Hand gesture recognition is an intuitive and effective way for humans to interact with a computer due its high processing speed accuracy. This paper proposes novel approach identify hand gestures in complex scenes by the Single-Shot Multibox Detector (SSD) deep learning algorithm 19 layers of neural network. A benchmark database used, general scene are chosen as objects. real-time system based on SSD constructed tested. The experimental results show that quickly identifies humans’ hands...

A multifunctional fluorescent probe P based on a naphthalimide derivative for the detection of Ag+ and Hg2+ through dual-signal was designed characterized. exhibited large Stokes shift (107 nm), high selectivity, good sensitivity, fast response time. By adjusting testing medium order reagent addition, with achieved. The addition or to solution in either ethanol an ethanol–water mixture resulted significant quenching fluorescence emission at 537 nm caused decrease absorbance 440 accompanied...

The reduction in rheological parameters and dissolution rate of precursors geopolymer coatings during early hydration significantly contributes to sagging. This study aims improve the sag resistance these by incorporating diatomite filler. Rheological testing was conducted assess impact its concentration on yield stress, plastic viscosity, thixotropy coatings. results indicated that diatomite's large specific surface area high reactivity have a significant influence precursors. With...

Multilabel image annotation is one of the most important open problems in computer vision field. Unlike existing works that usually use conventional visual features to annotate images, based on deep learning have shown potential achieve outstanding performance. A multimodal framework proposed, which aims optimally integrate multiple neural networks pretrained with convolutional networks. In particular, proposed explores a unified two‐stage scheme consists (i) fune‐tune parameters network...

For hyperspectral imagery (HSI) classification, most of the studies focus on how to improve classification accuracy, while influence sampling strategy for performance attracts little attention. now, random (RS) is adopted strategy. That is, a image, certain number labeled samples are randomly selected as training set, and remaining taken test set. However, RS will produce over optimistic results when used evaluation because overlap between set Though spectral-spatial methods benefit from...

As one of the core content intelligent monitoring, target detection and tracking is basis for video analysis understanding. Tracking-Learning-Detection considered as a highly efficient algorithm single target. Although this can re-track when occluded by other targets, there still exists many shortcomings. This paper deals with issue fusing kalman filter tracking-learning-detection algorithm. Specifically, an improved Kalman first utilized to enhance reliability algorithm; then, area...

In recent years, images have played a more and important role in our daily life social communication. To some extent, the textual information contained pictures is an factor understanding content of scenes themselves. The accurate text detection natural is, semantic will be. Thus, scene has also become hot spot domain computer vision. this paper, we presented modified network which based on further research improvement Connectionist Text Proposal Network (CTPN) proposed by previous...

Retracted.

Edge detection is a basic problem in computer vision and image processing. The main purpose of edge to identify points with obvious brightness changes digital images. At present, there are many good methods, but most them do not consider the correctness crispness edges at same time. In order address this problem, paper proposes method based on deep convolution neural network. mainly dense network that then combined single structure backward refinement path module. former can detect retain...

To deal with the issue of multi-target tracking, this paper proposes a hierarchical correlation tracking trajectory generation method. On basis target detection and initial trajectory, AdBoost algorithm combined online discriminant analysis apparent model is first utilized to achieve trajectories; then, Hungarian here optimize fragmented discontinuous trajectories stable accurate fragments; finally, intelligent extrapolation based on energy minimization final smoother longer trajectories....

This paper proposes a massive Remote sensing tile data organization and management strategy based on file tree. The storage rule of tree-structured directory is: root directory\tile pyramid hierarchy\subdirectory\tile data. Here, subdirectory is dynamically created by the alternating growth. can maintenance dynamic balance between numbers under hierarchy. However, there are some problems need to be elaborated.

In the traditional routing model of Wireless Sensor Networks (WSN), intermediate nodes can store and forward messages sent by other nodes. Existing research shows that using network coding in WSN improve throughput communication performance, reduce node energy consumption. Based on above points, this paper mainly improves stage spread interest target area GEAR protocol, introduces to process, proposes a new coding-based wireless sensor protocol NC-GEAR (Network Coding-GEAR). According...

Due to the complexity of human motion, target detection results by using traditional ViBe algorithm are not so satisfactory. Therefore, this paper proposes a method deal with issue fusing an improved Canny operator Vibe algorithm. Specifically, is utilized achieve initial foreground region moving object; then, applied extract edge information finally, extracted and fused obtain more accurate region. The experimental performed on KTH behavior database demonstrate effectiveness proposed scheme.

Due to the complexity of human motion, target detection results by using traditional ViBe algorithm are not so satisfactory.Therefore, this paper proposes a method deal with issue fusing an improved Canny operator Vibe algorithm.Specifically, is utilized achieve initial foreground region moving object; then, applied extract edge information finally, extracted and fused obtain more accurate region.The experimental performed on KTH behavior database demonstrate effectiveness proposed scheme.

Multilabel image annotation is one of the most important open problems in computer vision field.Unlike existing works that usually use conventional visual features to annotate images, based on deep learning have shown potential achieve outstanding performance.In this work, we propose a multimodal framework, which aims optimally integrate multiple neural networks pretrained with convolutional networks.In particular, proposed framework explores unified two-stage scheme consists (i) fune-tune...

Image segmentation is a fundamental process in many image, video, and computer vision applications.It very essential critical to image processing pattern recognition, determines the quality of final result analysis recognition.This paper presents semi-supervised strategy deal with issue segmentation.Each first segmented coarsely, represented as graph model.Then, algorithm utilized estimate relevance between labeled nodes unlabeled construct matrix.Finally, normalized cut criterion segment...

Based on spatial object and relation model of database management system, this paper propose an "integrated data-attribute data" standardized data structure multisource heterogeneous satellite data. The is described by table, provides a unified access processing interface Using structure, we can provide protection for application system "Single-requirement multi-satellite".

The recent pressure pipeline reform and China's opening to the outside world have become catalysts for further economic social development. oil fields in northeast pure water projects southwest benefited residents, however, due working problems transportation, which are caused by corrosion of materials various arising from operation pipeline, resulting long-term safety problems. Therefore, importance checking cannot be missed. In order ensure natural gas pipelines under pressure, we need...