Wenming Huang

Algorithm for 3D Point Cloud Denoising

The raw data of point cloud produced by 3D scanning tools contains additive noise from various sources. This paper proposes a method for 3D unorganized point cloud denoising by making full use of the depth information of unorganized points and space analytic geometry theory, applying over-domain average method for 2D image of image denoising theory to 3D point data. The point cloud noises are filtered by using irregular polyhedron based on the limited local neighborhoods. The experiment shows that the proposed method successfully removes noise from point cloud with the features of the scattered point model reserved. Furthermore, the presented algorithm excels in its simplicity both in implementation and operation.

Discussion of Ability Cultivation of Computational Thinking in Course Teaching

Computational thinking is taken for the fundamental skills for everyone, not just computer scientist, to learn and use. It remains to be explored in many aspects how to understand the essence of computational thinking and how cultivate the computational thinking ability of students. In the view of course teaching of computer science, this paper discusses the perception of the essence of computational thinking and presents a method for the cultivation of the computational thinking ability by using courses as carrier, and a case study of the course Data Structures and Algorithms.

2-D maximum entropy spermatozoa image segmentation based on Canny operator

Aiming at the problems of sperm microscopic imaging that is unclear, low-contrast and involves many impurities, a two-dimensional maximum entropy segmentation based on canny operator is presented. First, the initial sperm image segmentation is made according to its features. Next, the real spermatozoa edge is detected by using Canny operator. Finally, the 2-D maximum entropy segmentation is used for the segmentation of sperm image targets and their edges according to different thresholds, thus solving the imprecise segmentation problems because of the local blurred and overlapping spermatozoa image. The experiments show that this spermatozoa image segmentation is of good robustness and high segmentation accuracy.

Trust-Based Ant Colony Optimization for Grid Resource Scheduling

An Ant Colony Optimization (ACO) algorithm is proposed to improve grid resource scheduling, for the performance of grid scheduling affects the applications of grid platform directly. According to the updating rule of local and global pheromone, ACO achieves load balance by incorporating resource-oriented trust mechanism. The evaluation criterion of system safety covers both direct and indirect trust of the resource nodes, thus increasing the reliability and validity of the grid system. The rescheduling mechanism is designed to guarantee the successful task completion at the single node failure, and increase the successful task ratio and fault-tolerant of the grid system.

A real-time medical assistance billing system

To further improve the urban and rural medical assistance system and make it much more convenient in the emergent case, we designed and developed a real-time medical assistance billing system for Linchuan county civil affairs bureau. This system consists of two medical assistance models, one of which is real-time assistance model that realizes a “seamless combination” of the civil affairs assistance system and the two existing systems: the rural cooperative medical care system and hospital billing system. It provides the direct settlement for the medical assistance at hospital, thus solving the problem that some poor people have no money to pay the bill for the medical treatment. This paper begins with the introduction of the design guideline of the medical assistance system, followed by the design of the system function modules. Furthermore, it presents the detailed solution to some key technology in the system, including system deployment, data security and data consistency problem of the civil affairs bureau database and the hospital database and so on.

Tensor voting based method for image denoising

The computation of voting field is the key computation step of image de-noising based on tensor voting that is a highly robust feature extraction method. As the traditional tensor voting field is approximately elliptic and symmetry, there will be redundant computation resulted during the course of tensor voting superposition. The voting field of the images of symmetry property can be simplified to half of the original traditional field. On the basis of this simplification idea, this paper presents an image de-noising method based on voting field, with a comparison of its result with those of median filter de-noising and of wavelet de-noising. The experiment shows that the method presented in this paper can efficiently reduce the image noise, with the key details of the image preserved.

A Trust Rating Model of Recommendation in Grid

The trust rating of recommendation for one entity greatly depends on the trust in the recommenders’ ability to provide recommendations. To prevent the cahoots, cheating and fluctuant fraudulence in recommendation, we studied the characteristics of the two classes of true and false recommendation. On the basis of the analysis of the feature set, a distinguishing model is created to effectively identify the baleful entities that provide false recommendations in grid system, thus increasing the veracity of trust rating. The model checking shows that this scheme is of high feasibility.

Exploration of ability development of engineering and computational thinking skills in software engineering majors

Software talents should have the engineering ability, a common sense in IT industry, which is characterized by the fact that design plays a very important role in software engineering. Basis on the enhancement of computational thinking, the reform ideas of the development of the students engineering ability are introduced. The detailed approaches include: 1) Extraction of fundamental discipline concept of engineering ability development; 2) Problem solving-centered organization of courses of software development tools; 3) Initiate the courses of software engineering as soon as possible; 4) Continual training of abstract logical thinking for the purpose of software abstract thinking.

Research on dynamic recommendation trust based on GC-TOPSIS in grid

In order to effectively reduce the impact of malicious recommendation, this paper designs a method of dynamic recommendation trust based on the gray correlation technique for order preference by similarity to ideal solution (GCTOPSIS). First, we determine the combined weights of each node attribute by using AHP and entropy method. Next, we use the gray correlation analysis to calculate the gray correlation relative closeness degree (GCRCD) of the recommendation nodes and the reference nodes, and sort all the recommendation nodes according to the GCRCD. Finally, according to the importance of the interaction between the nodes, we can dynamically select the recommendation nodes and obtain the set of the trusted recommendation nodes. The experiment shows the method can filter out the malicious nodes, thus effectively reducing the impact of malicious recommendation.

Surface reconstruction from scattered point cloud based on RBF and POU

A method for surface reconstruction from scattered point cloud based on the combination of RBF and POU is proposed. First, point cloud is decomposed into a set of slightly overlapping subdomains in the form of binary tree. Next, local reconstruction functions are evaluated by using radial basis functions in each subdomain. Then, the global reconstruction function is obtained by weighted blending of these local reconstruction functions. Finally, the implicit surface is visualized by using Bloomenthal method. The experiments prove that the surfaces resulted from the proposed method are continuous and smooth.