Search

Image classification tasks often compress the neural network model to reduce the number of parameters, which leads to a decrease in classification accuracy. herefore, we propose a novel ResNet50-based attention mechanism for image classification. ResNet50 network is used to extract image features and input the features into the graph neural network as node features. Then, packet convolution and depth-separable convolution are used to compress the residual network. The attention mechanism is introduced into the network backbone to make it focus on the important part of the neighborhood and help the branch network to extract key information. The accuracy of 5-way 1-shot task classification on three publicly available datasets reaches 86.32%, 92.21% and 92.19%, respectively. The proposed method has achieved remarkable results in image classification tasks.

Abstract Background Named entity recognition (NER) of electronic medical records is an important task in clinical medical research. Although deep learning combined with pretraining models performs well in recognizing entities in clinical texts, because Chinese electronic medical records have a special text structure and vocabulary distribution, general pretraining models cannot effectively incorporate entities and medical domain knowledge into representation learning; separate deep network models lack the ability to fully extract rich features in complex texts, which negatively affects the named entity recognition of electronic medical records. Methods To better represent electronic medical record text, we extract the text’s local features and multilevel sequence interaction information to improve the effectiveness of electronic medical record named entity recognition. This paper proposes a hybrid neural network model based on medical MC-BERT, namely, the MC-BERT + BiLSTM + CNN + MHA + CRF model. First, MC-BERT is used as the word embedding model of the text to obtain the word vector, and then BiLSTM and CNN obtain the feature information of the forward and backward directions of the word vector and the local context to obtain the corresponding feature vector. After merging the two feature vectors, they are sent to multihead self-attention (MHA) to obtain multilevel semantic features, and finally, CRF is used to decode the features and predict the label sequence. Results The experiments show that the F1 values of our proposed hybrid neural network model based on MC-BERT reach 94.22%, 86.47%, and 92.28% on the CCKS-2017, CCKS-2019 and cEHRNER datasets, respectively. Compared with the general-domain BERT-based BiLSTM + CRF, our F1 values increased by 0.89%, 1.65% and 2.63%. Finally, we analyzed the effect of an unbalanced number of entities in the electronic medical records on the results of the NER experiment.

Abstract Background Drug repositioning has caught the attention of scholars at home and abroad due to its effective reduction of the development cost and time of new drugs. However, existing drug repositioning methods that are based on computational analysis are limited by sparse data and classic fusion methods; thus, we use autoencoders and adaptive fusion methods to calculate drug repositioning. Results In this study, a drug repositioning algorithm based on a deep autoencoder and adaptive fusion was proposed to mitigate the problems of decreased precision and low-efficiency multisource data fusion caused by data sparseness. Specifically, a drug is repositioned by fusing drug-disease associations, drug target proteins, drug chemical structures and drug side effects. First, drug feature data integrated by drug target proteins and chemical structures were processed with dimension reduction via a deep autoencoder to characterize feature representations more densely and abstractly. Then, disease similarity was computed using drug-disease association data, while drug similarity was calculated with drug feature and drug-side effect data. Predictions of drug-disease associations were also calculated using a top-k neighbor method that is commonly used in predictive drug repositioning studies. Finally, a predicted matrix for drug-disease associations was acquired after fusing a wide variety of data via adaptive fusion. Based on experimental results, the proposed algorithm achieves a higher precision and recall rate than the DRCFFS, SLAMS and BADR algorithms with the same dataset. Conclusion The proposed algorithm contributes to investigating the novel uses of drugs, as shown in a case study of Alzheimer's disease. Therefore, the proposed algorithm can provide an auxiliary effect for clinical trials of drug repositioning.

Image classification indicates that it classifies the images into a certain category according to the information in the image. Therefore, extracting image feature information is an important research content in image classification. Traditional image classification mainly uses machine learning methods to extract features. With the continuous development of deep learning, various deep learning algorithms are gradually applied to image classification. However, traditional deep learning-based image classification methods have low classification efficiency and long convergence time. The training networks are prone to over-fitting. In this paper, we present a novel CapsNet neural network based on the MobileNetV2 structure for robot image classification. Aiming at the problem that the lightweight network will sacrifice classification accuracy, the MobileNetV2 is taken as the base network architecture. CapsNet is improved by optimizing the dynamic routing algorithm to generate the feature graph. The attention module is introduced to increase the weight of the saliency feature graph learned by the convolutional layer to improve its classification accuracy. The parallel input of spatial information and channel information reduces the computation and complexity of network. Finally, the experiments are carried out in CIFAR-100 dataset. The results show that the proposed model is superior to other robot image classification models in terms of classification accuracy and robustness.

Abstract Although recently a series of deep‐learning based methods have promoted the performance of face super resolution (FSR), most of these methods cannot recover essential face attributes accurately, especially to super‐resolve a very low‐resolution (LR) face image (16×16 pixels) to its (8×) high‐resolution (HR) version. To address this issue, a novel alternating optimisation algorithm to estimate facial attributes and restore facial images in a single network is presented . Specifically, two convolutional neural modules (denoted as Restorer and Corrector) are constructed and these two modules are alternated repeatedly to form an end‐to‐end trainable network. The Restorer module reconstructs face images based on estimated facial attributes, while the Corrector module corrects the estimated attributes with the help of restored face image. Since Corrector can utilise information from previous estimated attributes and the FSR image, the estimated attributes are iteratively corrected and gradually approach the ground truth. Moreover, a new attribute transformation scheme is designed to introduce attribution information into Restorer, in which the facial attribute vectors as the control conditions can guide the face image restoration explicitly. By conducting extensive experiments on the well‐known CelebA dataset, it is demonstrated that the proposed method can provide superior FSR performance in both quantitative and qualitative measurements.

Blind face restoration (BFR) from severely degraded face images is important in face image processing and has attracted increasing attention due to its wide applications. However, due to the complex unknown degradations in real-world scenarios, existing priors-based methods tend to restore faces with unstable quality. In this article, we propose a multi-prior collaboration network (MPCNet) to seamlessly integrate the advantages of generative priors and face-specific geometry priors. Specifically, we pretrain a high-quality (HQ) face synthesis generative adversarial network (GAN) and a parsing mask prediction network, and then embed them into a U-shaped deep neural network (DNN) as decoder priors to guide face restoration, during which the generative priors can provide adequate details and the parsing map priors provide geometry and semantic information. Furthermore, we design adaptive priors feature fusion (APFF) blocks to incorporate the prior features from pretrained face synthesis GAN and face parsing network in an adaptive and progressive manner, making our MPCNet exhibits good generalization in a real-world application. Experiments demonstrate the superiority of our MPCNet in comparison to state-of-the-arts and also show its potential in handling real-world low-quality (LQ) images from several practical applications.

Abstract Study design A single-institution, retrospective cohort study. Objective To compare the accuracy and short-term clinical outcomes of pedicle screw placement between robot-assisted (RA) and freehand (FH) technique in the treatment of adult degenerative scoliosis (ADS). Methods From February 2018 to October 2019, 97 adult patients with degenerative scoliosis admitted to our department were retrospectively reviewed. Thirty-one patients received robot-assisted pedicle screw placement (RA group), and 66 patients underwent freehand pedicle screw placement (FH group). Patient demographics and short-term clinical outcomes were recorded and compared between two groups. Gertzbein-Robbins grading system was adopted to evaluate the accuracy of pedicle screw placement by means of postoperative CT scan. Short-term clinical outcomes consist of operative time, intraoperative blood loss, length of hospital stay (LOS), radiological parameters, Scoliosis Research Society-22 (SRS-22) scores before the operation, 6 months after operation, adverse events, and revisions. Results The accuracy of screw placement was higher than that of the FH group (clinically acceptable 98.7% vs. 92.2%; P< 0.001). Intraoperative blood loss of the RA group was less than those in the FH group (499 vs. 573 ml; P < 0.001). Operative time (283.1 vs. 291.9 min; P = 0.31) and length of stay (12.8 vs. 13.7 days; P = 0.36) were compared between RA and FH groups. In terms of radiological parameters, both of groups were improved postoperatively. The SRS-22 scores at 6 months after operation from both groups were better than those before operation. For surgery-related complication, one case had pressure sores in the RA group while two cases developed dural tears in the FH group. No revision was required in both groups. Conclusion Combined with other surgical correction modalities, robot-assisted pedicle screw fixation is an effective and safe method of treating degenerative scoliosis. Due to its satisfactory surgical outcomes such as higher accuracy and less trauma, it provides a good alternative for clinical practice. Level of evidence 3.

Machine learning based bottom-up saliency detection (MLBU) methods are very popular recently. These MLBU methods firstly use prior knowledge to select some regions from the given image as training samples and label them. Based on training set, a saliency classifier is learned to classify salient object and background by applying machine learning algorithms in the given image. Nevertheless, training labels obtained by prior knowledge are not always accurate in some complex scenes, inaccurate training set is hard to make subsequent learning process succeed. To solve this problem, we propose an inaccurate supervised learning (ISL) based saliency detection framework, which assumes that training labels obtained by prior knowledge might be inaccurate and constructs three checking rules to remove mislabeled samples for more accurate training set construction. The refined training set is used to learn a saliency classifier which can better predict each image region. To obtain more accurate saliency inference, the proposed ISL process is introduced into a novel iterative feedback (IF) framework to generate better saliency result. Finally, we use smoothness operator to further smooth saliency result for performance improvement. Experimental results on three benchmark datasets demonstrate adequately the superiority of the proposed method.

Recently, single image super-resolution (SISR) has been widely applied in the field of remote sensing image processing and obtained remarkable performance. However, existing CNN-based remote sensing image super-resolution methods are unable to exploit shallow visual characteristics at global receptive fields, which results in the limited perceptual capability of these models. Furthermore, the low-resolution inputs and features contain abundant low-frequency information, which are weighed in channels and space equally, hence limiting the representational ability of CNNs. To solve these problems, we propose a non-locally up-down convolutional attention network (NLASR) for remote sensing image super-resolution. First, a non-local features enhancement module (NLEB) is constructed to obtain the spatial context information of high-dimensional feature maps, which allows our network to utilize global information to enhance low-level similar structured texture information with effect, overcoming the defects of deficiency perceptual ability of shallow convolutional layers. Second, an enhanced up-sampling channel-wise attention (EUCA) module and enhanced down-sampling spatial-wise attention (EDSA) module are proposed to weight the features at multiple scales. By integrating the channel-wise and multi-scale spatial information, the attention modules are able to compute the response values from the multi-scale regions of each neuron and then establish the accurate mapping from low to high resolution space. Extensive experiments on NWPU-RESISC45 and UCMerced-LandUse datasets show that the proposed method can provide state-of-the-art or even better performance in both quantitative and qualitative measurements.

Low-dose CT (LDCT) images have been widely applied in the medical imaging field due to the potential risk of exposing patients to X-ray radiations. Given the fact that reducing the radiation dose may result in increased noise and artifacts, methods that can eliminate the noise and artifacts in the LDCT image have drawn increasing attentions and produced impressive results over the past decades. However, recent proposed methods mostly suffer from noise remaining, over-smoothing structures or false lesions derived from noise. In this paper, we propose a generative adversarial network (GAN) with novel architecture and loss function for restoring the LDCT image. Firstly, the inception-residual block and residual mapping are incorporated in the U-Net structure. The modified U-Net is applied as the generator of the GAN network so that the noise feature can be eliminated during the forward propagation. Secondly, a novel multi-level joint discriminator is designed by concatenating multiple convolutional neural networks (CNNs) where the output of each deconvolutional layer in the generator is compared with the corresponding down-sampled ground truth image. The adversarial training can be sensitive to noise and artifacts in different scales with this discriminator. Thirdly, we novely define a loss function consisting of the least square adversarial loss, VGG based perceptual loss, MSE based pixel loss and the noise loss, so that the differences in pixel, visual perception and noise distribution are comprehensively considered to optimize the network. Experimental results on both simulated and official simulated clinical images have demonstrated that the proposed method can provide superior performance to the state-of-the-art methods in noise removal, structure preservation and false lesions elimination.

With the development of computer vision, high quality images with rich information have great research potential in both daily life and scientific research. However, due to different lighting conditions, surrounding noise and other reasons, the image quality is different, which seriously affects people's discrimination of the information in the image, thus causing unnecessary conflicts and results. Especially in the dark, the images captured by the camera are difficult to identify, and the smart system relies heavily on high-quality input images. The image collected in low-light environment has the characteristic with high noise and color distortion, which makes it difficult to utilize the image and can not fully explore the rich value information of the image. In order to improve the quality of low-light image, this paper proposes a Heterogenous low-light image enhancement method based on DenseNet generative adversarial network. Firstly, the generative network of generative adversarial network is realized by using DenseNet framework. Secondly, the feature map from low light image to normal light image is learned by using the generative adversarial network. Thirdly, the enhancement of low-light image is realized. The experimental results show that, in terms of PSNR, SSIM, NIQE, UQI, NQE and PIQE indexes, compared with the state-of-the-art enhancement algorithms, the values are ideal, the proposed method can improve the image brightness more effectively and reduce the noise of enhanced image.

Objective: This study was designed to investigate the relationship between the laminar slope angle (LSA) and the lumbar disc degenerative grade, the cross-section area (CSA) of multifidus muscle, the muscle-fat index, and the thickness of the ligamentum flavum. Methods: Retrospective analysis of 122 patients who were scheduled to undergo a lumbar operation for diagnoses associated with degenerative lumbar disease between January and December 2017. The L4–L5 disc grade was evaluated from preoperative sagittal T2-weighed magnetic resonance imaging of the lumber region; the CSA of the multifidus and muscle-fat index were measured at the L4 level, while the thickness of the ligamentum flavum was measured at the L4–L5 facet level from axis T2-weighed magnetic resonance imaging. The slope of the laminar was evaluated from preoperative three-dimensional computer tomography at the tip level of the facet joints and selected by the axis plane. Independent-sample T-tests were used to assess the association between age and measurement indices. Results: Our results showed that age was positively connected with the LSA of L4 and L5 in different patients, although there was no significant difference between age and the difference of the two segment LSA. Partial correlation analysis, excluding the interference of age, revealed a strong negative relationship between the LSA of L4 and the thickness of the ligamentum flavum, irrespective of whether we considered the left or right. However, there was no correlation with lumbar disc degenerative grade, the CSA of the multifidus, and the muscle-fat index. Conclusion: The thickness of the ligamentum flavum showed changes with anatomical differences in the LSA, but not the lumbar disc degenerative grade, the CSA of the multifidus, and the muscle-fat index. A small change in LSA may cause large mechanical stress; this may be one of the causative factors responsible for lumbar spinal stenosis.

Glaucoma is an eye disease which is one of the most common causes of blindness. Accurate optic disc (OD) and optic cup (OC) segmentation play a critical role for detecting glaucoma. Considering that the existing approaches can't effectively integrate the multi-view information deriving from shape and appearance to sufficiently describe OD and OC for segmentation, Locally Statistical Active Contour Model with the Information of Appearance and Shape (LSACM-AS) and Modified Locally Statistical Active Contour Model with the Information of Appearance and Shape (MLSACM-AS) are proposed in this paper. The main contributions are as below: (1) we introduce the Locally Statistical Active Contour Model (LSACM) to address the commonly occurred intensity inhomogeneity phenomenon caused by imperfection of image devices or illumination variations. (2) In order to overcome the common effects caused by pathological changes (i.e., peripapillary atrophy (PPA)) and vessel occlusion in OD and OC segmentation, we integrate the local image probability information around the point of interest from a multi-dimensional feature space into our model to preserve the integrity of the OD and OC structures. (3) Since the segmentation objects have the similar ellipse shape structure, we incorporate the shape priori constraint information into our model to further improve the robustness of the variations found in and around objects regions. To evaluate the effectiveness of the proposed models, an available publicly DRISHTI-GS database is employed in this paper. Seen from the abundant experiments, the proposed models outperform the state-of-the-art approaches in terms of the obtained qualitative and quantitative results.

A large class of Person Re-identification (ReID) approaches identify pedestrians with the TriHard loss. Though the TriHard loss is a robust ReID method, pose variance and viewpoint in pedestrians constrain the performance. To address this problem, we introduce a spatial transformer network (STN) to align pedestrians. Then, we illustrate the generality of the STN module in pose variance problem through the evaluations on feature representation network (FRN) like VGG, ResNet and DenseNet architectures respectively. Furthermore, based on the evaluation results, we propose a robust and high-performance ReID model which consists of the STN module, DenseNet backbone and TriHard loss. And finally, we prove that our ReID model is whole differentiable by formula derivation, therefore achieving an end-to-end high-performance ReID system. The experiments show that our ReID system outperforms the state-of-art methods on Market-1501, DukeMTMC-reID and CUHK03 datasets.

Background: Intervertebral disc degeneration impairs the quality of patients lives. Even though there has been development of many therapeutic strategies, most of them remain unsatisfactory due to the limited understanding of the mechanisms that underlie the intervertebral disc degeneration.Questions/purposes: This study is meant to identify the key modules and hub genes related to the annulus fibrosus in intervertebral disc degeneration (IDD) through: (1) constructing a weighted gene co-expression network; (2) identifying key modules and hub genes; (3) verifying the relationships of key modules and hub genes with IDD; and (4) confirming the expression pattern of hub genes in clinical samples.Methods: The Gene Expression Omnibus provided 24 sets of annulus fibrosus microarray data. Differentially expressed genes between the annulus fibrosus of degenerative and non-degenerative intervertebral disc samples have gone through the Gene Ontology (GO) and pathway analysis. The construction of a gene network and classification of genes into different modules were conducted through performing Weighted Gene Co-expression Network Analysis. The identification of modules and hub genes that were most related to intervertebral disc degeneration was proceeded. In order to verify the relationships of the module and hub genes with intervertebral disc degeneration, Ingenuity Pathway Analysis was operated. Clinical samples were adopted to help verify the hub gene expression profile.Results: One thousand one hundred ninety differentially expressed genes were identified. Terms and pathways associated with intervertebral disc degeneration were presented by GO and pathway analysis. The construction of a Weighted Gene Coexpression Network was completed and clustering differentially expressed genes into four modules was also achieved. The module with the lowest P-value and the highest absolute correlation coefficient was selected and its relationship with intervertebral disc degeneration was confirmed by Ingenuity Pathway Analysis. The identification of hub genes and the confirmation of their expression profile were also realized.Conclusions: This study generated a comprehensive overview of the gene networks underlying annulus fibrosus in intervertebral disc degeneration.Clinical Relevance: Modules and hub genes identified in this study are highly associated with intervertebral disc degeneration, and may serve as potential therapeutic targets for intervertebral disc degeneration.

Although the development time of cross-border e-commerce in China is very short, the scale of its transactions and the speed of development are amazing, and as a supporting foundation for promoting economic and trade globalization, cross-border e-commerce has an extremely important strategy for this guiding role. This not only brings new opportunities to cross-border e-commerce companies but also excavates a huge potential market for the logistics industry. Cross-border e-commerce not only breaks through the trade barriers between countries; it makes trade move towards borderlessness and at the same time triggers major changes in international trade. This paper introduces partial differential equations into the video surveillance image enhancement system of cross-border e-commerce logistics. Aiming at the shortcomings of the contrast enhancement method based on gradient field equalization, this paper proposes a partial differential enhancement method based on histogram equalization. By proposing a gradient transformation function, the edges and textures with relatively small gradient values are enhanced to make the original weaker texture details clearer. In order to better adjust the brightness and contrast of the image, combined with histogram equalization, we propose an inverse equalization transformation. When the histogram equalization and the inverse equalization transform are combined reasonably, the brightness and contrast of the image can be adjusted very well. In this paper, the finite difference method is used for discretization when solving partial differential equations, and Euler’s equation is obtained by applying the principle of least squares. By introducing the heat equation, the direct solution of Euler’s equation is converted into an iterative form, which greatly reduces the amount of calculation. This article uses statistical methods to obtain the empirical formula of the fractional differential order. This empirical formula makes the calculation of the order of the fractional derivative easy and can be extended to other fractional image enhancement models and overcomes the shortcomings of the traditional fractional derivative order obtained through experience or a large number of experiments. Experiments show that the proposed algorithm not only enhances detailed texture information but also improves image clarity, overall brightness, and contrast without color distortion. The objective evaluation indicators also show the superiority of the algorithm.

Abstract Feature selection, which aims to select an optimal feature subset to avoid the “curse of dimensionality,” is an important research topic in many real-world applications. To select informative features from a high-dimensional dataset, we propose a novel unsupervised feature selection algorithm called Double Regularized Matrix Factorization Feature Selection (DRMFFS) in this paper. DRMFFS is based on the feature selection framework of matrix factorization, but extends this framework by introducing double regularizations (i.e., graph regularization and inner product regularization). There are three major contributions to our approach. First, for the sake of preserving the useful underlying geometric structure information of the feature space of the data, we introduce the graph regularization to guide the learning of the feature selection matrix, making it more effective. Second, in order to take into account the correlations among features, an inner product regularization term is imposed on the objective function of matrix factorization. Therefore, the selected features by DRMFFS cannot only represent the original high-dimensional data well but also contain low redundancy. Third, we design an efficient iteratively update algorithm to solve our approach and also prove its convergence. Experiments on six benchmark databases demonstrate that the proposed approach outperforms the state-of-the-art approaches in terms of both the classification and clustering performance.

Structured light (SL) has been extensively researched and developed due to reconstruction the fast and highly accurate depth sensing. In order to achieve this goal, many approaches have been taken. In this article, we have taken some innovative work in the field for building a new type of fast, high-precision depth perception based on structured light. In addition, the basic ideas we propose are as follows: our model for slow motion dynamic scenes, the sparse depth maps are obtained based on the pair of unstructured rigid pattern with different color, and the high-quality depth images are optimized by the ridge line extraction. Finally, the calculated two depth maps based on the different color pattern merged, which to acquire the optimization depth. We further propose a model for the high-speed motion dynamic scene which apply a pair of projectors structure light. We show that, we actively use motion blur to estimate depth which is different from the previous approaches. The patterns are illuminated from two projectors and the motion blur of each line is accurately measured. The scene depth information is estimated by analyzing the length of the blur. As a result, we describe two experiments that we conducted on two different experimental equipment (ball and planner board). The first experiment was conducted with the planer board, measuring the slow moving object. In addition, the second experiment scene was conducted with the ball, which obtained the depth of the ultra-fast object. Our experimental results demonstrate that our approach achieves effective real-time measurement, can successfully obtain accurate depth information of moving objects. Keywords: Structured light, Dynamic scene, Ridge detection, Light flow, Unstructured grid

Pulmonary nodule detection in chest computed tomography (CT) is of great significance for the early diagnosis of lung cancer. Therefore, it has attracted more and more researchers to propose various computer-assisted pulmonary nodule detection methods. However, these methods still could not provide convincing results because the nodules are easily confused with calcifications, vessels, or other benign lumps. In this paper, we propose a novel deep convolutional neural network (DCNN) framework for detecting pulmonary nodules in the chest CT image. The framework consists of three cascaded networks: First, a U-net network integrating inception structure and dense skip connection is proposed to segment the region of lung parenchyma from the chest CT image. The inception structure is used to replace the first convolution layer for better feature extraction with respect to multiple receptive fields, while the dense skip connection could reuse these features and transfer them through the network. Secondly, a modified U-net network where all the convolution layers are replaced by dilated convolution is proposed to detect the “suspicious nodules” in the image. The dilated convolution can increase the receptive fields to improve the ability of the network in learning global information of the image. Thirdly, a modified U-net adapting multi-scale pooling and multi-resolution convolution connection is proposed to find the true pulmonary nodule in the image with multiple candidate regions. During the detection, the result of the former step is used as the input of the latter step to follow the “coarse-to-fine” detection process. Moreover, the focal loss, perceptual loss and dice loss were used together to replace the cross-entropy loss to solve the problem of imbalance distribution of positive and negative samples. We apply our method on two public datasets to evaluate its ability in pulmonary nodule detection. Experimental results illustrate that the proposed method outperform the state-of-the-art methods with respect to accuracy, sensitivity and specificity.

Accurate optic disc (OD) and optic cup (OC) segmentation play a critical role in automatic glaucoma diagnosis. In this paper, we present an automatic segmentation technique regarding the OD and the OC for glaucoma assessment. First, the robust adaptive approach for initializing the level set is designed to increase the performance of contour evolution. Afterwards, in order to handle the complex OD appearance affected by intensity inhomogeneity, pathological changes, and vessel occlusion, a novel model that integrates ample information of OD with the effective local intensity clustering (LIC) model together is presented. For the OC segmentation, to overcome the segmentation challenge as the OC’s complex anatomy location, a novel preprocessing method based on structure prior information between the OD and the OC is designed to guide contour evolution in an effective region. Then, a novel implicit region based on modified data term using a richer form of local image clustering information at each point of interest gathered over a multiple-channel feature image space is presented, to enhance the robustness of the variations found in and around the OC region. The presented models symmetrically integrate the information at each point in a single-channel image from a multiple-channel feature image space. Thus, these models correlate with the concept of symmetry. The proposed models are tested on the publicly available DRISHTI-GS database and the experimental results demonstrate that the models outperform state-of-the-art methods.

Visual object tracking methods based on wireless multimedia sensor network is one of the research hotspots while the present linear method for processing feature vectors often lead to the tracking drift when tracking object with significant nonplanar pose variations through wireless sensor networks. In this article, we propose a novel nonlinear algorithm for tracking significant deformable objects. The proposed tracking scheme has two filters. On one hand, considering that Grassmann manifold is one of entropy manifold in Lie group manifold, which can describe and process the data of appearance feature more accurately, one filter is designed on it, to estimate the object appearance, by making full use of the transformation relationship between the point on manifold and its corresponding point on tangent space. On the other hand, considering that the process of objects imaging is essentially projection transformation process, the other filter is designed on projection transformation (SL(3)) group, describing the geometric deformation of the objects. The two filters execute alternatively to mitigate tracking drift. Extensive experiments prove that the proposed method can realize stable and accurate tracking for targets with significant geometric deformation, even obscured and illumination changes.

Anomaly detection in crowded scenes is an important and challenging part of the intelligent video surveillance system. As the deep neural networks make success in feature representation, the features extracted by a deep neural network represent the appearance and motion patterns in different scenes more specifically, comparing with the hand-crafted features typically used in the traditional anomaly detection approaches. In this paper, we propose a new baseline framework of anomaly detection for complex surveillance scenes based on a variational auto-encoder with convolution kernels to learn feature representations. Firstly, the raw frames series are provided as input to our variational auto-encoder without any preprocessing to learn the appearance and motion features of the receptive fields. Then, multiple Gaussian models are used to predict the anomaly scores of the corresponding receptive fields. Our proposed two-stage anomaly detection system is evaluated on the video surveillance dataset for a large scene, UCSD pedestrian datasets, and yields competitive performance compared with state-of-the-art methods.

Computed tomography (CT) imaging technology has been widely used to assist medical diagnosis in recent years. However, noise during the process of imaging, and data compression during the process of storage and transmission always interrupt the image quality, resulting in unreliable performance of the post-processing steps in the computer assisted diagnosis system (CADs), such as medical image segmentation, feature extraction, and medical image classification. Since the degradation of medical images typically appears as noise and low-resolution blurring, in this paper, we propose a uniform deep convolutional neural network (DCNN) framework to handle the de-noising and super-resolution of the CT image at the same time. The framework consists of two steps: Firstly, a dense-inception network integrating an inception structure and dense skip connection is proposed to estimate the noise level. The inception structure is used to extract the noise and blurring features with respect to multiple receptive fields, while the dense skip connection can reuse those extracted features and transfer them across the network. Secondly, a modified residual-dense network combined with joint loss is proposed to reconstruct the high-resolution image with low noise. The inception block is applied on each skip connection of the dense-residual network so that the structure features of the image are transferred through the network more than the noise and blurring features. Moreover, both the perceptual loss and the mean square error (MSE) loss are used to restrain the network, leading to better performance in the reconstruction of image edges and details. Our proposed network integrates the degradation estimation, noise removal, and image super-resolution in one uniform framework to enhance medical image quality. We apply our method to the Cancer Imaging Archive (TCIA) public dataset to evaluate its ability in medical image quality enhancement. The experimental results demonstrate that the proposed method outperforms the state-of-the-art methods on de-noising and super-resolution by providing higher peak signal to noise ratio (PSNR) and structure similarity index (SSIM) values.

Following the development of wireless multimedia sensor networks (WMSN), the coverage of the sensors in the network constitutes one of the key technologies that have a significant influence on the monitoring ability, quality of service, and network lifetime. The application environment of WMSN is always a complex surface, such as a hilly surface, that would likely cause monitoring shadowing problems. In this study, a new coverage-enhancing algorithm is presented to achieve an optimal coverage ratio of WMSN based on three-dimensional (3D) complex surfaces. By aiming at the complex surface, the use of a 3D sensing model, including a sensor monitoring model and a surface map calculation algorithm, is proposed to calculate the WMSN coverage information in an accurate manner. The coverage base map allowed the efficient estimation of the degree of monitoring occlusion efficiently and improved the system’s accuracy. To meet the requests of complex 3D surface monitoring tasks for multiple sensors, we propose a modified cuckoo search algorithm that considers the features of the WMSN coverage problem and combines the survival of the fittest, dynamic discovery probability, and the self-adaptation strategy of rotation. The evaluation outcomes demonstrate that the proposed algorithm can describe the 3D covering field but also improve both the coverage quality and efficiency of the WMSN on a complex surface.

Image super resolution (SR) based on example learning is a very effective approach to achieve high resolution (HR) image from image input of low resolution (LR). The most popular method, however, depends on either the external training dataset or the internal similar structure, which limits the quality of image reconstruction. In the paper, we present a novel SR algorithm by learning weighted random forest and non-local similar structures. The initial HR image patches are obtained from a weighted forest model, which is established by calculating the approximate fitting error of the leaf nodes. The K-means clustering algorithm is exploited to get a non-local similar structure inside the initial HR image patches. In addition, a low rank constraint is imposed on the HR image patches in each cluster. We further apply the similar structure model to establish an effective regularization prior under a reconstruction-based SR framework. Comparing with current typical SR algorithms, the results of comprehensive experiments implemented on three publicly datasets show that peak signal-to-noise ratio (PSNR) has been effectively promoted by the presented SR approach, and a better visual effect has been realized.

Superpixel segmentation is a widely used preprocessing method in computer vision, but its performance is unsatisfactory for color images in cluttered indoor environments. In this work, a superpixel method named weighted coplanar feature clustering (WCFC) is proposed, which produces full coverage of superpixels in RGB-depth (RGBD) images of indoor scenes. Basically, a linear iterative clustering is adopted based on a cluster criterion that measures the color similarity, space proximity and geometric resemblance between pixels. However, to avoid the adverse impact of RGBD image flaws and to make full use of the depth information, WCFC first preprocesses the raw depth maps with an inpainting algorithm called a Cross-Bilateral Filter. Second, a coplanar feature is extracted from the refined RGBD image to represent the geometric similarities between pixels. Third, combined with the colors and positions of the pixels, the coplanar feature constructs the feature vector of the clustering method; thus, the distance measure, as the cluster criterion, is computed by normalizing the feature vectors. Finally, in order to extract the features of the RGBD image dynamically, a content-adaptive weight is introduced as a coefficient of the coplanar feature, which strikes a balance between the coplanar feature and other features. Experiments performed on the New York University (NYU) Depth V2 dataset demonstrate that WCFC outperforms the available state-of-the-art methods in terms of accuracy of superpixel segmentation, while maintaining a high speed.

A fundamental issue in sensor network is the coverage problem. Since the distribution of sensor nodes is not usually uniform due to random deployment and node failures, the coverage holes are hardly avoided in sensor network. And the coverage holes are important health indicators of the sensor network. This paper firstly proposes a level set based coverage holes detection algorithm for hybrid sensor network. This algorithm could estimate the number of holes and the size of the holes. Then we propose genetic algorithms based coverage holes healing algorithm. This algorithm could leverage mobility to optimize the average coverage rate and the average movement distance of the mobile nodes. Simulation results show that the proposed method could detect the holes efficiently. The holes healing algorithm outperforms the Random and Delaunay methods.