Hardware implementation of SNN-Based

Remote sensing scene classification plays an important role in many fields, but the traditional remote sensing classification methods face the problems of computational complexity and energy consumption, and it is difficult to meet the real-time requirements. Brain-inspired computing, especially spike neural networks (SNNs), has become an effective way to solve this problem due to its low power consumption and high reliability. In this paper, we propose an FPGA-Based SNN accelerator to implement an SNN-Based Remote sensing scene classification model to improve the real-time and resource efficiency of remote sensing image processing. We use the LeNet-5 network architecture as the core, build SNN based on the LIF model, and realize parallel processing and pipeline design by optimizing the data flow and hardware architecture, so as to improve the data throughput and processing speed. On the FPGA platform, our SNN accelerator achieves a high processing speed of 4690 frames/s while maintaining low power consumption (0.887W), which is approximately 3.75x faster than existing processors. At the same time, the overall system was evaluated using the adopted network architecture, and our architecture can achieve large-scale SNN network model inference with low power, low overhead, and high throughput.