The rapid development of remote sensing technology has brought about a sharp increase in the amount of remote sensing image data. However, due to the satellite’s limited hardware resources, space, and power consumption constraints, it is difficult to process massive remote sensing images efficiently and robustly using the traditional remote sensing image processing methods. Additionally, the task of satellite-to-ground target detection has higher requirements for speed and accuracy under the conditions of more and more remote sensing data. To solve these problems, this paper proposes an extremely efficient and reliable acceleration architecture for forward inference of the YOLOX-s detection network an on-orbit FPGA. Considering the limited onboard resources, the design strategy of the parallel loop unrolling of the input channels and output channels is adopted to build the largest DSP computing array to ensure a reliable and full utilization of the limited computing resources, thus reducing the inference delay of the entire network. Meanwhile, a three-path cache queue and a small-scale cascaded pooling array are designed, which maximize the reuse of on-chip cache data, effectively reduce the bandwidth bottleneck of the external memory, and ensure an efficient computing of the entire computing array. The experimental results show that at the 200 MHz operating frequency of the VC709, the overall inference performance of the FPGA acceleration can reach 399.62 GOPS, the peak performance can reach 408.4 GOPS, and the overall computing efficiency of the DSP array can reach 97.56%. Compared with the previous work, our architecture design further improves the computing efficiency under limited hardware resources.

An invalid cloud region masking method based on local edge context is proposed for the compression of remote sensing images. Through analyzing the characteristics of various compression algorithms and taking the local edge information of the cloud region into account, the decompression quality is improved. First, according to the cloud mask information, labeling the connected cloud region, second, performing region growth on the labeled mask image, then differing the two mask images to obtain the local edge context, and finally different invalid cloud regions are filled with the average of the respective local edge context pixels. Using the image testing set generated from QuickBird and OrbView images, our method’s impact on six common remote sensing image compression algorithms is analyzed experimentally. The experimental results show that our masking method can improve the decompressed image quality when the compression ratio is certain. When the image quality is fixed, it can further reduce the compressed bitrate. For onboard application, our masking method can increase the onboard imaging time of the satellites, and eventually improve the onboard specifications of remote sensing satellites.

由于空间探测任务中需处理的遥感图像数量增多, 对目标检测系统的鲁棒性和时效性要求越来越高, 将大量遥感数据传输给地面再进行处理已经无法满足空间探测任务的需求. 针对此问题, 本文开展了基于宇航级NPU的遥感图像在轨目标检测系统研究, 以Yolov5s网络为基础, 替换与NPU适配度低的部分并引入注意力机制提升精度, 解决深度学习算法模型复杂和计算量大导致其在星上难以部署的问题. 优化后的网络迭代训练后部署在开发板上, 经CPU - NPU并行协同处理, 并行执行图像处理三部分, 充分利用Yulong810A平台资源. 实验显示, 优化后的网络部署到星上平台后参数量减少75%, mAP 值达71.25%, 检测速度为47.67 frame·s^–1, 均超过原版Yolov5s网络, 成功构建了一个更轻量快速的目标检测系统.