UAV Night Target Detection Method Based on Improved YOLOv5

Responding to the problem of low accuracy of UAV night target detection technology, existing night target detection methods and YOLOv5 algorithm are briefly analyzed, and a night target detection approach based on improved YOLOv5 is presented. Firstly, the improved Retimex algorithm is added into the original YOLOv5 network to dynamically enhance the original network. Secondly, the spatial channel attention mechanism is added into the backbone network to improve the model expression ability, and then during the model training process, the Focus layer is replaced with CBS layer to optimize the model training effect. Finally, the proposed improved YOLOv5 algorithm is simulated on the performance of UAV nighttime target detection, and the experimental results indicate that the improved YOLOv5 algorithm improves 11.5% and 7.9% over the original YOLOv5 algorithm in terms of checking accuracy and checking completeness, respectively, and effectively enhances the UAV night target detection capability, which is of strong reference value for the research of UAV experimental night target detection and tracking.