Steel Coil Recognition Neural Networks for Intelligent Cranes

Steel coil recognition in factories is important for an intelligent crane. In recent years, the deep neural network has become a popular tool in industry. The YOLO family of object detection methods has been widely applied. However, the original YOLOv8 has certain limitations in dealing with light change, occlusive targets, and small long-distance targets. To achieve a better recognition effect and improve the working efficiency, in this paper we make some improvements. Firstly, Space-to-depth (SPD) and Large Selective Kernel Networks (LSK) are added to the network Backbone. Secondly, the BiFormer attention mechanism is added to the neck part of the network, and the self-attention mechanism is used to obtain the global relationship between input to improve the network performance, so that the network can receive information at any time and autonomously adjust its receptive field in the spatial dimension. The improved YOLOv8 has better recognition accuracy for low-resolution occluded objects and small target objects. By experiments, the improved YOLOv8 network has 84.3% mAP on the self-built datasets, which is better than the 82.1% mAP of the original YOLOv8 network.