BINGBING DAN | 淡 冰 冰

In infrared dim-small target detection, large background group and small clutter group are the key components. However, existing methods usually consider the detection progress in the original image space, which limits the separability of the target from the two components and leads to missed detection and false alarms. In response to this issue, we propose an innovative infrared dim-small target detection method via chessboard topology, which mines potential differences, such as distribution density and scale trends in the topological space. Specifically, the core of our approach lies in the construction of the chessboard topology space, where each ”point set” serves as a basic unit that is a mapping result of pixels in the original image space. The chessboard’s horizontal divisions are based on the scale space, where pixels undergo multiscale transformations to emphasize scale invariance at smaller scales, resulting in rows that capture scale variation trends. Meanwhile, the vertical divisions are based on the gray space, with pixels rearranged to accentuate gray level variations, thereby forming columns that highlight distribution density disparities. To separate the target pixels, we design two complementary strategies for pixel scoring within the chessboard topological space. The first, scoreS, evaluates pixel consistency across multiple scales, aiming to eliminate inconsistent pixels that often represent false positives. The second, scoreL, focuses on measuring the density level of point sets to enhance target visibility by filtering out pixels within less dense point sets. The final detection results are derived from the dot product of these two scores, ensuring a robust differentiation of small targets from background and noise. Comprehensive experiments demonstrate that the proposed method achieves better performance than baselines in six real infrared dim-small target scenarios.

Small target detection against strong vignetting background poses two major challenges: the submergence of the target in high brightness background and the difficult separability between dense noise and weak target. These challenges have not been adequately addressed by existing methods, which heavily rely on the low-rank prior of background. In this paper, our focus shifts to the smoothness of background and propose a dynamic weight guided smooth-sparse decomposition model (DW-SSD). First, we adopt a surface-fitting perspective to reconstruct the smooth vignetting, where B-spline is introduced as the basis, and total variation (TV) is utilized to regularize the coefficients of the B-spline basis. This term provides a flexible description for different background local regions with low computation cost, effectively suppressing the high brightness vignetting and facilitating the emergence of target. Second, a dynamic weight based on spatial-temporal prior characteristic is added to the target component, where the differences in spatial contrast and temporal profile are jointly considered to guide a precise separation of the target from dense noise. Finally, a customized solution is designed based on the alternating direction multiplier method (ADMM) to separate small targets from strong vignetting rapidly and precisely. Comprehensive experiments on five real sequences demonstrate that DW-SSD method effectively overcomes the challenges posed by strong vignetting and performs well in real-time small target detection.

Stochastic resonance (SR) is usually used to enhance the signal with the help of noise. Inspired by this, we find that the SR can also handle the problem of dim-small target detection under the low local signal-to-noise ratio (LSNR) situation. In this letter, we propose a novel spatial–temporal SR (STSR) model for dim-small target detection. First, we select the SR as the core model to enhance the salience of the target by the inherent strong noise. With the help of the Poisson distribution prior, we use the multiple adjacent frames as the input of the SR model, improving the LSNR of the resonance state through the temporal accumulation of photons. Then, we introduce the total variation (TV) regularization in the variational framework to remove the false alarm points by spatial smoothing, while preserving the role of noise in SR. Finally, we customize an optimization process based on the alternating direction method of multiplier (ADMM) to solve the STSR variational minimization problem. Both the qualitative and quantitative experiments on real visible and infrared image sequences have demonstrated the superiority of the proposed model, especially in the low LSNR situation below 2 dB.

Star identification plays a key role in spacecraft attitude measurement. Currently, most star identification algorithms tend to perform well only in a scene without noise and are highly sensitive to noise. To solve this problem, this paper proposes a star identification algorithm based on the maximum spanning tree (MST) index and multi-order continuous cycle angle (CCA) intended for the lost-in-space mode. In addition, a neighboring star selection method named dynamic eight-quadrant (DEQ) is developed. First, the DEQ method is used to select high-confidence neighboring stars for the main star. Then, the star image is regarded as a graph, and the Prim algorithm is employed to construct the MST pattern for each guide star, which is then combined with the K vector index to perform the main star candidate search. Finally, the Jackard similarity voting for the multi-order CCA of the main star is used to identify the main star, and the global neighboring star identification is conducted by the multi-order CCA of neighboring stars. The simulated and real star images test results show that compared with five mainstream algorithms, when the position noise is 1 pixel, the number of false stars is five, the magnitude noise is 0.5, and the identification accuracy of the proposed algorithm is higher than 98.5%. Therefore, the proposed algorithm has excellent anti-noise ability in comparison to other algorithms.

The construction of a guide star catalog is crucial for a star sensor to achieve accurate star map recognition and attitude determination. At present, the methods of a guide star catalog for a large field of view (FOV) star sensor have been relatively mature. However, for a small-FOV star sensor, there are still certain problems, such as a large storage capacity of a guide star catalog, uneven distribution of stars, and easy occurrence of voids. To address these problems, we propose a construction method of a small-FOV star sensor guide star catalog based on the combination of the improved spherical spiral method (ISSM) and extreme learning machine (ELM) named the ISSM-ELM. First, a spiral reference point is used as an optical axis pointing of the star sensor, and the guide stars are preliminarily screened based on the star-diagonal distance between the star and the reference point, and the star-density and magnitude characteristics of the guide star. Then the ELM is used to supplement the guide star empty sky area to construct an integrity guide star catalog. The experimental results demonstrate that the proposed method can reduce the storage capacity of the guide star catalog and improve its uniformity, integrity, and average brightness.