Entity

Time filter

Source Type


Shen B.,Xiangtan University | Zhang D.,Xiangtan University | Zhang D.,Key Laboratory of Intelligent Computing and Information Processing of Ministry of Education | Peng Y.,Xiangtan University
Communications in Computer and Information Science | Year: 2012

A novel registration algorithm combining global and local registration technology is proposed. In this algorithm, bifurcation structures with 4 connected bifurcation points rather than single bifurcation points are used as the registration features. By reducing feature pairs, the efficiency of registration is improved. Moreover, due to non-linear deformation, there are some register errors appear in some local region. To solve this problem, a local registration technology is introduced to improve registration precision. Experiment results show that the algorithm effectively achieve fundus image registration with higher efficiency and precision. © 2012 Springer-Verlag.


Zhang D.-B.,Xiangtan University | Zhang D.-B.,Key Laboratory of Intelligent Computing and Information Processing of Ministry of Education | Yi Y.,Xiangtan University | Zhao Y.-Y.,Xiangtan University
Chinese Journal of Biomedical Engineering | Year: 2013

In order to utilize the vessel distribution and optic disc (OD) appearance characteristics effectively, a novel OD localization method based on 1 - D projection was proposed in this work. First, vertical windows with 2 main vessel widths were used to compute the vascular scatter degree of every horizontal position, and corresponding horizontal projection curve was plotted, the horizontal location of OD was identified as the location of the minimum position of the horizontal projection curve. Next, a rectangular window slid up to down at determined horizontal location to obtain the brightness and edge gradient features of local regions, and corresponding vertical projection curve was plotted. The maximum peak of the vertical projection curve was set as the vertical position of OD. The proposed method was tested with four publicly-available databases and a self-selection database. The OD was successfully located in 357 images out of 380 images (94%). And the average time for one image in STARE dataset was 13.2 s. Regarding to accuracy and efficiency, the proposed method was easy to implement and present good robust performance.


Li Z.-Q.,Xiangtan University | He Y.,Xiangtan University | Tian Z.-J.,Key Laboratory of Intelligent Computing and Information Processing of Ministry of Education
Journal of Software | Year: 2012

Low efficiency of interference calculation has become the bottleneck that restricts further development of the performance of evolutionary algorithm for the polygon layout. To solve the problem, in this paper, we propose an algorithm of calculating overlapping area between two irregular polygons. For this algorithm, at first, two irregular polygons are respectively decomposed into the minimum number of convex polygons; afterwards, each pair of the overlapping convex polygons from two resulting partitions is clipped and their overlapping area is calculated. Because through a fast non-overlapping test to all pairs of convex polygons to be clipped, invalid computation is decreased; by making use of simple internal vertex judgment and sheartransformation based on intersecting test and intersection calculation between a line segment and any convex broken line segment, its speed of clipping overlapping convex polygons is improved. The time complexity analysis and numerical experiments indicate that the performance of our presented algorithm superiors to the existing algorithms. © 2012 ACADEMY PUBLISHER.


Zhang D.,Xiangtan University | Zhang D.,Key Laboratory of Intelligent Computing and Information Processing of Ministry of Education | Liu X.,Xiangtan University | Song D.,Xiangtan University | And 2 more authors.
Yi Qi Yi Biao Xue Bao/Chinese Journal of Scientific Instrument | Year: 2014

In the production process of mobile phone, scratches, sand residues, stains and other quality problems often appear on the partition of the mobile phone, which affect the assembly of mobile phone and the final product quality. Sand residues appear as round local peak regions, and have omnidirectional symmetric structure characteristic and gray scale Gaussian distribution characteristics. According to these observations, a novel rotating line scan operator is proposed to achieve the feature analysis and detection of the sand residue. Experiments show that the proposed method is superior to traditional top-hat transformation, Gauss template matching and double ring filtering methods. The method is robust to contrast variation, has low false negative rate and false positive rate and presents good adaptability and practicability.


Tang H.,Xiangtan University | Tang H.,Key Laboratory of Intelligent Computing and Information Processing of Ministry of Education | Xu Y.,Xiangtan University | Xu Y.,Key Laboratory of Intelligent Computing and Information Processing of Ministry of Education | And 4 more authors.
Information and Control | Year: 2013

A novel object data correction based diversity neural network ensemble method is proposed by analyzing the object data correction principle. In this method, the individual networks are trained by correcting expectative output of individual network dynamically and set as new training sets, and all the individual networks in the ensemble are guided to realize diversity learning. The method is applied to the fault diagnosis of power transformer, the experiment results show that accuracy rate of the method is superior to the neural network ensemble method which is trained by individual networks, in comparison with ADL (active diverse learning), the communication cost of ensemble network is greatly reduced, it is a extremely efficient neural network ensemble method.

Discover hidden collaborations