With success of convolutional neural networks (CNNs) in computer vision, the CNN has attracted great attention hyperspectral classification. Many deep learning-based algorithms have been focused on feature extraction for classification improvement. In this letter, a novel learning framework based fully is proposed. Through convolution, deconvolution, and pooling layers, features data are enhanced. After enhancement, optimized extreme machine (ELM) utilized The proposed outperforms existing...

In this letter, we propose a novel automatic target recognition (ATR) method based on dictionary learning and joint dynamic sparse representation (DL-JDSR) for synthetic aperture radar (SAR) images. First, in the feature extraction step, extract two kinds of features, i.e., image domain amplitude scale-invariant transform (SIFT) feature, which describes intensity information SIFT gradient information. These features will be jointly utilized to combine SAR ATR. Second, introduce...

A hyperspectral image (HSI) super-resolution (SR) is a highly attractive topic in computer vision. However, most existed methods require an auxiliary high-resolution (HR) with respect to the input low-resolution (LR) HSI. This limits practicability of these HSI SR methods. Moreover, often destroy important spectral information. letter presents deep difference convolutional neural network (SDCNN) combination spatial-error-correction (SEC) model for SR. method allows full exploration and...

This article presents an intrafusion network (IFN) for hyperspectral image (HSI) super-resolution (SR). Given that the HSI is a 3-D data cube with both spatial information and spectral information, key challenge to construct SR how efficiently exploit among consecutive low-resolution (LR) bands, besides information. The proposed IFN consists of three modules, including difference module, parallel convolution which directly utilizes reconstructing high-resolution HSI. Different from most...

Download This Paper Open PDF in Browser Add to My Library Share: Permalink Using these links will ensure access this page indefinitely Copy URL DOI

In the construction site with complex environment, there are a variety of potential risk factors threatening personal safety. Since head is most critical part human body and vulnerable to fatal injuries, lots accidents caused by workers not wearing safety helmets have occurred from time time. So, staff working in has requirement wear helmet. order reduce incidence protective helmet, realtime detection application for helmet based on YOLO (You Look Only Once) v3 was proposed encapsulated into...

Thermal issue is a major concern in 3D integrated circuit (IC) design. optimization of IC often requires massive expensive PDE simulations. Neural network-based thermal prediction models can perform real-time for many unseen new designs. However, existing works either solve 2D temperature fields only or do not generalize well to designs with design configurations (e.g., heat sources and boundary conditions). In this paper, the first time, we propose DeepOHeat, physics-aware operator learning...

Over 50 million people worldwide suffer from epilepsy. Traditional diagnosis of epilepsy relies on tedious visual screening by highly trained clinicians lengthy EEG recording that contains the presence seizure (ictal) activities. Nowadays, there are many automatic systems can recognize seizure-related signals to help diagnosis. However, it is very costly and inconvenient obtain long-term data with activities, especially in areas short medical resources. We demonstrate this paper we use...

The transmission of virtual reality (VR) videos requires huge bandwidth, which brings great challenges for the system to implement real-time applications. This letter proposes a scalable full-panorama video coding method adapt insufficient in user's movement information is utilized as feedback from VR device encoder. regions that user interested are first mapped and then coded high quality, while others low quality. different-quality achieved through efficiency (SHVC), only makes limited...

Motivated by the problem of radar target recognition, we develop a label-aided factor analysis (LA-FA) model for statistical modeling high-resolution range profile (HRRP) under prerequisite that HRRP data are Gaussian distributed. The LA-FA is extension multitask learning-based (MTL-FA) model, which mainly applied to recognition with small training size. Compared MTL-FA our introduces discrete class labels via Sigmoid-Bernoulli hierarchy restrict learning parameters, offers potential enhance...

Limited by the existing imagery sensors, hyperspectral images are characterized high spectral resolution but low spatial resolution. The super-resolution (SR) technique aiming at enhancing of input image is a hot topic in computer vision. In this paper, we present (HSI) SR method based on deep information distillation network (IDN) and an intra-fusion operation. Specifically, bands firstly selected certain distance super-resolved IDN. IDN employs blocks to gradually extract abundant...

We propose to use two parameters, the scene complexity (C) and level of motion (M), capture video content characteristics that matter most quality compressed videos transmitted over IP networks, either with or without packet losses. Two simple but robust formulas are designed calculate these parameters from following variables: bits coded Intra (I-) frames (Bits <inf xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">I</inf> ), Inter (P-)...

A few emerging medical imaging methods are being developed for breast imaging. Electrical impedance tomography (EIT) is an excellent candidate safe, low cost, and noninvasive cancer monitoring. Despite early promises, the EIT faces a challenges application. It mainly due to its limited resolution especially tumors in depth. However, unlike other applications of EIT, such as brain thorax, tissues deformable. This article exploits deformation shape enhance depth detection. Exterior boundary...

Hyperspectral image (HSI) super-resolution has gained great attention in remote sensing, due to its effectiveness enhancing the spatial information of HSI while preserving high spectral discriminative ability, without modifying imagery hardware. In this paper, we proposed a novel method via gradient-guided residual dense network (G-RDN), which gradient is exploited guide process. Specifically, there are three modules super-resolving Firstly, mapping between low-resolution and desired...

<title>Abstract</title> Heart sound auscultation plays a crucial role in the early diagnosis of cardiovascular diseases. In recent years, great achievements have been made automatic heart classification; however, most methods are based on segmentation features and traditional classifiers, do not fully use existing deep networks. This paper proposes cardiac audio classification method image expression multidimensional (CACIEMDF). First, 102-dimensional feature vector is designed by combining...

Daylighting performance is one of the primary metrics used for evaluation and selection window systems. Radiance most accurate software tools lighting simulations buildings. However, it difficult to use in early stages daylighting design because complexity tool uncertainty variability parameters. Repeated modifications different parameters are required order select appropriate design. As a result, simulation runtime significantly increased. To solve conflict between accuracy length runtime,...

As wireless local area networks (WLANs) become a part of our network infrastructure, it is critical that we understand both the performance provided to end users and capacity these WLANs in terms number supported flows (calls). Since clear video traffic, as well voice data, will be carried by networks, particularly important investigate issues for packetized video. In this paper, user subject multiuser perceptual quality constraint. particular example, study transmission AVC/H.264 coded...

We revisit the classic problem of developing a spatial correlation model for natural images and videos by proposing conditional relatively nearby pixels that is dependent upon five parameters. The conditioning on local texture optimal parameters can be calculated specific image or video with mean absolute error usually smaller than 5%. use this to calculate rate distortion function when universal side information available at both encoder decoder. demonstrate information, available, save as...

To deal with the serious visual artifacts caused by consistent restoration algorithm on local motion blurred images, a novel two stage blur region detection method including coarse location and refinement for images is proposed in this paper. First, feature discriminant criteria frequency domain spatial defined to generate image; then, binarization used image obtain map; subsequently, morphological methods are process clear areas trimap obtained. Finally, of achieved combination automatic...