Lightweight container code recognition based on multi-reuse feature fusion and multi-branch structure merger
Dapeng Yang,Guanghui Wang,Mingtang Liu,Shuang Yue,Hao Zhang,Xiaokang Chen,Mengxiao Zhang
DOI: https://doi.org/10.1007/s11554-023-01364-x
IF: 2.293
2023-10-08
Journal of Real-Time Image Processing
Abstract:In the field of global shipping, intelligent container code recognition has become a popular research topic. To address the issues of poor recognition accuracy and real-time performance in the complex port environment, this paper proposes a real-time container code recognition method based on YOLOV5. The proposed method consists of two stages: container code localization and container code recognition. First, for container code localization, the backbone component consists of MobileOne blocks to enhance feature extraction capability and accelerate inference speed. Second, for container code recognition, a lightweight MobileNetV3 is used as the backbone. Third, a feature extraction module, called Multiple Reuse Feature Pyramid Networks (MRFPN), is designed to extract sufficient semantic information. Experimental results show that the algorithm achieves accuracy rates of 96.5% and 93.5% in the container code localization and container code recognition stages, respectively, while significantly reducing the number of parameters, computations, and model size, and improving inference speed. Additionally, the proposed method performs better in terms of real-time performance and can meet the requirements of practical applications.
computer science, artificial intelligence,engineering, electrical & electronic,imaging science & photographic technology