Infrared small target detection faces the inherent challenge of precisely localizing dim targets amidst complex background clutter. Traditional approaches struggle to balance detection precision and false alarm rates. To break this dilemma, we propose SeRankDet, a deep network that achieves high accuracy beyond the conventional hit-miss trade-off, by following the “Pick of the Bunch” principle. At its core lies our selective rank-aware attention (SeRank) module, employing a nonlinear Top-K selection process that preserves the most salient responses, preventing target signal dilution while maintaining constant complexity. Furthermore, we replace the static concatenation typical in U-Net structures with our large selective feature fusion (LSFF) module, a dynamic fusion strategy that empowers SeRankDet with adaptive feature integration, enhancing its ability to discriminate true targets from false alarms. The network’s discernment is further refined by our dilated difference convolution (DDC) module, which merges differential convolution aimed at amplifying subtle target characteristics with dilated convolution to expand the receptive field, thereby substantially improving target-background separation. Despite its lightweight architecture, the proposed SeRankDet sets new benchmarks in state-of-the-art performance across multiple public datasets. The code is available athttps://github.com/GrokCV/SeRankDet.

Single-frame infrared small target (SIRST) detection has been a challenging task due to a lack of inherent characteristics, imprecise bounding box regression, a scarcity of real-world datasets, and sensitive localization evaluation. In this article, we propose a comprehensive solution to these challenges. First, we find that the existing anchor-free label assignment method is prone to mislabeling small targets as background, leading to their omission by detectors. To overcome this issue, we propose an all-scale pseudobox-based label assignment scheme that relaxes the constraints on the scale and decouples the spatial assignment from the size of the ground-truth target. Second, motivated by the structured prior of feature pyramids, we introduce the one-stage cascade refinement network (OSCAR), which uses the high-level head as soft proposal for the low-level refinement head. This allows OSCAR to process the same target in a cascade coarse-to-fine manner. Finally, we present a new research benchmark for infrared small target detection, consisting of the SIRST-V2 dataset of real-world, high-resolution single-frame targets, the normalized contrast evaluation metric, and the DeepInfrared toolkit for detection. We conduct extensive ablation studies to evaluate the components of OSCAR and compare its performance to state-of-the-art model- and data-driven methods on the SIRST-V2 benchmark. Our results demonstrate that a top-down cascade refinement framework can improve the accuracy of infrared small target detection without sacrificing efficiency. The DeepInfrared toolkit, dataset, and trained models are available athttps://github.com/YimianDai/open-deepinfrared.

Although Cramér-Rao Bounds (CRB) for angle-of-arrival (AOA) based position estimation have been extensively studied for decades, the existing works mainly focused on target direction finding errors, neglecting the deviations of mutual measurements in collaborative environment of surface ships. The deviations of mutual direction-finding and ranging between cooperative ships should be taken into account for source localization using mutual measurements in collaborative environment of surface ships. In this paper, we derive quantitative expressions of the CRB on the positioning accuracy while considering the deviations of mutual measurements, and calculate the lower bound on the circular error probability (CEP) of the cooperative localization system. The derived bounds are simple to evaluate and provide a good prediction of the actual AOA based cooperative localization system performance in collaborative environment of surface ships. Numerical examples illustrate our theoretical results.

Azimuth periodically gapped sampling occurs in satellite Synthetic Aperture Radar (SAR) due to new radar mission or sensor geometries. For resolving this issue, a novel strategy is presented in this paper for obtaining high-resolution satellite SAR image via processing azimuth periodically gapped echo data by segmented recovery. To avoid azimuth frequency aliasing, the proposed method splits echo data into several sub-blocks in azimuth direction. Then, sub-blocks are compensated with curved orbit phase compensation function and recovered with iterative adaptive approach. Afterwards, sub-blocks are combined as the whole echo data in time domain. Subsequently, the recovered data is processed by curved orbit deramping-based approach and modified range migration algorithm. The proposed strategy is validated with simulation and actual SAR data experiment.

Existence of periodical interruption in Synthetic Aperture Radar (SAR), which is caused by various avenues, leads to challenges on SAR data imaging. To resolve this issue, a novel approach is developed in this paper. The proposed approach deal with periodically interrupted raw SAR data by complex deconvolution. Complex Fast Iterative Shrinkage-Thresholding Algorithm (FISTA) is operated to perform complex deconvolution and restore the azimuth spectrum from periodically interrupted SAR echo. Then, conventional SAR focusing method is able to attain imaging result from interrupted echo data. The proposed method processes interrupted echo with lower fake targets and operates more quickly than approach on basis of Iterative Shrinkage-Thresholding Algorithm (ISTA). Experiments on simulation data and actual SAR echo illustrate the effectiveness of the proposed approach.

A novel algorithm is proposed in this paper to focus the azimuth periodically gapped Synthetic Aperture Radar (SAR) raw data with suppressed artificial targets. The proposed method mainly consists of phase compensation in range frequency domain, subsequent sparse recovery via complex Fast Iterative Shrinkage-Thresholding Algorithm (FISTA) and focusing recovered SAR data with traditional SAR imaging algorithm. With phase compensation, the gapped data becomes sparser in range Doppler domain and complex FISTA is utilized to restore complete data from gapped data in range Doppler domain. Afterwards, the conventional SAR focusing algorithms are capable of coping with the recovered data. The proposed method focuses azimuth gapped data with lower artificial targets and performs faster than method based on complex Iterative Shrinkage- Thresholding Algorithm (ISTA). In experiments, point target simulation and real SAR data processing are utilized to demonstrate the validity of the proposed method.

Algorithms for high resolution spotlight Synthetic Aperture Radar (SAR) focusing typically face challenges of curved orbit and azimuth aliasing. To conquer these difficulties, a novel algorithm, which is named as Modified Generalized Omega-K (MGOK) algorithm, is proposed in this paper for focusing low-orbit high resolution spotlight spaceborne Synthetic Aperture Radar (SAR). The proposed method modifies the Generalized Omega-K algorithm with orbital state vectors and deramping-based approach in order to accommodate the scenario of Low Earth Orbit (LEO) high resolution spaceborne SAR. Experiments are performed on point target simulation for assessing the effectiveness of the proposed method. In simulation, the range and azimuth resolution can both obtain 0.5m with scene size of 4km by 4km.

Video synthetic aperture radar (VideoSAR) possesses the capability of imaging and continuously monitoring the scenario from a wide-aspect interval for enhancing the performance of information interpretation. In this paper, we propose a complete imaging scheme to achieve the unified video coordinate system in high-resolution airborne VideoSAR configuration. Comprehensive postprocessing video imaging (PPVI) framework built on range Doppler algorithm and range migration algorithm is elaborated especially in terms of complex measured data, which is divided into three parts for ensuring the stability of video background: full-aperture imaging, 2-D autofocus technique, and Doppler spectrum segmentation. Experimental results utilizing the measured airborne data have demonstrated the effectiveness of PPVI scheme for sequential VideoSAR formation.

As one of the major search engines in the world, Baidu's Sponsored Search has long adopted the use of deep neural network (DNN) models for Ads click-through rate (CTR) predictions, as early as in 2013. The input futures used by Baidu's online advertising system (a.k.a. "Phoenix Nest'') are extremely high-dimensional (e.g., hundreds or even thousands of billions of features) and also extremely sparse. The size of the CTR models used by Baidu's production system can well exceed 10TB. This imposes tremendous challenges for training, updating, and using such models in production. For Baidu's Ads system, it is obviously important to keep the model training process highly efficient so that engineers (and researchers) are able to quickly refine and test their new models or new features. Moreover, as billions of user ads click history entries are arriving every day, the models have to be re-trained rapidly because CTR prediction is an extremely time-sensitive task. Baidu's current CTR models are trained on MPI (Message Passing Interface) clusters, which require high fault tolerance and synchronization that incur expensive communication and computation costs. And, of course, the maintenance costs for clusters are also substantial. This paper presents AIBox, a centralized system to train CTR models with tens-of-terabytes-scale parameters by employing solid-state drives (SSDs) and GPUs. Due to the memory limitation on GPUs, we carefully partition the CTR model into two parts: one is suitable for CPUs and another for GPUs. We further introduce a bi-level cache management system over SSDs to store the 10TB parameters while providing low-latency accesses. Extensive experiments on production data reveal the effectiveness of the new system. AIBox has comparable training performance with a large MPI cluster, while requiring only a small fraction of the cost for the cluster.