scispace - formally typeset
Search or ask a question
Author

Hong Tian

Bio: Hong Tian is an academic researcher from Dalian Jiaotong University. The author has contributed to research in topics: Ontology Inference Layer & Process ontology. The author has an hindex of 1, co-authored 2 publications receiving 16 citations.

Papers
More filters
Journal ArticleDOI
TL;DR: A novel multiplex classifier model, which is composed of two multiplex cascades parts: Haar-like cascade classifier and shapelet cascade classifiers, which filters out most of irrelevant image background and detects intensively head-shoulder features.

18 citations

Proceedings ArticleDOI
22 Nov 2010
TL;DR: Experimental results illustrate that this method for choosing the set of candidate similar concepts is effective for computing the concept similarity and ontology can be transferred successfully to learn.
Abstract: Recently, more and more research is devoted for ontology in the semantic web domain. Firstly, a method for choosing the set of candidate similar concepts is presented based on ontology graphical structural features and data mining. Secondly, a calculation method of conceptual similarity is proposed based on the characteristics of the concept ontology and information content. Finally, the optimized ontology can be transferred into learning. Experimental results illustrate that this method is effective for computing the concept similarity and ontology can be transferred successfully to learn.

Cited by
More filters
Journal ArticleDOI
TL;DR: A hierarchical inspection framework containing bearing cap (BC) detection, fault region localization, and BBK classification is proposed, and a BBK classifier based on the GCCM features and a support vector machine is used to process the fault region to identify the missing of BBK.
Abstract: With the development of hardware and software in camera and computing units, visual inspection system (VIS) plays an increasing significant role in fault inspection task. This paper proposes a VIS to inspect the missing of bogie block key (BBK) used on freight trains. BBK is an important component to keep wheel sets from separating out of bogies. The missing of BBK is one of the most common faults threatening the running safety. VIS first acquires a bogie image by the image acquisition system, and then a hierarchical inspection framework containing bearing cap (BC) detection, fault region localization, and BBK classification is proposed. Specifically, a cascaded detector trained by the AdaBoost approach combined with the gradient coded co-occurrence matrix (GCCM) features is used to achieve fast and accurate BC detection. Then, a fault region is located on the basis of the relationship between the BC and BBK. Finally, a BBK classifier based on the GCCM features and a support vector machine is used to process the fault region to identify the missing of BBK. The proposed system has been applied on a long sequence of real images showing high inspection speed and accuracy.

66 citations

Journal ArticleDOI
TL;DR: The progressive subspace ensemble learning approach (PSEL) which takes into account the data sample space and the feature space at the same time and outperforms a number of state-of-the-art classifier ensemble approaches.

38 citations

Journal ArticleDOI
Chao Mi1, Zhiwei Zhang1, He Xin1, Youfang Huang1, Weijian Mi1 
TL;DR: The experimental results in Tianjin port show that the two-stage classifier can improve the classification accuracy of human detection obviously.
Abstract: Abstract With the development of automation in ports, the video surveillance systems with automated human detection begun to be applied in open-air handling operation areas for safety and security. The accuracy of traditional human detection based on the video camera is not high enough to meet the requirements of operation surveillance. One of the key reasons is that Histograms of Oriented Gradients (HOG) features of the human body will show great different between front & back standing (F&B) and side standing (Side) human body. Therefore, the final training for classifier will only gain a few useful specific features which have contribution to classification and are insufficient to support effective classification, while using the HOG features directly extracted by the samples from different human postures. This paper proposes a two-stage classification method to improve the accuracy of human detection. In the first stage, during preprocessing classification, images is mainly divided into possible F&B human body and not F&B human body, and then they were put into the second-stage classification among side human and non-human recognition. The experimental results in Tianjin port show that the two-stage classifier can improve the classification accuracy of human detection obviously.

23 citations

Journal ArticleDOI
TL;DR: This work proposes a learning strategy aimed at maximizing the node classifiers ranking capability rather than their accuracy, and provides an efficient implementation yielding the same time complexity of the original Viola-Jones cascade training.

22 citations

Journal ArticleDOI
TL;DR: It is argued that 2D image models greatly reduce the time cost of scene segmentation with a little loss of accuracy and the usage of 2Dimage models is not limited inscene segmentation since robust features can be extracted from 2D picture models to accomplish laser point classification and scene understanding.

8 citations