Search

Graph convolutional neural networks (GCNs) have been widely used in hyperspectral images (HSIs) classification for their superiority in processing non-Euclidean structure data. The performance of GCNs relies on the initial graph structure. Most GCN models only utilize spectral information to construct a graph, which is inaccurate because they fail to take the relationship between adjacent nodes into consideration. In addition, due to the over-smooth phenomenon, most GCN models are shallow and unable to extract effective features. To address these issues, a dual graph u-nets is proposed by integrating spatial graph and spectral graph for HSIs classification, denoted by DGU-HSI. To integration the spectral and spatial information, two graphs are constructed for feature extraction. The spectral graph is created by spectral similarity among all pixels where multirange spectral information is retained, and the spatial graph is constructed by exploiting the neighborhood relationship of the center pixel, which describes spatial information. Then, a dual GCN is utilized to extract spatial and spectral graph features simultaneously. To relieve the over-smooth phenomenon, the graph u-nets structure is applied on constructed spectral and spatial graph to extract effective features. Finally, the extracted spectral and spatial features are fused for classification. Experiments conducted on the public datasets demonstrate the effectiveness of the proposed method on HSIs classification.

Abundant spectral features are the precious wealth of hyperspectral images (HSI). Nevertheless, well-designed spectral feature is still a challenge that affects the performance of the classifier, especially with insufficient number of training samples. To make up the poor discriminability of spectral feature, double-branch methods are proposed by fusing parallel spectral and spatial branches. However, this structure does nothing to improve the quality of spectral feature, which is regarded as the most valuable information for HSI information. In this article, we propose a siamese spectral attention network with channel consistency (SSACC) to focus on obtaining discriminative spectral features, thus improving the generalization ability of the classifier. Two kinds of HSI cubes with different patch sizes are generated as the input of SSACC. The two cubes are divided into top and bottom branches and then be fed into the siamese network to obtain the refined spectral features. Then, self-attention is conducted to interacting with each channel for the spectral features enhancement. Meanwhile, two attention maps are obtained to display the spectral structures of each branch. A channel consistency regularization is performed on the two attention maps by enforcing the two branches to possess similar spectral patterns when identifying the same centric pixel. Extensive experiments conducted on the three HSI datasets verify the superiority of the obtained spectral feature. Furthermore, the proposed method applying convolution only on the spectral domain outperforms the state-of-the-art double-branch methods which integrate the spectral and spatial features simultaneously.

Convolutional Neural Network (CNN) is widely used in Hyperspectral Images (HSIs) classification. However, the fine-grained spatial (FGS) details are discarded during a sequence of convolution and pooling operations for most of CNN-based HSIs classification methods. To address this issue, a unified encoder-decoder framework is proposed to integrate high-level semantics and FGS details for HSIs classification, denoted by FGSCNN. The encoder, including a series of convolution and pooling layers, captures the high-level semantic information with low resolution feature maps. The decoder fuses the high-level low-resolution semantic and the fine-grained high-resolution spatial information, namely, to get the FGS features with high-level semantics. The deconvolution layers and skip connection are used in the decoder to retain the FGS details, while, convolution layers are also used to combine the FGS features with high-level semantics. Based on the encoder-decoder framework, a unified loss function is exploited to integrate the high-level semantic information and FGS details with an end-to-end manner for HSIs classification. Experiments conducted on the three public datasets, i.e. the Indian Pines, Pavia University and Salinas, demonstrate the effectiveness of the proposed method on HSIs classification.

Variational Autoencoders (VAEs) and Generative Adversarial Networks (GANs) have been widely used in hyperspectral image classification (HSIC) tasks. However, the generated HSI virtual samples by VAEs are often ambiguous, and GANs are prone to the mode collapse, which lead the poor generalization abilities ultimately. Moreover, most of these models only consider the extraction of spectral or spatial features. They fail to combine the two branches interactively and ignore the correlation between them. Consequently, the variational generative adversarial network with crossed spatial and spectral interactions (CSSVGAN) was proposed in this paper, which includes a dual-branch variational Encoder to map spectral and spatial information to different latent spaces, a crossed interactive Generator to improve the quality of generated virtual samples, and a Discriminator stuck with a classifier to enhance the classification performance. Combining these three subnetworks, the proposed CSSVGAN achieves excellent classification by ensuring the diversity and interacting spectral and spatial features in a crossed manner. The superior experimental results on three datasets verify the effectiveness of this method.

Food recognition is one of the core functions for a smart refrigerator. But there are many challenges for accurate food recognition due to reasons of too many kinds of food inside the refrigerator which tends to obscure each other, and they may look very similar. This paper proposes a fruit recognition approach that fuses weight information and multi deep learning models. The proposed approach can remarkably improve recognition accuracy. We have extensively evaluated the proposed approach for its performance and accuracy, which demonstrate the effectiveness of the proposed approach.

Recently, hyperspectral image (HSI) classification has attracted increasing attention in the remote sensing field. Plenty of CNN-based methods with diverse attention mechanisms (AMs) have been proposed for HSI classification due to AMs being able to improve the quality of feature representations. However, some of the previous AMs squeeze global spatial or channel information directly by pooling operations to yield feature descriptors, which inadequately utilize global contextual information. Besides, some AMs cannot exploit the interactions among channels or positions with the aid of nonlinear transformation well. In this article, a spectral-spatial network with channel and position global context (GC) attention (SSGCA) is proposed to capture discriminative spectral and spatial features. Firstly, a spectral-spatial network is designed to extract spectral and spatial features. Secondly, two novel GC attentions are proposed to optimize the spectral and spatial features respectively for feature enhancement. The channel GC attention is used to capture channel dependencies to emphasize informative features while the position GC attention focuses on position dependencies. Both GC attentions aggregate global contextual features of positions or channels adequately, following a nonlinear transformation. Experimental results on several public HSI datasets demonstrate that the spectral-spatial network with GC attentions outperforms other related methods.

Summary Accurate resource requests prediction is essential to achieve optimal job scheduling and load balancing for cloud Computing. Existing prediction approaches fall short in providing satisfactory accuracy because of high variances of cloud metrics. We propose a deep belief network (DBN)-based approach to predict cloud resource requests. We design a set of experiments to find the most influential factors for prediction accuracy and the best DBN parameter set to achieve optimal performance. The innovative points of the proposed approach is that it introduces analysis of variance and orthogonal experimental design techniques into the parameter learning of DBN. The proposed approach achieves high accuracy with mean square error of [10−6,10−5], approximately 72% reduction compared with the traditional autoregressive integrated moving average predictor, and has better prediction accuracy compared with the state-of-art fractal modeling approach. Copyright © 2016 John Wiley & Sons, Ltd.

Context-awareness for big data applications is different from that of traditional applications in that it is getting challenging to obtain the contexts from big data due to the complexity, velocity, variety, and other aspects of big data, especially big video data. The awareness of contexts in big data is more difficult, and should be more in-depth than that of classical applications. Therefore, in this paper, we propose an in-depth context-awareness framework for a pervasive video cloud in order to obtain underlying contexts in big video data. In this framework, we propose an approach that combines the historical view with the current view to obtain meaningful in-depth contexts, where deep learning techniques are used to obtain raw context data. We have conducted initial evaluations to show the effectiveness of the proposed approach in terms of performance and also the accuracy of obtaining the contexts. The evaluation results show that the proposed approach is effective for real-time context-awareness in a pervasive video cloud.

Machine learning libraries are integral to a big data mining platform. There are three limitations on adopting current machine learning libraries in such a platform. First, these algorithms are not implemented for handling both online and offline big data analysis. Second, libraries exist in a variety of frameworks using different programming languages, which make it difficult in integrating several algorithms. Third, most machine learning libraries provides APIs for programming only, thus not user-friendly for those do not have a sufficient understanding of algorithms and those lack of programming skills. In this paper, we implement a comprehensive machine learning library including common algorithms and deep learning algorithms. We integrate this library at a platform level that allows both online and offline data analysis using this library. We further design a user-friendly portal that enables quick and agile data analysis practices. All of these form an Online-Offline Combined Big Data Mining Platform (OOBDP). We present a demonstration of big oil data analysis using this platform. We observe the that OOBDP can easily accommodate industrial requirement for adaptable data mining process, with personalized usage scenarios, and easy to use experiences.

Phellinus is a kind of fungus and known as one of the elemental components in drugs to avoid cancer. With the purpose of finding optimized culture conditions for Phellinus production in the lab, plenty of experiments focusing on single factor were operated and large scale of experimental data was generated. In previous work, we used regression analysis and GA Gene-set based Genetic Algorithm (GA) to predict the production, but the data we used depended on experimental experience and only little part of the data was used. In this work we use the values of parameters involved in culture conditions, including inoculum size, PH value, initial liquid volume, temperature, seed age, fermentation time and rotation speed, to establish a high yield and a low yield classification model. Subsequently, a prediction model of BP neural network is established for high yield data set. GA is used to find the best culture conditions. The forecast accuracy rate more than 90% and the yield we got have a slight increase than the real yield.

Welan gum is a kind of novel microbial polysaccharide, which is widely produced during the process of microbial growth and metabolism in different external conditions. Welan gum can be used as the thickener, suspending agent, emulsifier, stabilizer, lubricant, film-forming agent and adhesive usage in agriculture. In recent years, finding optimal experimental conditions to maximize the production is paid growing attentions. In this work, a hybrid computational method is proposed to optimize experimental conditions for producing Welan gum with data collected from experiments records. Support Vector Regression (SVR) is used to model the relationship between Welan gum production and experimental conditions, and then adaptive Genetic Algorithm (AGA, for short) is applied to search optimized experimental conditions. As results, a mathematic model of predicting production of Welan gum from experimental conditions is obtained, which achieves accuracy rate 88.36%. As well, a class of optimized experimental conditions is predicted for producing Welan gum 31.65g/L. Comparing the best result in chemical experiment 30.63g/L, the predicted production improves it by 3.3%. The results provide potential optimal experimental conditions to improve the production of Welan gum.

Identifying protein coding regions in DNA sequences by computational methods is an active research topic. Welan gum produced by Sphingomonas sp. WG has great application potential in oil recovery and concrete construction industry. Predicting the coding regions in the Sphingomonas sp. WG genome and addressing the mechanism underlying the explanation for the synthesis of Welan gum metabolism is an important issue at present. In this study, we apply a self adaptive spectral rotation (SASR, for short) method, which is based on the investigation of the Triplet Periodicity property, to predict the coding regions of the whole-genome data of Sphingomonas sp. WG without any previous training process, and 1115 suspected gene fragments are obtained. Suspected gene fragments are subjected to a similarity search against the non-redundant protein sequences (nr) database of NCBI with blastx, and 762 suspected gene fragments have been labeled as genes in the nr database.

There arises the needs for fast processing of continuous video data using embedded devices, for example the one needed for UAV aerial photography. In this paper, we proposed a distributed embedded platform built with NVIDIA Jetson TX1 using deep learning techniques for real time video processing, mainly for object detection. We design a Storm based distributed real-time computation platform and ran object detection algorithm based on convolutional neural networks. We have evaluated the performance of our platform by conducting real-time object detection on surveillance video. Compared with the high end GPU processing of NVIDIA TITAN X, our platform achieves the same processing speed but a much lower power consumption when doing the same work. At the same time, our platform had a good scalability and fault tolerance, which is suitable for intelligent mobile devices such as unmanned aerial vehicles or self-driving cars.

In this paper, we propose a food image recognition system with convolutional neural networks(CNN), which has been applied to image recognition successfully in the literature. A CNN which consists of five layers has been built and two group of controlled trials have been processed on it. Two datasets are prepared: one is UEC-FOOD100 dataset which is an open 100-class food image dataset including about 15000 images and the other is a fruit dataset that established by ourselves including over 40000 images. We have achieved the best accuracy of 80.8% on the fruit dataset and 60.9% on the multi-food dataset. In addition, we validate the method on two groups of controlled trials and discover the effect of color under various conditions that the color feature is not always helpful for improving the accuracy by comparing the results of two group of controlled trials. As future work, we will combine image segmentation with image recognition to get a better performance.

Vehicle detection and type recognition are important for intelligent transportation systems in smart cities. The real time high accuracy recognition with affordable hardware is a challenging issue due to the complexities of video data. In this paper, we propose an integrated approach that combining traditional three-frame difference and deep Convolutional Neural Networks (DCNNs) to detect vehicle and recognize vehicle type in traffic videos captured with fixed mounted cameras. This integrated approach can take advantage of the real-time motion detection ability of three-frame difference and capabilities of image recognition of DCNNs. We have evaluated the proposed approach using road traffic videos in terms of accuracy and performance, which show very promising results.