Small-scale helmet detection based on improved YOLOv5 and moving object detection

In industrial human-computer interaction, the safety of human is particularly important, helmet wearing plays a crucial role in ensuring construction safety and security. In smart industrial systems, intelligent helmet detection is a fundamental task. However, in practical scenarios, safety helmets often appear as small objects in the monitoring scenes, posing challenges for traditional object detection methods. This paper addresses the enhancement of small-scale helmet detection in intelligent monitoring. We achieve this by combining unsupervised foreground detection with an improved object detection network. Experimental results demonstrate significant improvements in detection accuracy compared to the original YOLOv5 algorithm.