Search

Traditional visual simultaneous localization and mapping (SLAM) systems mostly based on small-area static environments. In recent years, some studies focused on combining semantic information with visual SLAM. However, most of them are hard to obtain better performance in the large-scale dynamic environment. And the accuracy, rapidity of the system still needs to strengthen. In this paper, we develop a more efficient semantic SLAM system in the two-wheeled mobile robot by using semantic segmentation to recognize people, chairs, and other objects in every keyframe. With a preliminary understanding of the environment, fusing the RGB-D camera and encoders information, to localization and creating a dense colored octree map without dynamic objects. Besides, for the incomplete identification of movable objects, we used image processing algorithms to enhance the semantic segmentation effect. In the proposed method, enhanced semantic segmentation in keyframes dramatically increases the efficiency of the system. Moreover, fusing the different sensors can highly raise localization accuracy. We conducted experiments on various datasets and in some real environments and compared them with DRE-SLAM, DS-SLAM, to evaluate the performance of the proposed approach. The results suggest we significantly improve the processing efficiency, robustness, and quality of the map.

Since generative adversarial network (GAN) was first proposed, the processing of face images, especially the research of facial attribute editing, has attracted much interest. It not only can alleviate the problems associated with data deficiency, but also has great applications in the field of entertainment. However, existing approaches have limited scalability in the processing of newly-added face attributes, and the quality of generated images is poor. To solve these problems, FEGAN is proposed in this paper to achieve the accurate editing of multi-attribute faces by modifying feature vectors in the latent space. Firstly, a trained generator is used, which greatly reduces the training difficulty of GANs, and the inverse of the generator is used to establish the unique correspondence between the input image and the latent code. Secondly, a linear guide is applied to the latent code, and thus the same distribution as the target image in the latent space is assured. Finally, a generator is used to generate a face image from the guided latent code. The proposed method is utilized for a large number of attribute editing experiments, and the results show that FEGAN can flexibly perform accurate attribute editing while guaranteeing that other areas are not changed. Both qualitative and quantitative results demonstrate its advantages over existing methods.

Presently, finger-vein recognition is a new research direction in the field of biometric recognition. The Gabor filter has been extensively used for finger-vein recognition; however, its parameters are difficult to adjust. To solve this problem, an adaptive-learning Gabor filter is presented herein. We combine convolutional neural networks with a Gabor filter to calculate the gradient of the Gabor-filter parameters, based on the objective function, and to then optimize its parameters via back-propagation. The parameter θ of Gabor filter can be trained to the same angle as the vein texture of finger vein image. The parameter σ of Gabor filter has a certain relation with λ, and the parameter λ of Gabor filter can converge to the optimal value well. Using this method, we not only select appropriate and effective Gabor filter parameters to design the filter banks, we also consider the relationship between those parameters. Finally, we perform experiments on four public finger-vein datasets. Experimental results demonstrate that our method outperforms state-of-the-art methods in finger-vein classification.

The use of infrared or electroluminescence(EL) images of solar cell modules for defect detection is a very important method in non-destructive testing. Traditionally, this work is done by skilled technicians, which is time-consuming and susceptible to subjective factors. The surface defect detection method of solar cells based on machine learning has become one of the main research directions because of its high efficiency and convenience. For this reason, this paper proposes an improved fusion model based on VGGNet and U-Net++, which is used for defect detection and segmentation of EL images of solar cells. In the defect detection stage, the input image is processed pertinently, and by modifying the convolutional layer and the fully connected layer of the network, while improving the performance of the algorithm, it accelerates the convergence and avoids the phenomenon of over-fitting. In the defect segmentation stage, the defect location is marked based on the public data set, which is used for the training of each segmentation model, and the effect of different segmentation networks is compared to select a reasonable model. The experimental results show that the defect detection accuracy of the improved VGG16 network on the elpv-dataset is 95.2%, and the U-Net++ defect segmentation model has an average MIoU value of 0.955, which is better than other existing methods.

Local feature descriptors exhibit great superiority in finger vein recognition due to their stability and robustness against local changes in images. However, most of these are methods use general-purpose descriptors that do not consider finger vein-specific features. In this work, we propose a finger vein-specific local feature descriptors based physiological characteristic of finger vein patterns, i.e., histogram of oriented physiological Gabor responses (HOPGR), for finger vein recognition. First, a prior of directional characteristic of finger vein patterns is obtained in an unsupervised manner. Then the physiological Gabor filter banks are set up based on the prior information to extract the physiological responses and orientation. Finally, to make the feature robust against local changes in images, a histogram is generated as output by dividing the image into non-overlapping cells and overlapping blocks. Extensive experimental results on several databases clearly demonstrate that the proposed method outperforms most current state-of-the-art finger vein recognition methods.

Knowledge graphs have become increasingly popular over the past years. The most popular data model for representing knowledge graphs is the Resource Description Framework (RDF). In an open Web environment, it is common case that the RDF knowledge graphs data from different sources may often contain inconsistent or imprecise information. Efficient querying of fuzzy RDF knowledge graphs is therefore of increasing importance. In this paper, we firstly introduce practical extensions of the RDF model to represent vagueness based on fuzzy graph. Then we extend the concept of SPARQL graph pattern making it possible to query an element-level fuzzy RDF knowledge graphs with high degree of truth. Finally, we present a backtracking algorithm for calculating the set of homomorphisms from fuzzy SPARQL graph into a fuzzy RDF graph to demonstrate our proposed methods.

Image quality assessment is an important and necessary task in the field of image processing. It can simulate human visual perception accurately and effectively to ensure the credibility of information. Although the existing IQA algorithm based on CNN has achieved excellent success, the generalization and robustness of the algorithm are limited due to the loss of image information during feature extraction. The research shows that the phase and amplitude of image frequency domain will change with the quality, so we proposes an image quality assessment algorithm based on dual domains fusion (DualD-IQA). Using the frequency domain and spatial domain as multiple inputs of convolutional neural network, we can complement each other to represent image quality related information. Moreover, the input of any scale can be accepted by adding bilinear pooling, so as to ensure the reliability and robustness of the quality evaluation results. Experimental results show that the algorithm in this paper achieves higher consistency and accuracy in two commonly used public databases, and has higher robustness in different distortion types and cross databases.

Face recognition system performance is affected by face image quality. In this paper, a face matching score, instead of a subjective perceptual score, is used as the image quality label to build a face image quality database; therefore, the ground truth quality achieved high accordance with the recognition performance. The database contains five common practical application distortion types for 19 people and three to five collection scenes for each. In addition, a face image quality algorithm based on a local gradient is proposed to judge the image quality. Making use of the fact that face landmarks can locate contour areas that have rich gradient information, a new mask based on human visual characteristics is designed to evaluate the image quality by using the local gradient information of landmark neighborhoods; this method was used to control for the influence of the image content and background. Our algorithm does not rely on a large amount of training data, and it has low computational complexity. The experimental results show that the algorithm is effective and efficient for evaluating image quality and improving the performance of face recognition systems.

Spatiotemporal information and application have received considerable attention. Spatiotemporal information is often imprecise or uncertain; therefore, the establishment of a reasonable and effective fuzzy spatiotemporal data model is vital. The traditional tool-based modeling method cannot meet the needs of fuzzy spatiotemporal data modeling. XML has become the standard of Web data representation and exchange. Based on the fuzzy set and probability theory, this paper analyzes the characteristics of fuzzy spatiotemporal data and establishes a fuzzy spatiotemporal XML data model. We present a fuzzy spatiotemporal UML data model and provide the rules for converting the fuzzy spatiotemporal UML data model to the fuzzy spatiotemporal XML model. We demonstrate our model and the conversion by applying them to a meteorological event.

Recently, deep neural networks (DNNs) have been applied in various areas in computer vision. However, due to the simple neuron configuration, the existing DNNs rely heavily on the diversity of the data distribution and have poor generalization performance. In addition, the excessive network complexity also leads to low learning efficiency. To solve above problems, a topological higher-order neuron model was proposed, which is modeled according to the point-set topology in high-dimensional space. The proposed neuron is determined by the adaptively changed center and the direction weights with strong generalization ability. The high-dimensional feature space of the samples is constructed by a category-based covering learning method, thus achieving an optimal coverage of each category. Experiments on several different classical datasets demonstrate the generalization of the proposed neuron model, which provides a new approach for the further development of DNNs and can be widely used in computer vision.

On-disk data copy is commonly performed in various software. Its current implementation with read and write commands needs to go through the I/O stack, introducing significant performance overhead. Actually when only one disk is involved in a copy operation, the data do not have to be read out of the device and the operation can be all performed within it. Furthermore, if the device is an SSD supporting flexible mapping from logical to physical addresses, the data do not need to be physically duplicated and the data copy purpose can be fulfilled by establishing a new address mapping. In the paper, we propose a copy primitive for SSD to support copy of block(s) of data within a device with an almost zero cost regardless of amount of data. We show that it is relatively easy to implement the primitive, named copyless copy, or CC in short. We also evaluate two example uses of CC, database journaling and block-level deduplication, where the new primitive can help for efficient data copy as the case study. Our experiments with the use cases show that CC can dramatically increase the software's performance, reduce data writes to the device for improving SSD's endurance and space efficiency.

Since there exist some issues with traditional search engine in information retrieval, such as huge numbers of results, poor profession, low precision rate and other issues, in this paper, we proposed a Vertical Search Engine based on Nutch and Solr. We used forward iteration most granular segmentation algorithm based on dictionary to achieve Chinese word segmentation, employed Vector Space Model (VSM) based on keywords to implement topic relevance, extended the user search module and the tourism domain word library to collect information, filter information retrieval, and relate word various stages. Experiments were also conducted in order to evaluate the algorithm and the results show that the vertical search engine based on Nutch and Solr which is used for tourism information retrieval can improve the user retrieval precision and meet the professional demand of user retrieval.

Finger-vein recognition has become an important branch in the field of biometrics. Convolutional neural networks (CNNs) have achieved remarkable success in finger-vein recognition, but their models have many problems, such as high complexity, large parameters, running time slowly and so on. To solve these problems, we proposed adaptive Gabor convolutional neural networks (AGCNN). We replaced the normal convolution layer of CNNs with Gabor convolution layer, and systematically investigate comparative performance using AGCNN and CNNs in the raw image and region of interest (ROI) image of finger vein database respectively. Experimental results show that the parameter complexity of AGCNN is significantly less than that of CNNs with slight performance decrease. And the experimental results of the raw image database show that the processing speed has been greatly improved.

I/O deduplication is an important technique for saving I/O bandwidth and storage space for storage systems. However, it requires a new level of address mapping, and consequently needs to maintain corresponding metadata. To meet requirements on data persistency and consistency, the metadata writing is likely to make deduplication operations much fatter, in terms of amount of additional writes on the critical I/O path, than one might expect. In this paper we propose to compress the data and insert metadata into data blocks to reduce metadata writes. Assuming that performance-critical data are usually compressible, we can mostly remove separate writes of metadata out of the critical path of servicing users' requests, and make I/O deduplication much thinner. Accordingly we name the scheme ThinDedup. In addition to metadata insertion, ThinDedup also uses persistency of data fingerprints to evade enforcement of write order between data and metadata. We have implemented ThinDedup in the Linux kernel as a device mapper target to provide block-level deduplication. Experimental results show, compared to existing deduplication schemes, ThinDedup achieves (much) higher (up to 3X) I/O throughput and lower latency (reduced by up to 88%) without compromising data persistency.

In a multi-tenant cloud environment, tenants are usually hosted by virtual machines. Cloud providers deploy multiple virtual machines on a physical server to better utilize physical resources including CPU, memory, and storage devices. SSDs are often used as an I/O cache shared among the tenants for large storage systems using hard disk drives (HDDs) as their main storage devices, which can receive much of SSD's performance benefit and HDD's cost advantage. A key challenge in the use of the shared cache is to ensure strong performance isolation and maintain its high utilization at the same time. However, conventional SSD cache management approaches cannot effectively address this challenge. In this paper, we propose OC-Cache, an open-channel SSD cache framework which utilizes SSD'd internal parallelism to adaptively allocate cache to tenants for both good performance isolation and high SSD utilization. In particular, OC-Cache uses a tenant's miss ratio curve to determine the amount of cache space allocation and where the allocation is (in dedicated or shared SSD channels) and dynamically manages cache space according to the workload characteristics. Experiments show that OC-Cache significantly reduces interference among tenants, and maintains high utilization of the SSD cache.

A large number of datasets in various applications are imbalanced in which majority samples dominate minority samples. The skewed distribution poses a difficulty for existing learning approaches. Oversampling techniques address this concern by replicating original samples or adding new synthetic samples of minority class. Even with success, they raise the problems of over-generation and overlapping. In this paper, we propose an entropy difference-based oversampling approach (EDOS) for imbalanced learning using a novel metric, termed entropy difference (ED). First, given a dataset, EDOS measures the imbalance degree between the majority and the minority with ED. Second, EDOS creates synthetic minority samples. For each synthetic sample, EDOS evaluates its retention capability and remains the informative sample. Third, original and qualified synthetic samples are combined to train the classifiers. In the experiments, we demonstrate the effectiveness of the proposed EDOS method on several UCI datasets.

Many applications require the use of spatiotemporal data, which imposes increasing requirements on spatiotemporal data modeling. Fuzziness widely exists in spatiotemporal data and the existing studies consider only fuzzy temporal and spatial topological relations. Therefore, it is imperative to investigate fuzzy temporal relations and fuzzy spatial relations based on representation of conceptual data. This paper proposes an UML class diagram model to represent fuzzy spatiotemporal objects and relations. We review traditional direction, metric and topological relations. On the basis, we explore fuzzy temporal relations and fuzzy spatial relations as combinations of these three classic relations. Furthermore, we provide a representation of fuzzy temporal relations and fuzzy spatial relations in UML class diagrams. Finally, we apply our observations to analyze meteorological phenomena.

As a new biometric technology, finger vein recognition has received much attention in recent years. Since deep learning is an end-to-end system, and has achieved very good results in such fields as face recognition and target detection, we try to apply it to finger vein recognition. In this paper, we adopt seven layers of CNN which include 5 convolution layers and 2 fully-connected layers. This network obtains a recognition rate of 99.53%, which proves to be better performing than traditional algorithm.

As a method of noninvasive fetal monitoring, fetal electrocardiogram (FECG) signal processing has become a focus in current research, resulting in many methods of acquiring FECG. The extraction of FECG using blind source separation algorithms is feasible, which based on linear mixed model, and usually is assumed only one fetal signal mixed with the maternal signal. However, pregnant women may harbor multiple births in reality, so previous conventional methods are not applicable. Therefore, a novel method of estimating the number of FECG signals is proposed in this paper, which can prejudge the number of mixed-signals, thereby selecting corresponding FECG extraction algorithms or using other better extraction algorithms. Finally, the simulation shows that the algorithm is feasible and accurate to estimate the number of FECG signals.

This paper proposed a image fog removal method. The method based on the disinhibition properties of retinal neurons concentric receptive field whose function is three Gaussians. Firstly, we enhance original fog image by using contrast limited global histogram equalization. Secondly, a local image enhancement method is carried out to restore the image's details and deepness information. According to comparison experiment, the results of this paper's method were compared with the results of dark channel prior based method and the results of CLAHE (Contrast Limited Adaptive Histogram Equalization). The experimental results proved that the method of contrast limited global histogram equalization was able to avoid enhancing too light and to avoid adding noising. In addition, local details enhancement by three Gaussians strengthened the image's depth information. The method of this paper worked well in images fog removing and enhanced fog images' visibility.

Extraction of fetal electrocardiogram (FECG) is currently focused as a prominent issue of medical research. Normally, FECG can be extracted from the maternal abdominal recordings by conventional blind source separation algorithm. However, the recordings need to meet the linear mixing model of blind source separation (BSS), which makes it difficult to apply in the practical environment. In view of this fact, our paper introduces a discriminant method of BSS based on FECG correlations analysis. The sparse areas of signals are located firstly, and then the correlation values of these areas are calculated out. A judging criterion based on these values can be built to decide that if the recordings meet the model of BSS. Therefore, FECG can be extracted from these observations by conventional BSS methods in practical. The simulation experiments also show that our discriminant method is available and can obtain better performance of extraction.

A method is presented for obtaining the equilibrium outcomes of hypergame models with fuzzy strategy and preference perceptions in this paper. Firstly, for a decision maker specified credibility level λ, the crisp strategy set and outcome preference perception are determined based on fuzzy cuts and a relative distance measure for fuzzy numbers. Then, the hypergame-preserving λ-equilibrium and hypergame-destroying λ-equilibrium are defined and obtained by hypergame stability analysis. A simulation example illustrates the proposed method at the end.

The 3rd Generation network rely on a key protocol—the GPRS Tunneling Protocol (GTP). However, GTP itself is not designed to be secure. So it is necessary to inspect the legal GTP channels and protect the backbone. Because a great deal of GTP-U packets transferring in the channels and the 3G firewall is required to check ten thousand GTP channels which needs mass bytes, so a fast and saving space algorithm is needed. A Bloom Filter algorithm is used to inspect the GTP-U flows. This method takes advantages of multi-threads and hash units of network processor. The analyses prove that the method only store some bit vectors and corresponding counters which needs less space. The experiments show this method can inspect of 1Gbps flow in the GTP channel.

An effective face detection system used for detecting multi-pose frontal face in gray images is presented. Image preprocessing approaches are applied to reduce the influence of the complex illumination. Eye-analog pairing and improved multiple related template matching are used to glancing and accurate face detecting, respectively. To shorten the time cost of detecting process, we employ prejudge rules in checking candidate image segments before template matching. Test by our own face database with complicated illumination and background, the system has high calculation speed and illumination independency, and obtains good experimental results.