Safety Helmet Wearing Detection Based on Super-resolution Reconstruction
Hailan Yu, Yi-Ran Sun, Chao Jia · 2021 IEEE 5th Information Technology,Networking,Electronic and Automation Control Conference (ITNEC) · 2021
As the most commonly used labor protection products, safety helmets can effectively reduce accident damage, and it is particularly necessary to supervise the wearing of workers' helmets. This paper designs and implements a safety helmet wearing detection system optimized based on super-resolution reconstruction, obtains images through the RTSP video stream, detects and marks whether the characters in the video image wearing a helmet, and outputs the detection results in the form of a real-time video stream. The detection system applies the ESPCN super-resolution reconstruction neural network algorithm to preprocess the image data, enhances the subtle characteristics of the distant characters in the original image, and improves the detection accuracy of whether the distant target characters wear a helmet. The test results show that the method can achieve the image and video detection of hard hat wearing well, and the detection rate and detection accuracy of small targets in the distance are significantly better than the results of single YOLOv3 neural network detection.