification method and device, vehicle navigation equipment and unmanned vehicle
CN110688992A