| With the rapid development of economy and the increasing improvement of people’s quality of life,the requirement of living environment and travel facilities is becoming higher and higher.Under this background,the construction industry has developed vigorously.Because of its high-altitude operation and high-intensity physical labor,construction operation is a work with high potential safety hazards.In this paper,a cascaded small target detection network of YOLO-person and YOLO-helmet is built to detect workers and workers’ safety hats at the construction site,so as to find out workers without safety hats in time,ensure workers’ safety and avoid construction accidents.Based on the carefully designed cascade detection algorithm network,this paper builds a complete small target detection system.The system we built mainly consists of resource module,camera module,algorithm module,web page module and alarm module: resource module manages the data of the whole system,including data collection,storage,processing and transmission;camera module extracts the monitoring video of construction site to provide original input data for the system;algorithm module completes the worker and worker safety helmet detection;Web page module shows the recognition effect,which is convenient for real-time viewing the situation of workers wearing safety caps on the construction site.When the alarm module discovers workers without safety caps,it pushes the specific location of workers to the management of the construction site to give an alarm.Cascade detection algorithm network consists of worker detection network and safety hat detection network: YOLO-person network is responsible for detecting workers in the image and marking their specific location with white boundary box;YOLO-helmet network is responsible for detecting the safety caps worn by workers on the basis of YOLO-person detection,and marking their specific location with red boundary box.Through the construction of the system and the design of the target detection algorithm network,a complete target detection system based on the construction site is realized and successfully applied to a local subway construction site. |
