Research on Safety Belt Wearing Detection Algorithm based on Artificial Intelligence Working at Altitude

Under the call of the corresponding country, urban construction has been attached importance. The following is the safety of construction workers, safety helmet, safety belt, safety chain and other safety needs a new digital transformation. Given that there are few researches on high-altitude belt wear detection at present, this paper proposes an improved Yolov5 model cae pre-training Yolov5_MobileNetv3 detection algorithm based on self-supervised visual pre-training. Firstly, unmarked self-supervised pre-training cae model was used to learn the characterization ability of complex scenes in construction sites. Then, multi-layer feature distillation technology is used. Finally, for some small targets in dense scenes, Softnms algorithm is used to improve the detection ability of the model in complex environment for human body and seat belt, and enhance the generalization and robustness of the model. The experimental results show that compared with Yolov5_MobileNetv3 model, cae-pretrained Yolov5_MobileNetv3 model significantly improves the detection performance of human body, safety belt, safety hat and other human attributes. Based on its own data, the accuracy rate is 94.35%, recall rate is 92.92%, F1 value is 93.63%, which meets the requirements of belt wearing detection in construction scenarios.