Aiming at the problem that low identification accuracy,high false identification rate and missed identification rate and slow identification speed in helmet wear identification under complex industrial scene,an improved YOLOv3 helmet wear identification algorithm with high recognition accuracy and fast identification speed was proposed.The traditional YOLOv3 trunk network was trimmed and modified to improve the identification speed.The spatial pyramid pooling module was introduced to integrate local features and global features more effectively.In order to improve the fitting effect between the target prediction frame and the real target frame,the loss function was improved to CIoU.Finally,the fourth feature fusion scale was extended for small target identification to improve the accuracy of small target recognition.The results show that the improved YOLOv3 improves the average identification accuracy of helmet wear identification by 2.37% and the identification speed by 2.7 times in complex industrial environment,and at the same time,the missed identification rate and false identification rate of helmet wear identification are reduced.