We proposed compressive data gathering (CDG) that leverages sampling (CS) principle to efficiently reduce communication cost and prolong network lifetime for large scale monitoring sensor networks. The capacity has been proven increase proportionally the sparsity of readings. In this paper, we further address two key problems in CDG framework. First, investigate how generate RIP (restricted isometry property) preserving measurements readings by taking multi-hop into account. Excitingly,...

Because of the various appearance (different writers, writing styles, noise, etc.), handwritten character recognition is one most challenging task in pattern recognition. Through decades research, traditional method has reached its limit while emergence deep learning provides a new way to break this limit. In paper, CNN-based framework proposed. framework, proper sample generation, training scheme and CNN network structure are employed according properties characters. experiments, proposed...

There is an emerging trend to leverage noisy image datasets in many visual recognition tasks. However, the label noise among severely degenerates performance of deep learning approaches. Recently, one mainstream introduce latent handle noise, which has shown promising improvement network designs. Nevertheless, mismatch between labels and still affects predictions such methods. To address this issue, we propose a probabilistic model, explicitly introduces extra variable represent...

Deep learning methods have recently achieved impressive performance in the area of visual recognition and speech recognition. In this paper, we propose a hand- writing method based on relaxation convolutional neural network (R-CNN) alternately trained (ATR-CNN). Previous regularize CNN at full-connected layer or spatial-pooling layer, however, focus layer. The convolution adopted our R-CNN, unlike traditional does not require neurons within feature map to share same kernel, endowing with...

Cloud and cloud shadow detection in remote sensing imagery is important for its wide range of applications. Traditionally, the usually based on manually designed thresholds from multiband, which complicated multistage. To simplify process improve performance, we propose a multilevel feature fused segmentation network (MFFSNet), can be trained end-to-end without any hand-tuned parameters. Specifically, fully convolutional proposed features learning. Then, utilize novel pyramid pooling module...

Learning with noisy labels is imperative in the Big Data era since it reduces expensive labor on accurate annotations. Previous method, learning noise transition, has enjoyed theoretical guarantees when applied to scenario class-conditional noise. However, this approach critically depends an pre-estimated which usually impractical. Subsequent improvement adapts preestimation form of a Softmax layer along training progress. parameters are highly tweaked for fragile performance and easily get...

A digital terrestrial television broadcasting (DTTB) standard named "Frame structure, channel coding and modulation for system" was published in China August 2006. This is the first paper of a series that provide complete depth description to including laboratory field measurement results, detailed analysis on technologies achieving stable fixed reception fast mobile reception, as well methodologies spectrum allocation, principles performing single frequency network operation. Among hundreds...

It is well known that the handwritten Chinese text recognition a difficult problem since there are large number of classes. In order to solve this problem, we proposed whole new framework for unconstrained recognition. The core module heterogeneous CNN trained by deep knowledge. experimental results showed our method could achieve much better performance than state-of-the-art methods (96.28% vs. 91.39% CR on CASIA test set). Moreover, general, it can also be applied other time sequence...

Scene text information extraction plays an important role in many computer vision applications. Most features existing algorithms are only applicable to one stage (text detection or recognition), which significantly weakens the consistency end-to-end system, especially for complex Chinese texts. To tackle this challenging problem, we propose a novel structure feature extractor based on component detector (TSCD) layer and residual network Inspired by three-layer cognition model of human,...

Radar emitter classification (REC) is an essential part of electronic warfare (EW) systems. In REC tasks, after deinterleaving, the intercepted radar signals are classified into specific types. With new types arising and electromagnetism environment getting complicated, has become a big data problem. Meanwhile, there exist inconsistent features among samples. These two problems can affect performance classification. this work, first, authors designed novel encoding method to deal with...

Text detection in a natural environment plays an important role many computer vision applications. While existing text methods are focused on English characters, there strong application demands other languages, such as Chinese. In this paper, we present novel algorithm for Chinese characters based specific designed convolutional neural network (CNN). The CNN contains structure component detector layer, spatial pyramid and multi-input-layer deep belief (DBN). is pre-trained via sparse...

In the current natural environment, due to complexity of background and high similarity color between immature green tomatoes plant, occlusion key organs (flower fruit) by leaves stems will lead low recognition rates poor generalizations detection model. Therefore, an improved tomato organ method based on convolutional neural network (CNN) has been proposed in this paper. Based original Faster R-CNN algorithm, Resnet-50 with residual blocks was used replace traditional vgg16 feature...

Glioblastoma (GBM) tumor is the most common primary brain malignant tumor. The precise identification of GBM very important for diagnosis and treatment. Hyperspectral imaging a fast, non-contact, accurate safety modern medical detection technology, which expected to be new tool intraoperative diagnosis. In order make full use spectral spatial information hyperspectral images (HSIs) achieve identification, method based on fusion multiple deep models (FMDM) proposed in-vivo human HSI...

Intracranial aneurysms (IAs) can lead to subarachnoid hemorrhage, a life-threatening event associated with high morbidity and mortality. Identifying individuals at elevated risk is crucial for guiding timely interventions improving patient outcomes. In this retrospective cohort study, 850 patients who received interventional or surgical treatment IAs between January 2018 2024 were included. Demographic data (e.g., age, sex), lifestyle factors, comorbidities recorded. Hematologic,...

As an important protein encoded by hepatitis B virus (HBV), HBV X (HBx) plays role in the development of hepatocellular carcinoma (HCC). It has been shown that seven absentia homologue 1 (SIAH1) could regulates degradation HBx through ubiquitin-proteasome pathway. However, as a member SIAH family, regulatory effects SIAH2 on remain unclear. In this study, we first confirmed reduce levels depending its E3 ligase activity. Moreover, interacted with and induced K48-linked polyubiquitination...

Currently, most electrophotographic printers use halftoning technique to print continuous tone images, so scanned images obtained from such hard copies are usually corrupted by screen like artifacts. In this paper, a new model of halftone image is proposed consider both printing distortions and patterns. Based on model, an adaptive filtering based descreening method recover high quality contone the images. Image redundancy denoising algorithm first adopted reduce noise attenuate distortions....

Non-contact imaging devices such as digital cameras and overhead scanners can convert hardcopy books to images without cutting them individual pages. However, the captured have distinct distortions. A book dewarping system is proposed remove perspective geometric distortions automatically from single images. boundary model extracted, a 3D surface reconstructed. And then horizontal vertical metrics of each column are restored it. Experimental results show good speed performance. Since no...