Video anomaly detection is of critical practical importance to a variety real applications because it allows human attention be focused on events that are likely interest, in spite an otherwise overwhelming volume video. We show applying self-trained deep ordinal regression video overcomes two key limitations existing methods, namely, 1) being highly dependent manually labeled normal training data; and 2) sub-optimal feature learning. By formulating surrogate two-class task we devise...

Person Re-IDentification (ReID) aims at re-identifying persons from different viewpoints across multiple cameras. Capturing the fine-grained appearance differences is often key to accurate person ReID, because many identities can be differentiated only when looking into these differences. However, most state-of-the-art ReID approaches, typically driven by a triplet loss, fail effectively learn features as they are focused more on differentiating large To address this issue, we introduce...

Occluded person re-identification (ReID) aims at re-identifying occluded pedestrians from or holistic images taken across multiple cameras. Current state-of-the-art (SOTA) ReID models rely on some auxiliary modules, including pose estimation, feature pyramid and graph matching to learn multi-scale and/or part-level features tackle the occlusion challenges. This unfortunately leads complex that (i) fail generalize challenging occlusions of diverse appearance, shape size, (ii) become...

Background: Microbial communities have important influences on our health and disease. Identifying potential human microbe-drug associations will be greatly advantageous to explore complex mechanisms of microbes in drug discovery, combinations repositioning. Until now, the mechanism remains unknown. Objective: Computational models play an role discovering hidden associations, because biological experiments are time-consuming expensive. Based chemical structures drugs KATZ measure, a new...

Person Re-IDentification (ReID) aims at re-identifying persons from non-overlapping cameras. Existing person ReID studies focus on horizontal-view tasks, in which the images are captured by cameras a (nearly) horizontal view. In this work we introduce new task, bird-view ReID, searching for gallery of with query taken bird's-eye view, i.e., an elevated view object above. The task is important because there large number video surveillance capturing such public places. However, it challenging...

In recent years, more attention has been attracted on the classification of polarimetric SAR (PolSAR) images and a lot methods have proposed. With resolution increasing, pixel-based reveal insufficiency, therefore, this paper proposes an improved SLIC superpixel algorithm for PolSAR classification. Firstly, effective polarization features, such as scattering power typical mechanism based target decomposition spatial texture information statistical analysis, are extracted from these features...

In this paper, we utilize deep reinforcement learning algorithm Soft Actor-Critic (SAC) to solve the autonomous decision-making problem of Unmanned Aerial Vehicle (UAV). Firstly, UAV is abstracted into a game scenario, and simulation environment built based on tensorflow, pygame, etc. Secondly, modeled as Markov Decision Process (MDP), framework constructed, SAC connected with environment. Finally, model begins be trained, so that learns actions task, finds optimal strategy, generates...

The synthetic aperture radar scale-invariant feature transform (SAR-SIFT) algorithm has been applied to (SAR) image registration and good performance. However, compared with single polarization images, polarimetric (PolSAR) images have more combinations, which can obtain information. Therefore, based on the scattering mechanism of PolSAR, this paper proposes a suitable for PolSAR images. improves SAR-SIFT uses information introduce new gradient definition construct scale spaces. proposed in...

This study examines the influence of ligands with different N-substitution modes and heavy atom effects on transition metal complexes.

In order to effectively assist the researchers conduct quantitative analysis of ceramic microstructures, a segmentation algorithm based on mean shift is used for microstructure image. Since collection and transfer process microscopic image will inevitably be subject uneven distribution light, electronic noise other interference factors which make quality deterioration, it necessary reduce noises enhance edges processing at first. Therefore, median filter remove in images. Then component with...

The adversarial example attack on SAR images has been widely used in military confrontation recent years. However, many existing generation methods still exist problems such as excessive perturbation, unstable training, and susceptibility to detection. In response the above issues, a method based Grad-CAM mask extraction proposed, which uses algorithm extract masks from images, optimizes pixel points for search. Finally, examples generated by proposed have structural similarity of over 0.99...

High-resolution PolSAR images are wildly used in land cover mapping. However, there two problems when working with large areas, i.e., long processing time and computer memory. Considering that properties of homogeneous area, such as forest, farm land, bare etc., almost the same under different resolutions, which means these areas can be processed relatively low-resolution without decreasing mapping accuracy, this paper proposes a new method, to shorten reduce demand memory high resolution...

Abstract the Task of Robot Path Planning Based on Artificial Intelligence Algorithm is to Determine an Optimal Strategy, with Goal Maximizing Return. When in a New State, It Needs Select Action from Set and Execute It. Order Improve Quality Algorithm, Necessary Apply Search Strategy Selection. System Robot, One Most Critical Links Realization Directional Reference Vehicle Scheduling. Generally, Widely Used Planning. This Paper, Optimize Scheduling Problem. Variable Problem Can Be Defined...

PolSAR images have been used extensively for various surface features recognition and buildings is an important research topic of image interpretation. Traditional methods are only based on characteristics lack subjective knowledge human cognition, making low target rate algorithm redundancy. To overcome this shortcoming, cognition mechanism, a new method in proposed paper. The utilizes hierarchical model to identify buildings: the first layer visual second logical cognition. In sensitive...

Abstract. How to obtain accurate difference map remains an open challenge in change detection. To tackle this problem, we propose a detection method based on saliency and wavelet transformation. We do frequency-tuned initial image (IDI) obtained by logarithm ratio get salient (SDI). Then, calculate local entropy of SDI entropic (ESDI). The final (FDI) is the fusion IDI ESDI, Otsu thresholding used extract from FDI. Experimental results validate effectiveness feasibility.

This paper proposes a geometric correction method for vehicle chassis images acquired by linear charge coupled device (CCD) camera using speeded up robust features (SURF) and phase-only correlation (POC). We use feature matching based in SURF to estimate an approximate homography between standard real-time limited number of corresponding points, dense sub-pixel correspondence search POC adjust precise position points. Experimental result shows the can achieve high-accuracy.