Search

Pedestrian object detection and tracking have attracted much attention in the computer vision field. An improved multi-pedestrian tracking model is proposed, which improves the basic framework of Deep SORT and integrates Mask R-CNN to realize the detection, tracking and pose estimation f pedestrian. The anchor boxes with the more suitable aspect ratio for pedestrian target are adopted, which replace the anchor boxes of RPN to speed up the model and improve performance without complex calculation. In addition, attention mechanism is introduced into the deep residual network, i. e., the lightweight SKNet is used to choose the best convolution kernel adaptively to improve the feature representation for target detection. The histogram of gradient feature combined with color information is adopted instead of the convolution feature, which improves appearance feature association matching in the Deep SORT model so as to track pedestrian targets effectively under occlusion. The impact of various improvements on the model are verified through ablation studies, and the proposed model is compared with the current mainstream model. Experimental results show that the improved models are effective, which improves MOTA of NSH by 6% on the MOT16 tracking data set. The test performance of our proposal on the public datasets is superior to that of the compared models. The proposed model can still rack pedestrian targets effectively when the background moves or pedestrian targets are occluded. [ABSTRACT FROM AUTHOR]

Solving a large-scale convex quadratic programming problem is the core of the support vector machine. An equivalence complementarity problem can be obtained based on an amended model of the support vector machine(SVM), therefore we propose a non-monotonic trust region algorithm for solving the complementarity problem based on the Fischer-Burmeister complementarity function. The new algorithm need not compute any Hesse or the inverse matrix, thus reducing the amount of computational work. Global convergence of the algorithm is proved without any assumptions. Numerical experiments show that the running speed of the algorithm is faster than that of the LSVM algorithm and the descent algorithm when solving large-scale nonlinear classification problems and thus it provides a feasible method for solving SVM. [ABSTRACT FROM AUTHOR]

Locality sensitive hashing(LSH)is quite popular in high dimensional data indexing. The Hash function in the original LSH algorithm is generated based on p-stable distribution. So the number of hash tables must be increased in order to improve the performance of the algorithm, which however leads to a high space complexity. An improved LSH(I-LSH)algorithm is proposed, which does not require the labeled samples but only uses the distribution of data to construct the projection direction. The experimental results show that I-LSH can greatly save the memory without degrading its retrieval performance. [ABSTRACT FROM AUTHOR]

Endometrial carcinoma (EC) is one of the most common malignant female genital carcinomas, and the incidence has been increased recently. Although it occurs mainly in the postmenopausal women, there are still 25% patients are premenopausal and 3%-14% patients are below 40 years old. Nevertheless, the standard surgical treatment is associated with the loss of their ability to fertility and the endocrine function of the ovaries among young women, which seriously affect the quality of their life. Nowadays, to preserve the fertility for the younger EC patients, the long-term use of high-dose progesterone is the main therapy of atypical endometrial hyperplasia and EC. However, progesterone resistance appears in many cases during the different stages, which reduces the effect of hormone therapy and influences the patients' prognosis. Therefore, the further study of progesterone resistance and sensitization mechanisms is of great significance to improve the curative effect and preserve the fertility and endocrine function for the early stage EC patients. This article reviewed the progesterone resistance mechanisms and sensitizing ways of EC. [ABSTRACT FROM AUTHOR]

To solve the security problem of start-up procedure in virtualization system, bringing the idea of trust measurement and trust chain transfer in the boot process of virtual machine monitor, this paper proposed a trusted boot scheme of collaborative VMM. Considering the complexity of start process and diversity of measurements, the scheme used SHA-1 algorithm to test the integrity of the kernel, the program code and the configuration files on the key points of boot process. The system would terminate its startup process when found the data abnormal so that it could avoid disclosure of sensitive information and enforce the system security. The test shows that the scheme of trusted start-up procedure of virtualization system can ensure the security and credibility during system startup and kernel self-starting process. The scheme also provides a feasible solution for trusted boot of virtualization system while the 9. 8% of time overhead on system performance is not obvious. [ABSTRACT FROM AUTHOR]

In order to study the performance of concrete frame beams reinforced with HR.B500 steel bars, three single story, two-bay concrete frames were tested under static loading. During testing the displacements, rotation angles, cracks, and strains of concrete and steel of the beam were measured under mid-span loading. It is shown by calculation that in serviceability limit state, the maximum measured crack width at the mid-span of the frame beam is generally smaller than that calculated according to the Chinese design code, while the crack width measured' at the beam end is larger than that calculated according to the code. The mechanism of the crack at the beam end was analyzed. On the basis of bond-slip theory the slip between longitudinal steel bar and concrete was divided into two parts: slip in the beam-column joint and in end region of the beam. Then the slip model of tension steel bar in the negative bending moment region and the formula of maximum crack width were developed. Based on the statistic analysis of test results the parameters in the model were calibrated. The proposed formula of crack width of the frame beam under maximum negative bending moment can be used in engineering practice and as a reference for the modification of the code. [ABSTRACT FROM AUTHOR]

Sunipia including about 23 species was restricted to Southeast Asia. In China, genus of Sunipia is represented by twelve species. During our fieldwork in Pu'er Prefecture southern Yunnan from 2011 to 2013, one new recorded species from China was discovered and here reported. Its Chinese name was entitled by meaning of specific names. This species is characterized by purple perianth, unbranched pollinarium stipe terminated with 4 spherical solid pollinias differed from others of this genus. [ABSTRACT FROM AUTHOR]

This paper proposed a nonmonotonic trust region algorithm for solving inverse quadratic programming problems. It transformed the inverse quadratic programming problem into its dual problem which had a relatively small number of decision variables. Then it transformed the sub-problem into unconstrained optimization problem by using the general augmented Lagrange method. Numerical experiments show that the proposed algorithm runs faster than Newton algorithm and alternating direction method with Gaussian back substitution, it is more effective for large scale quadratic programming problems. [ABSTRACT FROM AUTHOR]

Currently, 3D CT images play an important role in the lossless detection of industry, and the focus analysis of medicine are widely used in the industrial and the medicinal field. In order to detect the line targets widely existed in the CT images, traditional methods were Hough transform and Radon transform. However the complexity of them were huge in the 3D space, what's more, they were not applicable by curve targets. So this paper proposed a method based on distance transform and terminal node, turned nodes detection to detect line targets. It was not only suitable to the line, but also to the curve. Furthermore, it could get the thickness of the line targets automatically from the step of distance transform. The result of experiments shows the availability of this method. [ABSTRACT FROM AUTHOR]

Support vector machine (SVM) parameter optimization selection is an important research direction, but there is still no systematic theory to guide the selection of the SVM parameters. Since optimizing the SVM parameters by the artificial fish-swarm algorithm tends to fall into the small neighborhood of the approximate optimal solution, we design the AFMC algorithm for the SVM parameter optimization. At the early stage, we use the better parallel optimization performance of the fish-swarm algorithm to quickly gain the approximate optimal solution. Then we use the MonteCarlo algorithm for local searching to achieve a quick and effective strong-approximate optimal solution. The numerical experiments show that the proposed algorithm has better classification performance and faster searching speed, and it is effective and feasible in the SVM parameter optimization. [ABSTRACT FROM AUTHOR]

This paper proposed a small current to ground fault line selection method with time-frequency atom and grey correlation analysis. Firstly, it calculated the base values of fault line and non-fault line based on the number of branch line of small current to ground system. Secondly, it used the matching pursuit method to decompose the transient zero-sequence current of the first 1/4 cycle indexed by Gabor atom, and calculated the damped sinusoidal atoms of every branch line which expressed the characteristic of fault state. Thirdly, it used the theory of grey relational analysis to compute the correlation degree of every damped sinusoidal atom, and got the characteristic value of every branch line. Finally, it calculated the Euclidean distance value between the characteristic values and the base values of fault line and non-fault line, and compared the size of Euclidean distance values to achieve the fault line selection accurately. The method has a lower computational complexity, an accurate fault selection proved by a large number of simulations. It is especially suitable for the multi-branch lines of small current to ground system. [ABSTRACT FROM AUTHOR]

A joint type-II censoring model was analyzed under the small sample size. Aiming at the problem that units may be exposed to more than one fatal failure factors, we focused on competing risks model with independent exponential data. Based on the maximum likelihood estimator (MLE), we developed the exact distributions of parameters and exact confidence intervals through the use of the moment generating functions. Furthermore, approximate confidence intervals were also constructed by Bootstrap method. A Monte-Carlo simulation was conducted to compare the performances of different estimations. And finally, an example was discussed to illustrate the validity of the method. [ABSTRACT FROM AUTHOR]

In order to measure the depths of the objective sound source in shallow water and virtual sound sources under muhi-path conditions, an abeam energy maximum measurement method was proposed by analyzing underwater image mechanism in depth measurement under the ideal and multi-path conditions. It can also realize the depth survey of level line-array by analyzing symmetric relation of the real source position and the virtual source position. In this paper, the simulation results and experimental data are analyzed and compared, the accuracies of the depth scanning under different conditions are discussed, and the accurate localization results for the target source are given. [ABSTRACT FROM AUTHOR]