Space resources are of economic and strategic value. Infrared (IR) remote sensing, unaffected by geography and weather, is widely used in weather forecasting and defense. However, detecting small IR targets is challenging due to their small size and low signal-to-noise ratio, and the resulting low detection rates (DRs) and high false alarm rates (FRs). Existing algorithms struggle with complex backgrounds and clutter interference. This paper proposes an adaptive differential event detection method for space-based aerial target observation, tailored to the characteristics of target motion. The proposed IR differential event detection mechanism uses trigger rate feedback to dynamically adjust thresholds for strong, dynamic radiation backgrounds. To accurately extract targets from event frames, a lightweight target detection network is designed, incorporating an Event Conversion and Temporal Enhancement (ECTE) block, a Spatial-Frequency Domain Fusion (SFDF) block, and a Joint Spatial-Channel Attention (JSCA) block. Extensive experiments on simulated and real datasets demonstrate that the method outperforms state-of-the-art algorithms. To advance research on IR event frames, this paper introduces SITP-QLEF, the first remote-sensing IR event dataset designed for dim and moving target detection. The algorithm achieves an mAP@0.5 of 96.3%, an FR of 4.3 $\times {10}^{-5}$, and a DR of 97.5% on the SITP-QLEF dataset, proving the feasibility of event detection for small targets in strong background scenarios.

Space-based infrared (IR) detection, with wide coverage, all-time operation, and stealth, is crucial for aerial target surveillance. Under low signal-to-noise ratio (SNR) conditions, however, its small target size, limited features, and strong clutters often lead to missed detections and false alarms, reducing stability and real-time performance. To overcome these issues of energy-integration imaging in perceiving dim targets, this paper proposes a biomimetic vision-inspired Infrared Temporal Differential Detection (ITDD) method. The ITDD method generates sparse event streams by triggering pixel-level radiation variations and establishes an irradiance-based sensitivity model with optimized threshold voltage, spectral bands, and optical aperture parameters. IR sequences are converted into differential event streams with inherent noise, upon which a lightweight multi-modal fusion detection network is developed. Simulation experiments demonstrate that ITDD reduces data volume by three orders of magnitude and improves the SNR by 4.21 times. On the SITP-QLEF dataset, the network achieves a detection rate of 99.31%, and a false alarm rate of $1.97\times {10}^{-5}$, confirming its effectiveness and application potential under complex backgrounds. As the current findings are based on simulated data, future work will focus on building an ITDD demonstration system to validate the approach with real-world IR measurements.

Detecting dim and small targets in complex scenes remains challenging due to low signal-to-noise ratio (SNR), weak texture, and cluttered backgrounds. Most existing methods are based on infrared images from frame-based cameras, with limited exploration of event-based approaches despite their potential advantages, such as high temporal resolution, low latency, and high dynamic range. In this paper, we propose the sparse convolutional spatio-temporal attention network (SC-STANet), a lightweight and efficient end-to-end framework for detecting dim targets from asynchronous event streams. Representing event data as 3D point clouds, our method preserves fine-grained spatio-temporal structures and leverages a hybrid sparse convolution (HSC) network for robust feature extraction. We further introduce a spatio-temporal attention module that independently captures spatial saliency and temporal consistency, enhancing target features while suppressing noise. In addition, a feature memory buffer is incorporated to reuse temporal features, significantly improving inference speed. Extensive experiments on our differential-image event synthesis dataset demonstrate that SC-STANet outperforms state-of-the-art single-frame and multi-frame methods in terms of both detection accuracy and computational efficiency, achieving an F1-score of 97.19%. Moreover, on the real-world event dataset EV-UAV, our model attains an F1-score of 99.49%, demonstrating its strong generalization capability and practical effectiveness in real-world scenarios.