Abstract For the extremely small size and low signal‐to‐clutter ratio, target detection in infrared images is still a considerable challenge. Specifically, it very difficult to detect point targets because there no texture shape information can be used. A target‐oriented shallow‐deep feature‐based algorithm proposed, opening up promising direction for convolutional neural network‐based dim algorithms. To ensure that instances used correctly networks, effective anchor designed according...

The detection of infrared small targets under low signal-to-clutter ratio (SCR) and complex background conditions has been a challenging popular research topic. In this article, spatial-temporal feature-based framework is proposed. First, several factors, such as the target's sample, sensitive size, usual sample selection strategy, that affect are analyzed. addition, intersection over union (IOU) which helps to solve false convergence misjudgment problem, Second, aiming at difficulties due...

To address the phenomenon of many small and hard-to-detect objects in drone images, this study proposes an improved algorithm based on YOLOv7-tiny model. The proposed assigns anchor boxes according to aspect ratio ground truth provide prior information object shape for network uses a hard sample mining loss function (HSM Loss) guide enhance learning from samples. This finds that difference vehicle under perspective is more obvious than scale difference, so assigned by can effective those...

In recent years, due to its strong nonlinear mapping and research capacities, the convolutional neural network (CNN) has been widely used in field of hyperspectral image (HSI) processing. Recently, pixel pair features (PPFs) spatial PPFs (SPPFs) for HSI classification have served as new tools feature extraction. this paper, on top PPF, improved subtraction (subtraction-PPFs) are applied target detection. Unlike original PPF SPPF, subtraction-PPF considers classes afford CNN, a detection...

The detection and recognition of dim small infrared (IR) targets across domains pose two formidable challenges: distributional discrepancies in samples scarcity or absence annotated instances the target domain. While current unsupervised domain adaptive object methods can somewhat alleviate performance degradation caused by these issues, they fail to address differences semantic content between background environments different application scenarios. This results a gap that impedes...

The infrared small target detection technology has a wide range of applications in maritime defense warning and border reconnaissance, especially the sky scenes for detecting potential terrorist attacks monitoring borders. However, due to weak nature targets presence background interferences such as wave reflections islands scenes, are easily submerged background, making hard detect. We propose multidimensional information fusion network(MIFNet) that can learn more from limited data achieve...

This paper proposes a hyperspectral target detection framework with convolutional neural network (CNN). The number of training samples is first sufficiently enlarged by subtraction method to maximize the advantages multilayer CNN. Next, CNN given function labelling new pixels subtracted between and background classes as 1, within both same different 0. Finally, for each testing pixel, difference central pixel its adjacent input into framework. If belongs target, output score close label....

This paper proposes an improved Mask RCNN and LMB algorithm for target detection tracking in complex backgrounds, which follows the architecture of Detect-Before-Track. First, novel adopts a two-stage neural network to improve positions' accuracy during detection. Then, label multi-Bernoulli filter, is suitable scenarios with unknown number targets intersection, utilized generate multiple-target trajectories stage. Experiments suggest algorithm's effectiveness superiority backgrounds'...

The detection and tracking of small targets under low signal-to-clutter ratio (SCR) has been a challenging task for infrared search track (IRST) systems. Track-before-detect (TBD) is widely-known algorithm which can solve this problem. However, huge computation costs storage requirements limit its application. To address these issues, dynamic programming (DP) multiple hypothesis testing (MHT)-based dim point target (DP–MHT–TBD) proposed. It consists three parts. (1) For each pixel in current...

Traditional data-driven algorithms suffer from data reliance, hyperparameter sensitivity, and faint characteristics in infrared (IR) "low, slow, small" unmanned aerial vehicle target detection recognition, resulting performance degradation complex backgrounds. Inspired by model-driven methods, this article proposes a learnable feature modulation module that uses prior knowledge to enhance representation. Specifically, method converts the local contrast measure into nonlocal quadrature...

In the field of infrared, detection dim small target has been a challenging topic. Especially in complex background, large number false alarms might appear when using traditional methods, this cannot meet dual requirements high rate and low alarm rate. Therefore, paper, CNN based method is proposed for infrared sequence. For an image to be detected, firstly, aiming at noise interference which are spatially non-stationary but temporally stable, registration frame difference used suppress...

An infrared small target detection and tracking method suitable for different scenes is proposed. For a given image sequence, 1) firstly, the complexity of scene judged according to number Harris corners; 2) sequence containing simple scenes, local adaptive threshold segmentation track association used complete multi-target; 3) complex SSD detector detect initial target, after detected, SiamRPN++ tracker started target. The result in process. When lost or wrong, restarted immediately capture...

Recently, the convolutional neural network (CNN) has been widely used in fields of hyperspectral image (HSI) processing. In this paper, a CNN-based target detection framework is presented. And subtraction model to sufficiently enlarge number training samples. The built from twenty-eight manually selected objects several AVIRIS date following three aspects: 1) new pixel made by any two pixels between 27 different classes labelled as 0; 2) within per class 3) pixels, which one and other...

Multiple hypotheses testing (MHT)-based Track-before-Detect (TBD) method is a widely known algorithm to detect infrared dim small target. However, the space in this kind of so big that computation cost and storage requirement inevitably increase, resulting hardly realized hardware. To solve problem, parallel two stage multiple model proposed. First, MHT designed save possible trajectories. By using proposed MHT, all trajectories can be parallelly processed search process simplified, greatly...

In the field of terrorism risk, terrorist attacks against urban targets have been a hot research topic. this paper, quantitative risk assessment and mitigation framework which consists attack security strategy optimization is proposed. First, conditional probability based description function constructed for loss assessment. Second, minimum to optimize anti-terrorist strategies. On these bases, established. To verify effectiveness proposed method, railway station used as an example analyze...