Research and application of operation field safety control based on YOLOv4 algorithm in Qt

Abstract

Safe production is the basis of all the work of enterprises, the occurrence of any accident, will directly affect production activities, and even cause property losses and casualties, therefore, the safety management of enterprises is particularly important. Combined with the current situation of safety production management at construction sites in the power industry, the safety of construction sites is mostly managed by the person in charge of on-site construction, which is prone to the lack of safety control and control, forming safety hazards or causing safety accidents. In order to improve the safety management level of the construction site and realize the transformation from human control to automatic control of equipment, this paper proposes a kind of safety monitoring and early warning based on network video to build production site, which is applied to on-site safety management. Through the inspection and survey of transmission ground line robot to carry the ball machine camera to monitor the site construction, when the safety hazard occurs, the safety supervisor can receive a safety warning, and through the background can be monitored in real time. First of all, the camera ball machine collects pictures and videos of workers wearing hard hats at the construction site to build a VOC dataset, and uses labelme software to label the data set with points of interest; Then, the target dataset is trained and tested by using the YOLOv4 deep learning target recognition algorithm in Qt; Then, the on-site safety helmet wearing is monitored, and finally the on-site wearing of the safety helmet is positioned and the coordinate display is displayed, that is, the additional target test is successful, and safety behavior supervision of the construction personnel at the construction site are realized. The results show that the trained safety supervision system accurately identifies the hard hats worn by workers in the video stream of the construction site, which provides ideas and directions for the development of the new model of safety supervision.