Target Detection Algorithm for Unmanned Vehicles in Complex Environments

In the realm of unmanned vehicles, challenges persist regarding their limited perception and real-time obstacle detection capabilities. This paper introduces the YOLOv5-SA object recognition algorithm, built upon YOLOv5, which addresses these issues. Leveraging Laplace transformation and SSR image enhancement on the training dataset, coupled with the integration of the SiAM attention mechanism at the network's output layer, and the replacement of the SPPF module with the AIFI module, the proposed model significantly enhances detection speed and accuracy. Experimental results demonstrate lower target omission rates and faster detection in various complex environments.