文献类型：会议论文

英文题名：Low Level Saliency Feature Extraction Method based on Hessian Threshold

作者：Du, Mingfang[1];Wang, Junzheng[1];Li, Jing[1];Cao, Haiqing[1];Cui, Guangtao[1];Fang, Jianjun[2];Lv, Ji[2];Pu, Jiantao[2]

第一作者：Du, Mingfang

通讯作者：Du, MF[1]

机构：[1]Beijing Inst Technol, Minist Educ, Key Lab Complex Syst Intelligent Control & Decis, Beijing 100081, Peoples R China;[2]Beijing Union Univ, Automat Sch, Beijing, Peoples R China

第一机构：Beijing Inst Technol, Minist Educ, Key Lab Complex Syst Intelligent Control & Decis, Beijing 100081, Peoples R China

通讯机构：[1]corresponding author), Beijing Inst Technol, Minist Educ, Key Lab Complex Syst Intelligent Control & Decis, Beijing 100081, Peoples R China.

会议论文集：Chinese Automation Congress (CAC)

会议日期：NOV 07-08, 2013

会议地点：Changsha, PEOPLES R CHINA

语种：英文

外文关键词：Visual saliency feature; Hessian threshold; SURF; Robot visual recognition

摘要：The local invariant feature extraction algorithm SRUF (Speeded Up Robust Features) is introduced firstly. Then the new method of finding low level visual saliency feature based on SURF is deduced. The new method pay attention to Hessian matrix threshold and extract image features through changing the Hessian threshold. The number of saliency feature points change with the change of Hessian threshold. The visual saliency feature points will become sparser when Hessian threshold becomes larger. When some certain extreme thresholds which are defined as Hessian threshold Nodes are reached, the retained feature points are remarkable discriminative and stable feature points which make up the best sparse saliency features set. The feature extraction, matching and object recognition experiments of robot vision are finished to verify the new method. Experiment results show that the method is very effective.