Port staff trajectory exploitation via an emsemble detection-tracking framework
Xinqiang Chen, Meilin Wang, Yongsheng Yang, Huafeng Wu, Jun Ling, Jiangfeng Xian · 2022
For the safety supervision of port environment, we propose a port staff detection and tracking framework. Firstly, the framework is based on Faster-RCNN (Faster Region Convolutional Neural Networks) detection algorithm to identify the port staff in the surveillance video, and obtain the target location information. Based on the Deep SORT (Deep Simple Online and Realtime Tracking) tracking algorithm, we add Gaussian noise reduction and histogram equalization to preprocess the input image, and then use Kalman filter and Hungarian algorithm to predict and match, so as to improve the accuracy of target tracking. The framework effectively improves the accuracy of the algorithm, provides a powerful technical support for port security monitoring, and provides a security guarantee for port staff.