NCSTP: A Benchmark Dataset for Non-Cooperative Space Target Perception

The automatic, accurate perception of targets in space is a crucial prerequisite for many aerospace missions, such as on-orbit maintenance and target monitoring. Therefore, research on perception technologies within images from spaceborne cameras, is of great significance. The recent, rapid development of deep learning has revealed its potential for application to space target perception. However, implementing deep learning models typically requires large-scale labeled datasets. In this study, we build a multitask benchmark space target dataset, NCSTP, to address the limitations of current datasets. First, we collect and modify various space target models for satellites, space debris, and space rocks. By importing them into a realistic space environment simulated by Blender, 200,000 images are generated with different target sizes, poses, lighting conditions, and backgrounds. Then, the data are annotated to ensure the dataset supports simultaneous space target detection, recognition and component segmentation. NCSTP has 10 fine-grained classes of satellites, 6 classes of space debris, and 4 classes of space rocks. All the data can be used for training space target detection and recognition models. We further annotate the body, solar panels, antennas, and observation payloads of each satellite for component segmentation. Finally, we test a series of state-of-the-art object detection and semantic segmentation models on the dataset to establish a benchmark.2025.6.16: A smaller version NCSTP-10000 is available now