Lightweight Real-Time Vehicle Detection and Recognition of UAV Images Based on Brain-Inspired Computing Architecture

Vehicle detection and recognition based on Unmanned Aerial Vehicle (UAV) remote sensing images is of great significance in both civilian and military domains. There is currently a growing demand for real-time, precise, and reliable vehicle detection and recognition technologies due to complex and diverse application scenarios. In contrast to traditional close-range imaging, UAV remote sensing images present a more complex background, and the vehicle targets exhibit characteris-tics such as diverse types, multi-scale variations, and occasional dense distributions. This study aims to address the challenges associated with vehicle detection and recognition in UAV remote sensing images. We have conducted targeted optimization designs based on the characteristics of UAV remote sensing imaging to enhance the computational efficiency, accuracy, and reliability of object detection and recognition. Our approach involves the development of an improved algorithm designed to enhance object detection and recognition in UAV remote sensing images. This algorithm has been optimized for brain-inspired chips, enabling acceleration in detection and recognition speed on UAV edge-computing terminals to meet real-time requirements. The experimental results conclusively indicate that the proposed algorithm in this paper significantly improves the accuracy and efficiency of vehicle target detection in UAV remote sensing images.