In this paper, we propose a novel feature learning framework for video person re-identification (re-ID). The proposed largely aims to exploit the adequate temporal information of sequences and tackle poor spatial alignment moving pedestrians. More specifically, exploiting information, design residual (TRL) module simultaneously extract generic specific features consecutive frames. TRL is equipped with two bi-directional LSTM (BiLSTM), which are respectively responsible describe in different...

Transforming infrastructures, buildings and services with the sensed data from Internet of Things (IoT) technique has drawn wide attention. Enormous video city surveillance cameras poses huge challenges transmission, storage analysis, which necessitates new compression technologies. The fusion generated smart could be used to support management urban policy. Based on specific characteristics video, are successive pictures have very strong correlations each picture can divided into background...

One impressive advantage of convolutional neural networks (CNNs) is their ability to automatically learn feature representation from raw pixels, eliminating the need for hand-designed procedures. However, recent methods single image super-resolution (SR) fail maintain this advantage. They utilize CNNs in two decoupled steps, i.e., first upsampling low resolution (LR) high (HR) size with techniques (e.g., bicubic interpolation), and then applying on upsampled LR reconstruct HR results. In...

In response to the annual occurrence of over 10 million gastrointestinal endoscopic examinations in China, this study proposes a personalized anesthesia management model based on multimodal deep learning. This was designed enhance efficiency and disease detection rates. collaboration with Department Anesthesiology at Renji Hospital, which is affiliated Shanghai Jiao Tong University School Medicine, data pertaining were collected from 398 patients, who undergoing endoscopy. yielded total 327...

Appearance-based human re-identification is challenging due to different camera characteristics, varying lighting conditions, pose variations across views, etc. Recent studies have revealed that color information plays a critical role on performance. However, two problems remain unclear: (1) how do descriptors perform under the same scene in problem? and (2) can we combine these without losing their invariance property distinctiveness power? In this paper, propose novel ensemble model...

ABSTRACT A liquid silicon/phosphorus containing flame retardant (DOPO–TVS) was synthesized with 9,10‐dihydro‐9‐oxa‐10‐phosphapheanthrene‐10‐oxid (DOPO) and triethoxyvinylsilane (TVS). Meanwhile, a modified epoxy resin (IPTS–EP) prepared by grafting isocyanate propyl triethoxysilane (IPTS) to the side chain of bisphenol (EP) through radical polymerization. Finally, incorporated into sol–gel reaction between ethyoxyl two intermediates obtain resin. The molecular structures DOPO–TVS, IPTS–EP...

Object detection and recognition is the premise foundation for intelligent service robot to understand surrounding environment make decisions. In this paper, aiming at accuracy real-time performance of object in complex scenes, an end algorithm based on deep learning proposed. Firstly, local multi branch convolution neural network adopted enhance feature representation capability model by enhancing module function. Then, combining anchor point mechanism, class position regression prediction...

Phosphoric triamide (PTA) and glycidyl polyhedral oligomeric silsesquioxane (POSS) were simultaneously incorporated into the cured network of a bisphenol F epoxy resin 4,4′-diaminodiphenyl methane (DDM) to improve thermal stability flame retardancy. PTA was synthesized by triethyl phosphate DDM, its chemical structure confirmed 1 H nuclear magnetic resonance (NMR) Fourier transform infrared (FTIR). The differential scanning calorimetric (DSC) results showed that introduction POSS slightly...

In this paper we present a novel combined tracking algorithm based on moving object detection technology. Particle filtering can work well when the gets an occlusion, it has difficulty in satisfying requirement of real-time computing. Meanshift solve problem easily, poor rubustness during mutual occlusion. meantime, because backgrounds many scenes include complex objects, methods only using color, texture and shape feature often have rubustness. Aiming at all above problems, detect codebook...

We present a novel visual object tracking algorithm based on two-dimensional principal component analysis (2DPCA) and maximum likelihood estimation (MLE). Firstly, we introduce regularization into the 2DPCA reconstruction develop an iterative to represent by bases. Secondly, model of sparsity constrained MLE is established. Abnormal pixels in samples will be assigned with low weights reduce their effects algorithm. The results are obtained using Bayesian posteriori (MAP) probability...

Abstract The build-up rate prediction is of great importance for trajectory control in the field drilling. However, it very difficult to achieve accurate due complexity, nonlinearity, and multiple uncertainties drilling system. As a consequence, novel hybrid model proposed, which uses feature selection methods, combination strategy based on machine learning, three models improve accuracy rate. More precisely, correlation analysis, statistical analysis are employed ensure effectiveness...

RGBT tracking has been widely used in various fields such as robotics, surveillance processing, and autonomous driving. Existing trackers fully explore the spatial information between template search region locate target based on appearance matching results. However, these have very limited exploitation of temporal information, either ignoring or exploiting it through online sampling training. The former struggles to cope with object state changes, while latter neglects correlation...