At present, modular satellite is an important development direction of satellite technology, and in order to realize the function of autonomous identification and transfer of satellite modules by robotic arm, a modular satellite component recognition algorithm based on improved Yolov7-tiny is proposed. By introducing the FasterNet structure, the backbone network is improved to reduce the number of parameters and calculations of the model, and improve the detection efficiency and frame rate of the model. At the same time, the loss function Repulsion Loss is introduced to improve the detection accuracy of occluded images in the dataset. In addition, a dataset containing multi-class modular satellite component models was created, and the improved algorithm on this dataset had an mAP of up to 95.6% and a FPS of 121.95 frames per second at an interaction ratio of 0.5, which was about 3% higher than that of the original algorithm, and the FPS was increased by about 12 frames per second. Finally, the improved algorithm model was pruned to further reduce the size of the model.