The detection of dim and small space-targets is crucial in space situational awareness missions; however, low signal-to-noise ratio (SNR) targets and complex backgrounds pose significant challenges to such detection. This paper proposes a space-target detection framework comprising a space-target detection network and a k-means clustering target centroid positioning method. The space-target detection network performs a three-dimensional convolution of an input star image sequence to learn the motion features of the target, reduces the interference of noise using a soft thresholding module, and outputs the target detection result after positioning via the offsetting branch. The k-means centroid positioning method enables further high-precision subpixel-level centroid positioning of the detection network output. Experiments were conducted using simulated data containing various dim and small space-targets, multiple noises, and complex backgrounds; semi-real data with simulated space-targets added to the real star image; and fully real data. Experiments on the simulated data demonstrate the superior detection performance of the proposed method for multiple SNR conditions (particularly with very low false alarm rates), robustness regarding targets of varying numbers and speeds, and complex backgrounds (such as those containing stray light and slow motion). Experiments performed with semi-real and real data both demonstrate the excellent detection performance of the proposed method and its generalization capability.

As a core technology of space situational awareness, dim and small space target detection plays a critical role in satellite surveillance and space defense. However, it remains highly challenging due to low signal-to-noise ratios of targets, stellar clutter interference, dynamic backgrounds, and imaging noise. Space-based imagery often lacks texture and contextual information of both targets and backgrounds. Most existing detection methods are developed for uncrewed aerial vehicle (UAV) detection in ground-based scenes with abundant spatial details. These methods often treat spatial and temporal features independently, neglecting the importance of spatiotemporal collaboration. Consequently, their effectiveness diminishes in space-based imagery, where spatial features are insufficiently reliable for accurate detection. To address it, we propose a temporal evolution-aware network with spatial guidance (TESNet). Rather than depending on conventional spatial backbone networks, TESNet employs a lightweight spatiotemporal backbone to capture spatial and temporal representations jointly from early stages. A dual-branch architecture is then designed to further enhance spatial and temporal features, where the temporal branch models cross-frame temporal evolution, and the spatial branch refines target spatial representations. In addition, a novel gated fusion module based on dual attention is introduced to effectively integrate spatial and temporal features, thereby strengthening deep spatiotemporal correlations. To train and evaluate our approach, we generate a simulated dataset, SIMDataset, specifically designed for dim and small target detection in space-based scenarios. Through a series of experiments and comparisons with current methods, we demonstrate the effectiveness of TESNet in addressing the challenges of dim and small space target detection under complex deep-space backgrounds. Our code and simulated dataset is publicly accessible online.

As a core technology of space situational awareness, dim and small space target detection plays a critical role in satellite surveillance and space defense. However, it remains highly challenging due to low signal-to-noise ratios (SNRs) of targets, stellar clutter interference, dynamic backgrounds, and imaging noise. And space-based imagery often lacks texture and contextual information of both targets and backgrounds. Most existing detection methods are developed for unmanned aerial vehicle (UAV) detection in ground-based scenes with abundant spatial details. These methods often treat spatial and temporal features independently, neglecting the importance of spatiotemporal collaboration. Consequently, their effectiveness diminishes in space-based imagery, where spatial features are insufficiently reliable for accurate detection. To address it, we propose a Temporal Evolution-aware Network with Spatial Guidance (TESNet). Rather than depending on conventional spatial backbone networks, TESNet employs a lightweight spatiotemporal backbone to capture spatial and temporal representations jointly from early stages. A dual-branch architecture is then designed to further enhance spatial and temporal features, where the temporal branch models cross-frame temporal evolution, and the spatial branch refines target spatial representations. Additionally, a novel gated fusion module based on dual-attention is introduced to effectively integrate spatial and temporal features, thereby strengthening deep spatiotemporal correlations. To train and evaluate our approach, we generate a simulated dataset, SIMDataset, specifically designed for dim and small target detection in space-based scenarios. Through a series of experiments and comparisons with current methods, we demonstrate the effectiveness of TESNet in addressing the challenges of dim and small space target detection under complex deep-space backgrounds. Our code and simulated dataset is publicly accessible at https://github.com/rcxlk/TESNet.