Research on construction machinery identification based on improved YOLOv5

In order to effectively control the security risks brought by the dynamic change of construction machinery, this paper mainly selected the dynamic change of small target machinery in the construction scene as the research object, through the object detection technology to identify all kinds Firstly, aiming at the problem of small sample data, the sample size was amplified by PCA three-channel color jitter. Secondly, aiming at the problem of small target recognition difficulty, the improved adaptive attention mechanism was integrated into the YOLOv5 model, and the test accuracy of the YOLOv5 model was improved, and the test accuracy of the YOLOv5 was optimized by changing the non-maximum suppression. Finally, The improved YOLOv5 detection model is trained and tested by combining the self-built data set. The experimental results show that the improved algorithm has improved average accuracy and recall rate compared with the original algorithm, and has strong generalization ability in small target detection in construction sites.