文档介绍:第 卷 第 期 集美大学学报 (自然科学版)
26 4 SUN Shidan , ZHENG Jiachun , ZHAO Shijia , HUANG Yiqi
( 1. School of Marine Information Engineering, Jimei University, Xiamen 361021, China;
2. Navigation Institute, Jimei University, Xiamen 361021, China)
Abstract:
Aiming at the problem of automatic detection of wearing helmets and masks at the same time in
the scene of construction site and dangerous area, an improved YOLOv3 algorithm is proposed to enhance the
detection accuracy of helmets and masks. Firstly, the clustering algorithm in network is optimized. The weighted
kernel K-means clustering algorithm is used to analyze the dataset, so as to select the anchor box more suitable
for small targets detection and improve the average accuracy and speed of detection. Secondly, it optimizes the
Darknet characteristic network layer in YOLO network. The extracted quadruple down sampling feature map is
up sampling once. The double up sampling is fused with the previous double down sampling, and then it is
transmitted to the subsequent network together with quadruple down sampling, eightfold down sampling and six-
teen times down sampling to reduce the miss detection rate of small targets. Experimental results show that, the
average