Abstract: In order to solve the problem that the object detection algorithm based on deep learning is difficult to deploy on a space-based image processing platform with limited resources due to the complex network structure and excessive computational cost, this paper proposes a convolutional neural network acceleration design based on aerospace-grade neural network processor (NPU), and uses the improved Yolov5s network to realize fast image processing function on the NPU. The optimized network is iteratively trained on the GPU through the public dataset VOC, and the three parts of image processing are executed in parallel after the CPU-NPU parallel collaborative processing design, making full use of the limited computing and storage resources of the Yulong810A platform. Experiments show that the optimized network not only reduces the number of parameters by 82%, but also improves the accuracy compared with the original Yolov5s network, with an mAP value of 82.35%. After the algorithm is deployed on the Yulong810A on-board processing platform, the target detection speed reaches 41.67fps/s, which is more than twice the speed of the original Yolov5s network, and realizes a lighter and faster object detection system.