Remote Sensing Target Detection System Based on Spiking Neural Networks

With the rapid development of Earth observation technology, remote sensing image object detection has been increasingly applied in fields such as national defense and urban planning. However, existing methods based on deep convolutional neural networks suffer from high computational energy consumption and long inference latency, which makes them difficult to deploy directly on resource-constrained edge platforms such as satellite- and aircraft-mounted systems. To address this challenge, this paper investigates and implements a remote sensing object detection system based on spiking neural networks. The proposed system adopts biologically inspired spike encoding and event-driven computing mechanisms, significantly reducing computational redundancy. At the hardware level, a spiking neural network–based remote sensing object detection system is constructed, achieving efficient SNN inference acceleration through the collaborative optimization of processing engine arrays, instruction control subsystems, and memory scheduling mechanisms. Experimental validation is conducted on the Xilinx XC7VX690T FPGA platform, and the results demonstrate that the proposed system significantly reduces power consumption and latency while maintaining high detection accuracy, with reasonable resource utilization, providing a feasible hardware–software co-design solution for low-power, real-time intelligent remote sensing processing at the edge.