Intelligent Person Monitoring system of the UHV Power Infrastructure Line Construction site based on Deep Neural Networks.

Recently, many types of workers are required for the ultra-high voltage (UHV) power line construction sites in China. For safety, each worker's real-time status should be monitored. The key point of this monitoring system is identifying and tracking each worker accurately and quickly, which is the special multi-objects tracking (MOT) problem at the large open scene. In this paper, an intelligent person monitoring system with sensors composed of a camera and the ultra-wide band (UWB) radars is proposed. The proposed MOT method is composed of two parts. The one is the method of detecting and identifying persons, which is composed of an image person recognition method that is a Faster R-CNN combined with a re-ID branch, a UWB worker localization method, and a person association method. And the other is tracking these persons, which is composed of an person-track association algorithm by a similarity matrix, and a tracking method by the Kalman Filter. The experiments illustrate that our method can monitor multi-persons accurately and quickly in real time.