Research on Unmanned Warehouse Scene Monitoring Method Based on Improved YOLO V5s

With the rapid development of industrial automation and intelligence, unmanned warehouse scene monitoring has a wide range of application prospects in many fields. However, the existing monitoring methods are often faced with problems such as imperfect attention mechanism, low utilization rate of feature information, and unsatisfactory monitoring effect. To solve these problems, we propose an unmanned scene monitoring method based on the integration of attention mechanism and dense multi-scale features. By integrating the attention mechanism, this method realizes the comprehensive attention to the objects in the monitoring scene. Firstly, this paper designs an improved YOLO v5s model to solve the problem of inaccurate prediction frame and low detection accuracy caused by the object detection algorithm under the large area occlusion of warehouse goods. This paper introduces ECA-Net attention mechanism in the neck network in order to strengthen the ability to extract the features of warehouse goods. Secondly, considering the impact of model size on real-time performance, this paper introduced a space void pyramid pool structure into the YOLO v5s backbone network to reduce the loss of context information, reduce the calculation amount of the model, and improve the recognition and classification ability of object features. In order to verify the feasibility and superiority of the proposed method, we conducted experiments on several public unmanned scene monitoring data sets. The experimental results show that our method has excellent performance in monitoring accuracy and real-time performance. Especially in the complex background and dynamic monitoring scene, our method can identify and track the target object more accurately than the traditional monitoring method. In conclusion, our proposed unattended scene monitoring method based on the integration of attention mechanism and dense multi-scale features can effectively solve the problems faced by existing monitoring methods and improve the monitoring effect. This method has a wide application prospect and can provide strong support for the development of industrial automation and intelligence.