Grenade Detection in Complex Environment Based on YOLOv5

The artificial blasting method of dealing with dud bombs has high risks and low efficiency. Based on the target detection technology, the realization of autonomous detonation by an unmanned platform can effectively reduce the risk of human operation and improve the efficiency of detonation. Aiming at the problem of grenade recognition in complex environments, this paper studies the target detection algorithm, analyzes the structure and principle of the algorithm, and finally verifies the feasibility of the algorithm on the self-made grenade data set. The results show that the weight of the YOLOv5 model is small, the inference speed is fast, and the model detection accuracy rate reaches 99.5%, which can effectively meet the needs of practical applications.