期刊:
Data Science Journal,2009年8:52-61 ISSN:1683-1470
通讯作者:
Ouyang, C.(ouyangcp@gmail.com)
作者机构:
[Zhang, Xiaoming; Ouyang, Chunping; Hu, Changjun; Zhao, Chongchong] School of Information Engineering, University of Science and Technology Beijing, No.30 Xueyuan Road, Haidian District, Beijing 100083, China;[Wu, Jinbin] School of Materials Science and Engineering, University of Science and Technology Beijing, No.30 Xueyuan Road, Haidian District, Beijing 100083, China;[Ouyang, Chunping] School of Computer Science and Technology, South-China University, Hengyang 421001, China
通讯机构:
[Ouyang, C.] S;School of Information Engineering, University of Science and Technology Beijing, No.30 Xueyuan Road, China
摘要:
The explosion of non-structured materials science data makes it urgent for materials researchers to resolve the problem of how to effectively share this information. Materials science image data is an important class of non-structured data. This paper proposes a semantic annotation method to resolve the problem of materials science image data sharing. This method is implemented by a four-layer architecture, which includes ontology building, semantic annotation, reasoning service, and application. We take metallographic image data as an example and build a metallographic image OWL-ontology. Users can accomplish semantic annotation of metallographic image according to the ontology. Reasoning service is provided in a data sharing application to demonstrate the effective sharing of materials science image data through adding semantic annotation.
摘要:
This paper presents a new approach based on maximum entropy and maximum entropy Markov model for web information extraction. This approach is not only able to overcome the shortcoming of the less precision and recall of the hidden Markov model. In addition, this approach can make the most of various kinds of contextual information from web. The experiments are found that the hybrid approach has an average precision rate of 87.516% while the hidden Markov model trained by the Baum-Welch algorithm has an average precision rate of 68.630%. This implies that the hybrid approach is more optimized than the hidden Markov model trained by the Baum-Welch algorithm.
摘要:
Motion estimation and motion compensation are included as major technologies into the existing video coding standards, which generate the motion vectors that determine how each motion compensated prediction frame is created from the reference frame. As traditional motion vectors are 2-dimensional and translational, they could not represent the actual moving directions of objects or blocks in frames. This paper proposes the special idea of GTMV (global translational motion vector), and then presents us the nearly Full Search Motion Estimation and Motion Compensation algorithms based on GTMV. Experimental results based on the H.264/AVC reference software of JM10.1 show that the proposed algorithms could get a better video compression performance and could improve the coding efficiency despite increasing some coding complexity.
期刊:
The Journal of Information and Computational Science,2007年4(3):1035-1043 ISSN:1548-7741
通讯作者:
Tan, M.(tanlntms@tom.com)
作者机构:
[Tang, Liang; Yu, Hongxiao; Zhao, Zhiguo; Hu, Yixiang; Tan, Minsheng] School of Computer Science and Technology, University of South China, Hengyang 421001, China;[Wan, Yaping] School of Computer Science and Technology, Huazhong University of Science and Technology, Wuhan 430074, China
通讯机构:
School of Computer Science and Technology, University of South China, China
摘要:
Most of spam filters work on single computer, so they can only obtain limited information and affect is not satisfying. Furthermore, many spam makers apply the new technologies and approaches to make and send spam constantly, which is challenging the traditional anti-spam technique based on contents. In this paper, we propose Identifying Technology of Spam Sending Behavior Based on P2P (ITSSB). According to bulk sending characteristics of spam and the advantages of mixed P2P network, an anti-Spam P2P network on the basis of super-peer is designed. Then, we put forward the algorithm for email conversation information examination, the algorithm for rvp query and the algorithm for email sending behavior judging. Finally, on the spam filtering capability, we compare ITSSB with Foxmail based on Bayesian. The experiment results show that: instead of depending on the email content, the language type and the format analysis, ITSSB, based on the characteristics of bulk email sending mails, works better than Foxmail in the recall, precision, F1, accuracy of filtering spam.