摘要
研究一种基于视觉词袋模型的图像筛选与搜索优化算法以提高机器人闭环检测质量和降低图像信息处理量.首先,通过SURF算子提取图像中的特征信息,构建对应的视觉词袋模型,并形成视觉词袋直方图;其次,基于视觉词袋直方图计算获得对应的图像混合显著度,进而筛选出信息量丰富且可区分度大的图像,并组成待搜索图像集合;然后,从视觉词袋直方图中提取图像中的显著主要特征类组成集合,并用其近似替代图像的特征分布情况,以降低图像特征信息处理量,加快图像搜索速度.最后,仿真实验证明本文提出的图像筛选和搜索方法的可行性和有效性.
An optimized method for image selection was proposed to improve the quality of robotic close-loop detection and to decrease the data volume of image information processing.Firstly,SURF algorithm was employed for the feature extraction,which could be used to set up a visual bag-of-word model and to form its histogram.On this basis,a mixed image saliency was achieved,and then the images with diverse features and obvious discriminations were picked up to construct the set of images which was pending to be searched.Then the main remarkable feature classes were extracted from the visual bag-of-word histogram,which were used to represent the information distribution of the image. So the image feature information to be processed was reduced and the speed in the image search was increased.Finally,several experiments were implemented to verify the feasibility and validity of the proposed optimized methods.
出处
《华中科技大学学报（自然科学版）》
EI
CAS
CSCD
北大核心
2015年第S1期233-236,共4页
Journal of Huazhong University of Science and Technology(Natural Science Edition)
基金
国家自然科学基金资助项目(51575034)
机器人技术与系统国家重点实验室开放课题(SKLRS-2013-ZD-03)
计算机辅助设计与图形学国家重点实验室开放课题(A1516)
