Video stream safety helmet identification method and device based on YOLO, equipment and storage medium

• Main Authors: LING XIAOBAO, LIU ZHEWEI, ZHANG JIANQIANG, ZHU CHENYUE, NIE YIHE, XIAO ZIKAI
• Format: Patent
• Language: Chinese; English
• Published: 24.08.2021
• Subjects: CALCULATING; COMPUTER SYSTEMS BASED ON SPECIFIC COMPUTATIONAL MODELS; COMPUTING; COUNTING; HANDLING RECORD CARRIERS; PHYSICS; PRESENTATION OF DATA; RECOGNITION OF DATA; RECORD CARRIERS

The invention relates to a video stream safety helmet identification method, device based on YOLO, equipment and a storage medium, and the method comprises the steps: firstly obtaining image information of construction of a worker at a construction site in the air through an unmanned plane, then carrying out the preprocessing of an image, adjusting the size of the input image to 416*416, segmenting the image into S*S grids, and if the center of an object falls on a certain grid, the grid being responsible for detecting the object. In addition, the network structure of the YOLO algorithm is adjusted, and a small target detection layer is added between a convolutional layer and a residual jump layer. Meanwhile, it is considered that the color of the safety helmet is a pure color, when the loss function is designed, the ratio of the maximum color pixel number to all pixels in the prediction frame is considered; moreover, most safety helmets are approximately circular, the length-width ratio of the bounding box i