Person detection and identification surrounding gas facilities based on Improved-Yolov4-tiny

Abstract

The safety of the gas facilities is an important guarantee for the production and life of the society, and the monitoring of people entering the area is an important means to guarantee the safety of the facility. In order to solve the problem of non-contact monitoring of people entering the gas facilities, an Improved-Yolov4-tiny method is proposed in this paper. This method can be used to monitor people entering the area in real time and determine whether the type of person is a worker or non-worker based on characteristics such as clothing. Among them, the model increases a large-scale feature layer to improve the recognition and detection of small targets. The introduction of attention mechanism for multi-scale feature fusion increases the effectiveness of fine-grained detection of person. The Improved-Yolov4-tiny detection algorithm not only meets the requirements of real-time monitoring, but also achieves better results at different resolutions. Compared with the traditional Yolov4-tiny detection algorithm, the Improved-Yolov4-tiny detection algorithm improved the mAP metric by 3.1%.