Accurate depth maps are essential for indoor navigation and modeling by robots, but raw depth maps often have missing areas due to sensor limitations, environmental factors, and distance constraints. Existing methods that fuse RGB images with depth maps usually cannot utilize spatial structural information and exhibit poor accuracy at object edges. To bridge this gap, a dual-branch fusion network with Mamba decoder, called DBFNM, is proposed for depth completion in this work. It consists of two complementary branches: one branch utilizes semantic and texture information from RGB images as visual guidance, while the other extracts spatial geometric structures from normal maps as structural guidance. In particular, a geometric gated encoder is utilized to fully leverage spatial information. In the dual-branch decoding stage, a dual-branch feature interaction alignment module is designed, which is composed of three components, including dual-branch edge feature alignment, dual-branch interaction, and global alignment. Then, the decoded dual-branch features are processed by a dual-modal fusion network based on a spatial propagation network to obtain dense depth map predictions. Extensive experimental results on the NYU-Depth V2 and SUN RGB-D datasets demonstrate that DBF achieves superior depth completion performance compared to existing methods in indoor scenes, particularly in handling large-scale missing depth regions and preserving edge details.

In the context of multiobject tracking using video synthetic aperture radar (Video SAR), Doppler shifts induced by target motion result in artifacts that are easily mistaken for shadows caused by static occlusions. Moreover, appearance changes of the target caused by Doppler mismatch may lead to association failures and disrupt trajectory continuity. A major limitation in this field is the lack of public benchmark datasets for standardized algorithm evaluation. To address the above challenges, we collected and annotated 45 video SAR sequences containing moving targets, and named the video SAR MOT benchmark (VSMB). Specifically, to mitigate the effects of trailing and defocusing in moving targets, we introduce a line feature enhancement mechanism that emphasizes the positive role of motion shadows and reduces false alarms induced by static occlusions. In addition, to mitigate the adverse effects of target appearance variations, we propose a motion-aware clue discarding mechanism that substantially improves tracking robustness in video SAR. The proposed model achieves state-of-the-art performance on the VSMB, and the dataset and model are released athttps://github.com/softwarePupil/VSMB

Data annotation is extremely difficult due to the satellite imaging conditions, under which the targets are usually small, obscured, and scattered. Therefore, satellite video data annotation inevitably has noise and errors. Moreover, the irregular acceleration, sudden turns, and stops of the target make prediction and trajectory maintenance highly challenging. In this study, we propose the Points of Interest Flows Track (PIFTrack) to address the aforementioned challenges. PIFTrack improves tracking accuracy by modeling target uncertainty distributions and nonlinear motion patterns, while leveraging the spatial inclusion relationships of points of interest (PoIs) across consecutive frames. Specifically, we eliminate the rigid Dirac-based labeling assumption by employing a set of PoIs to model the spatial probability distribution of the target. PoIs enable the model to infer optimal outputs in the vicinity of annotations, thereby improving robustness to annotation errors. Secondly, to capture the real motion transfer patterns of targets in the data, we introduce a diffusion-based ordinary differential equation (ODE) model. Ultimately, we alleviate the impact of tiny object localization drifts on association results by exploiting the inclusion relationship between PoIs. PIFTrack has been extensively evaluated on the VISO, AIR-MOT, CGSTL, and VSMB datasets, exhibiting competitive performance relative to contemporary studies. Our code is open-source and available at https://github.com/softwarePupil/PIFTrack.

Vehicle detection based on remote sensing images is widely used in urban traffic management and disaster rescue. RGB images, which are used more, lead to poor detection performance in low light conditions due to the imaging mechanism. At present, the main solution is to improve the detection performance in low light by fusing with infrared images. However, the current methods often overlook the impact of illumination changes on RGB images, and ignore the important role of high-frequency information for object detection, especially for low-light target detection. In this paper, we propose a Cross-modality and Cross-domain Low-light Detector (CCLDet) for low-light vehicle detection, including three improvements. First, an object illumination-aware module (OIAM) is proposed, which can adjust adaptively the weight of different modalities according to the object illumination intensity in the training phase and enables the detector to adapt to different lighting conditions. Second, we propose a visibility loss, which converts the position deviation into the illumination intensity deviation of each point in the object area. Compared with relying only on semantic information for object localization, the illumination makes the information that can be used for localization more abundant. Third, we design a cross-domain feature fusion module (CDFFM), which can enhance high-frequency features and enrich target information when low-frequency features are lost due to low light pollution. Extensive experiments on three challenging RGB-infrared objects detection datasets demonstrated the mAP and the parameter quantities of CCLDet over popular object detectors.

With the rapid advancement of satellite technology, satellite video has emerged as a key method for acquiring dynamic terrestrial information, facilitating multiple object tracking (MOT). Satellites are capable of surveying vast urban landscapes, yet the observed objects are small and dispersed among complex interference from the background, heightening the challenges in detection and association tasks for object tracking. However, current trackers often dissociate the classification task from the localization task, leading to drift in tiny object detection (TOD), and rely on prior knowledge for clue ranking, limiting model robustness. In this article, we introduce the graph-based multiclue fusion tracker (GMFTracker). Initially, we introduce a sparse sampling-based feature map correction approach to rectify the misalignment between the classification and localization feature maps. Furthermore, we developed graph neural networks (GNNs) for object relationship modeling, free from presuppositions, to tackle association challenges using relational features. GMFTracker was rigorously tested on VISO, CGSTL, and TinyPerson datasets, demonstrating its competitive performance relative to contemporary studies.

Oriented object detection in aerial images is a crucial link in earth observation. As a special parameter in oriented object representation, angle is the key to achieving high-precision detection. However, the widely-used regression-based methods suffer from boundary discontinuity problem due to the periodicity of angle. To address this issue, we proposed a novel angle prediction method called Fixed Step Trigonometric Coder (FSTC). Exploiting the innate periodicity of trigonometric functions, FSTC can encode angles cyclically in a succinct, continuous, and uniform manner. Based on FSTC, we designed a single-shot oriented object detector, namely, Trigonometric-coded Refined Detector (TRDet), for high-precision object detection in real-time. TRDet consists of two modules: the Angle Optimization Module (AOM) and the Object Detection Module (ODM). AOM employs FSTC to generate high-quality rotated anchors. In ODM, a Dynamically Weighted Loss (DWL) was proposed to make the model focus on hard samples with higher angle deviation. Extensive experiments on DOTA and HRSC2016 show that both FSTC and TRDet can achieve competitive performance compared with peer works.

Oriented object detection in optical remote-sensing images has been a challenging task due to arbitrary orientations and densely packed distribution of objects. Specifically, most existing methods lack adaptivity when regressing objects with different shapes and orientations. Although the point set representation is relatively flexible, the initial distribution of the point set is fixed in advance. In addition, some models based on the point set cannot get high location precision of points, affecting the bounding box generation. In this letter, we propose an Adaptive Point Set Network (APS-Net) for optical remote-sensing object detection, including three improvements. First, we propose the initial distribution learner (IDL) to learn the optimal initial aspect ratio, which helps the point set fit the object’s shape well. Second, we design the uncertainty measurement module (UMM), which considers the uncertainty of point location to improve location precision. Third, we introduce the local outlier factor (LOF) in the loss to punish outlier points more reasonably. Extensive experiments demonstrate that our proposed model achieves state-of-the-art performance on three commonly used datasets (i.e., DOTA-v1.0, UCAS-AOD, and HRSC2016) in the remote-sensing field.