I-SHOU University / Prof. CHUNG-KUN YEN

 Pain Points Solved 

With the advancement of intelligent transportation systems and urban management, conventional license plate recognition and surveillance systems are typically restricted to fixed installations, which limits their capability to continuously acquire dynamic information from moving vehicles. To overcome this constraint, a large-scale mobile vehicle feature recognition and recording system was developed. The proposed system enables real-time vehicle feature acquisition on mobile platforms, integrates structured data storage, and supports subsequent statistical analysis and computational processing.

 Technology Introduction 

The system is implemented using a Python-based architecture that integrates deep learning and advanced image processing techniques, with a specific focus on real-time detection and recognition of license plates on large articulated vehicles. Video streams are reliably acquired from cameras via an FFmpeg-based streaming module using the RTSP protocol, after which object detection is performed using the YOLO (You Only Look Once) framework.
To enhance recognition stability and accuracy, the system incorporates a multi-voting strategy in conjunction with tracking algorithms and applies PaddleOCR for optical character recognition. In addition, validation mechanisms supporting both legacy and current license plate formats are implemented to further improve recognition reliability. Recognition results are rendered in real time within the visual interface, while the most frequently detected outcomes are automatically logged into a database with corresponding timestamps. Concurrently, snapshot images of detected license plates are archived locally to support long-term management and traceability. Beyond license plate recognition, the proposed system integrates AI-based visual analysis with cloud-based data matching to identify extended vehicle attributes, including color, model, damage locations, decals, and modification features. These data are synchronously transmitted to a centralized data center to establish dynamic vehicle profiles. Collectively, this system overcomes the inherent limitations of conventional static surveillance approaches and enables large-scale data acquisition and analytics in wide-area mobile environments, thereby providing a scalable foundation for intelligent transportation and smart city applications.

Figure1. Upon successful license plate recognition, the system records the corresponding timestamp and synchronously writes the information to the database in conjunction with data obtained from the weighbridge system. Concurrently, snapshot images of the recognized license plates are automatically captured and archived as evidentiary records to ensure data integrity, traceability, and subsequent verification.

 Application Examples 

To strengthen factory access control, a recognition and logging system is introduced that utilizes license plates of large vehicles as primary identification features. The proposed system enables automated and accurate recognition of inbound and outbound vehicle license plates, while systematically managing access records to ensure both security and operational efficiency within the facility. Furthermore, the system is integrated with a weighbridge management platform, thereby facilitating interoperability and seamless data exchange among heterogeneous enterprise information systems.

 Related Links 

None

 Patent Name and Number 

In progress

 Industry-Academia / Tech Transfer Partner 

Industry-Academia: FULL MORE METAL INDUSTRY CO., LTD.

 Honors and Awards  

None

 Technical Contact  

Yu-Hui Huang, Manager

I-SHOU University
Tel: +886 7-6577711 ext. 2194
Email: yuhuihuang@isu.edu.tw