Taipei, Taiwan, November 16th – Wistron Intelligent Mobility Solutions (IMS) has worked with Wistron-NYCU Embedded Artificial Intelligence Research Center (Team NYCU) to advance their mobility inspection technology towards commercialization stage. This year the collaboration resulted in a successful automatic railway inspection proof of concept project for the Institute of Transportation, Ministry of Transportation and Communications (IoT-MOTC). The initiative utilized Wistron IMS’s AI-powered ZigNeurons and ZigRail platforms. The venture concluded stage three proof of concept phase at Taichung Maintenance Office of Taiwan Railway Administration. It demonstrated that the automatic railway inspection can successfully function at night and at speeds of 60 km/h. The resulting data has revealed a high detection rate of railway component faults, providing strong evidence that the technology is ready for commercialization.

Since 2020, Wistron IMS has been working with Team NYCU on the development of railway/road inspection systems, acquiring advanced AI techniques and knowledge. The railway/road inspection system is able to detect road faults via AI and Machine Learning (ML), integrating High Definition Positioning System (HDPS) developed by Wistron IMS. The system provides seamless positioning information and uploads the data of abnormalities and faults to the cloud, resulting in improved maintenance office operations.

Chun-Yin Kuo, the professor of NYCU, said: “By utilizing continuous learning and continuous deployment architecture aided by semi-automatic labelling tools and semi-monitoring learning, we’ve accelerated the modelling development cycle. As a result, the system is not only able to reduce human resource costs required by model training, but also continuously learn different categories of faults. The inspection system equipped with this architecture has already proven to have high performance detection rate in road surface cracks, abnormalities of road marked lines, and more.”

The IoT-MOTC’s project of automatic railway inspection system was based on the techniques and experiences from previous road surface inspections. With the assistance from Taichung Maintenance Office and Yilan Maintenance Office of Taiwan Railway Administration, Wistron IMS and IoT-MOTC implemented computer vision recognition and HDPS of ZigRail towards real railway inspection. By continuously collecting data and information from vehicles on rail tracks and training the AI models with ZigNeurons, the team raises the precision rate of the model and accumulates image data for AI model training. This solution is already validated to have high detection rate for detection items such as railway track cracks, rail retainer faults, fishplate faults, and other railway components. The teams will continue to collect and expand image data to further enhance the detection rate.

Taiwan Railway Administration believes that safety is always their first priority, and the railway inspection process runs non-stop. However, the most sophisticated inspections are still scheduled to be operated during the middle of the night. This gives Wistron IMS and IoT-MOTC an opportunity to validate the use of AI and machine learning technologies that can implement efficient, precise, and automatic railway inspection during nighttime. Furthermore, it also increases maintenance efficiency and operational safety with more accurate position identification of faults or other problems.

To learn more about how ZigRail products help IoT-MOTC implement railway inspection with AI and ML advanced technologies, or how ZigNeurons products continuously train models to optimize AI modeling precision rate, please visit our website.