Search

Recent research has shown Convolutional Neural Networks (CNNs) trained in a fully-supervised fashion achieve promising performance on extreme low-light image denoising task. However, a large amount of "noisy-clean" image pairs are required to train a network, which are difficult to obtain. In this paper, we propose a compact yet effective noise model to generate synthetic noisy images for training. Especially, we address the severe color distortion problem in low-light images by identifying a novel noise component, black calibration error, as its physical origin. We prove that a small error at the sensing stage will strongly affect the following in-camera signal processing (ISP) pipeline and eventually lead to color bias. Experiment results demonstrate that the proposed model is superior in preserving perceptual quality and achieves state-of-the-art performance among existing noise synthesis methods.

Surgery is the best treatment for liver canner, segmentation of livers is one of the most crucial steps in surgery. The segmentation task is often done manually, making it tedious, labor intensive, and subject to intra-/inter- operator variations. Automatic segmentation of livers remains intractable owing to their low tissue contrast with respect to the surrounding organs and their deformable shape in CT images. The paper proposes a new liver segmentation architecture named DU-Net, which is composed of cascade U-Net based on deep separable convolution and deconvolution. We make use of the Dice loss function for training. Finally, we evaluate the proposed DU-Net on liver segmentation task under the dataset provided by the public 3D-IRCADb (3D Image Reconstruction for Comparison of Algorithm Database).The experimental result demonstrates that the proposed network can outperform the U-Net, FCN and RV-FCN methods in term of the performance measures, such as Dice score, VOE, and RVD.

Emotion, as an advanced function of the human brain, affects kinds of human behaviors. Electroencephalographs (EEG) are widely used in the field of emotion classification owing to their low cost and portability. In this work, we study the effects of a non-linear EEG feature and a channel selection method on emotion recognition. First, the fractal dimension(FD) which could reflect the state of the brain is extracted with a sliding window. The top seven channels are screened out by calculating the F-score from the whole samples. Then, based on the signals from forehead channels, filtered channels and associated channels, emotions on valence and arousal are classified by Support Vector Machine(SVM) and K Nearest Neighbours(KNN). The result shows that the forehead channels Fp2, AF8, Fpz play an important role in valence classification. When combining the forehead channels with other channels that have higher F-score, the SVM classifier has a better accuracy on the whole set with 89.37% on valence and 87.07% on arousal. Besides, the overall accuracy calculated on each participants with associated channels get significant improvement. Especially, the KNN classifier has a much better result on every subject. This phenomenon indicates that by combining the higher F-score channels with the forehead channels, the associated channels can not only take advantage of the forehead channels' ability to categorize emotions but also consider individual differences.

Tens of millions of people suffer from depression worldwide. It is urgent to explore an effective method for diagnosing depression. This study developed a novel of multimodal feature fusion depression recognition method based on functional near-infrared spectroscopy (fNIRS). Sixty volunteers, including thirty patients with depression and thirty healthy controls, participated in the study. The 22-channel fNIRS device recorded the participants’ brain oxyhemoglobin (HbO) and deoxyhemoglobin (HbR) concentration changes in the positive, neutral and negative affective words’ stimulation. K-nearest neighbors (KNN) and support vector machine (SVM) classifiers were used to recognize depressed patients from normal people, and 10-fold cross-validation was used to verify the classification result. Under the three single-mode features, the accuracy rates were 85.69%, 88.32% and 86.77%, corresponding to the positive condition, neutral condition and negative condition. Then, we used concatenation and linear combination for feature fusion. For the concatenation fusion method, the principal component analysis (PCA) was used to reduce the dimension. The result showed that feature fusion can relatively improve the recognition rate of people with depression, compared with single-model features. The optimal feature fusion method is to concatenate the neutral features and negative features, and the best accuracy reaches 94.45%. The study may provide a more accurate and convenient method for depression detection.

The Type-I censoring test has the advantage of time control which makes it widely used in the engineering practice. The outputs of the current Type-I censoring statistical algorithm are the total test time and the receiving number of failures. However, the total test time has already been settled by the test profile under many circumstances. Furthermore, the current Type-I censoring statistical algorithm is unable to evaluate the system reliability when the number of failure is zero during the system reliability test. Aiming at the problems above, a new Type-I censoring statistical algorithm was proposed to calculate the number of failures to receive the results of the reliability test considering the fixed total test time. And an evaluation method for zero-failure system reliability evaluation was developed. Then an assistant software was built to implement the algorithm. At last, two practical cases showed the importance of the receiving number of failures which should be calculated in the test design stage before the reliability test begins. And they also proved the new algorithm and the software is effective and feasible.

A coupled-fed printed antenna for eight-band covering WWAN/WLAN/WiMAX operation in the mobile phone is presented. The proposed antenna consists of a long radiating strip, a short-circuited inductive shorting strip, and a coupling strip to capacitively excite the radiating strip. By using the capacitive excitation, the long radiating strip with the shorting strip can generate a wide lower band to cover the GSM850/900 operation, and the coupling strip can also form a wide upper band to cover GSM850/900/1800/1900/UMTS2100, WLAN2400, and WiMAX2300/2500 operations. The antenna is well suited to be placed on a small non-ground board space (15 × 45 mm 2 ) of the system circuit board and encased by an L-shaped system ground plane of the mobile phone. A universal series bus (USB) connector for the data port of the mobile phone is accommodated on the protruding ground of the L-shaped system ground plane. Details of the proposed antenna are described, and the obtained results are presented and discussed.

Benefiting from trustworthily tracking of the processes in the production, processing, storage, transportation and sale phases, an electronic pedigree system becomes an important technology of the Internet of Things. In an electronic pedigree system, small-sized but huge volume of electronic pedigrees in the XML format will be generated, stored, and retrieved. Unfortunately, study of these massive electronic pedigrees' storage in an electronic pedigree system, which is in the form of small XML files, is rarely concerned. We, therefore, try to leverage Hadoop to solve the storage problem of massive electronic pedigrees, by the optimization of storing and accessing massive small XML files in HDFS. First, all correlated small XML files of the same envelope are merged into a larger file to reduce the metadata occupation at NameNode. Second, a prefetching mechanism and a remerging mechanism are used to improve the efficiency of accessing small XML files. Finally, we implement a prototype to evaluate the effectiveness and efficiency comparing with the origin HDFS. The results show that the optimized approach is able to reduce the memory consumption of NameNodes by up to 50%, improve performance of storing by up to 91%, and accelerate accessing by up to 88% in Hadoop.

The authentication based on user's username and password is one of the most popular ways to verify a user when he or she enters an information system. Thus, in recent years, many attackers, e.g. phishers, aim to steal passwords to intrude information systems. To strengthen the password-based authentication, we introduce a method, named SmartID, where we use a Bluetooth-enabled Smart Phone as a platform to store user's username, password and their relevant information of login interface. SmartID can help a user authenticate both his or her identify and validity of login interfaces of web sites and desktops. The experiment and analysis show that SmartID can offer an excellent recall rate, and an acceptable precision rate for anti-phishing and anti-pharming. Thus, SmartID is applicable to strengthen password-based authentication.

In this dissertation, the research of distributed fiber grating spectrum form optimization is given. Spatial division multiplexing transform and time division multiplexing are the advantages of distributed fiber grating so that it can highly enhance efficiency and practicality, but much side lobes and bad precision are inherent disadvantages of distributed fiber grating. In accordance with those, BP neural network method is put forward. The result shows that BP neural network method possesses robustness and fault tolerance.

A study on the emergence is beneficial to the understanding and mastery of the basic rules governing the development of Research University's core competence system. The emergence of this system originates from its features, such as non-linearity, self-organized behavior, openness and so forth, as well as the complicated system environment, non-conservation of systematic knowledge and intelligence of the subject. Full play of the material effect, the scale effect, the structural effect and the environmental effect will contribute to the continuous emergence of Research University's core competence system, and therefore the new properties and hierarchies of core competence system emerge.

Along the bank of Songhua River, the aquifer of Xinren Irrigation District between low flood plain and high flood plain has such characteristics as followed: the landform appears long and narrow, the aquifer is with high water permeability, the groundwater table berries very shallow and the variation of surface water can easily influence the groundwater. After sorting out physical geographic information of the researching area, secondary salinization situation is analyzed according to groundwater table data and groundwater depth data. It refers that intensifying groundwater exploitation or water drainage, reducing seepage from surface water and increasing evaporation are 3 useful methods to decline groundwater table. Meanwhile, besides highlighting the importance of the dynamic control of groundwater, the secondary salinization soil problem can be better solved with the combination of agricultural structure adjustment, field management and other measures comprehensively. The methods given in this paper can provide scientific support for managing secondary salinization soil problem and other relevant researches.

a novel broadband single-layer aperture coupled patch antenna is proposed. A simple patch geometry that is based on a rectangular patch with two symmetry arms fed by aperture coupled is introduced in this paper. A new resonant mode is excited using this novel patch structure. The results shows that the bandwidth is 4.1 to 6.3 GHz (S11S11

Bio-mass power cogeneration plant, generating power and heat by using straws, stalks or other biomass energy, is one of the latest types of the energy conservation and pollutant reduction construction projects. Study on water resources feasibility demonstration is one essential approach to achieve sustainable utilization of water resources for construction projects. Based on a typical example of power cogeneration plant in cold region—Yi’an bio-mass power cogeneration plant, strategy and method of water resources study is explicitly analyzed and summarized, which provide an entire operational system of tactic for water resources feasibility study on bio-mass power cogeneration plant especially in cold area and enrich the research and practical theory for water resources feasibility study of construction projects.

With the depth and workload of water resources management increased, it is more efficient to build an automatic generation system to organize and manage the water resources bulletins. A practical automatic generation system for water resources bulletin has designed and built in this paper which includes 3 modules: Figure generation module, Data chart generation module and Bulletin text generation module. In addition, taking Jiamusi (a city in Heilongjiang Province) as an example, a Jiamusi automatic generation system for water resources bulletin has been designed, developed and analyzed, which is expected to provide references for relevant researches.

To precision guided munition (PGM), the sample size of batch of products for inspection is small, but the large number prior information is proposed for accuracy testing of PGM. According to these physical circumstances, the writer does some application research of Bayers` Theorem in accuracy testing assessment of PGM under the condition of small sample size. Firstly, we can gain the the credibility of each prior information source by consistent test, and then count the parameters of distribution on the assumption that the landing error deviation of PGM is in normal de-ganma distribution. Secondly we chose the iheweighted method based on credibility to syncretize the prior information of multiple sources. At last the accuracy of PGM can be calculated by Bayers algorithm. This article demonstrates that the new algorithm on Bayers` Theorem is more better than conventional method and it is more reasonably applied in accuracy testing assessment of PGM.