Safety suit detection method and system based on improved YOLO V5

Abstract

The invention discloses a security clothing detection method and system based on improved YOLO V5, and belongs to the field of target detection. The method comprises the steps that a safety clothing wearing state training set is adopted to train and improve YOLO V5, a training sample comprises a worker picture frame, a label is a safety clothing wearing state, and a trained detection model is obtained; and each frame of the industrial monitoring video is input into the trained detection model to obtain a safety suit detection result. According to the method, different neural network structures are used for replacing a Backbone module of an original YOLO V5 algorithm, the OfficientNet is used as the Backbone, the width and depth of the network structures and the resolution of the input image are uniformly scaled through the expansion coefficient of the composite model, and the effect superior to that of manual parameter adjustment of the YOLO V5 is obtained. ResNet50 is adopted as a Backbone, due to the fact that a residual block is added, feature information extracted by a network is completely reserved to a next layer, and gradient dispersion between network layers is effectively eliminated in the forward propagation process. ShuffleNet or MobileNet is used as a Backbone, so that the complexity of a network structure and the size of a model are reduced, and a lightweight model is obtained.

Description

technical field [0001] The invention belongs to the field of target detection, and more specifically relates to a safety clothing detection method and system based on the improved YOLO V5. Background technique [0002] In industrial production, safety issues are very important, and the awareness of safety production is deeply rooted in the hearts of the people. Safety clothing is the protective clothing that workers must wear in the production and operation area. Standardized wearing of safety clothing can effectively protect the body of the staff and reduce the damage of dangerous chemicals such as acids and alkalis to the skin. The online real-time detection method of workers wearing safety clothing based on industrial surveillance video is an important guarantee for the safety of workers in industrial scenes, and it is also important for standardizing industrial management and safe production. At present, most industrial management units mostly adopt the method of manual...

AI Technical Summary

Problems solved by technology

[0006] 1. The target detection technology used in existing solutions is relatively backward

Most of the algorithms currently used by researchers to complete related target detection tasks are improvements to the YOLOv3 and YOLOv4 algorithm models. Not long ago, there were relatively few studies on the specific use of the latest YOLOv5 algorithm;

[0007] 2. Most of the safety clothing objects detected by the existing research programs come from reflective vests collected by web crawling or shooting in daily life scenes, and there is a lack of safety clothing target detection methods for real industrial scenes;

[0008] 3. The algorithm model in the existing research scheme is poor for the detection of safety clothing targets in real industrial scenes. The background of real industrial scenes is complex, and the target objects of safety clothing to be detected are vulnerable to illumination changes, multiple occlusions, loss of monitoring images, and motion blur. The interference of various complex problems, such as interference, puts forward higher requirements for the robustness of the algorithm model

Images