Target Detection Algorithm for Drones Based on YOLOv5l_CA
To address the issue of detecting unauthorized "black flight" drones,a drone target detection method based on YOLOv5l is proposed. Firstly,a drone target dataset is established,and the drones are annotated. Next,YOLOv5l is applied as the base network for drone training,and the network model's hyperparameters are set for model training and evaluation. Then,YOLOv5l was improved to form an improved network YOLOv5l-CA,which introduced a coordinate attention mechanism to increase the detection accuracy of drones under the sky. Finally,comparative experiments are conducted with other network models. Experimental results show that the improved algorithm YOLOv5l-CA achieves an average accuracy of 94.6％,which is 2.4％,1.9％,and 0.8％ higher than YOLOv5s,YOLOv5m,and YOLOv5l algorithms,respectively. It performs well and meets the real-time detection requirements,verifying the feasibility of the improved algorithm for drone detection.
NETL
NSTL
万方数据
