基于颜色和SIFT特征的视频目标检索方法研究

Video Object Retrieval Method Based on Color and SIFT Features

提出并实现了一种结合颜色和SIFT特征的视频目标检索方法。首先,利用图像颜色直方图特征对目标图像进行初步检索和排序;然后对检索结果中与目标图像相似度较高的图像进行SIFT特征提取,并与目标图像中截取的目标区域子图的SIFT特征进行匹配;最后根据SIFT特征匹配结果对初步检索图像再进行二次排序,并反馈目标位置和偏移信息。仿真实验结果表明提出的基于颜色和SIFT特征的视频目标检索方法对目标亮度、位置平移、缩放、形态变化等有较好的适应性,能准确检索出目标,具有较高的实用性。

A method of video object retrieval which is combined color and SIFT features is proposed and achieved.Firstly,color histogram is used to search the target and sort the results of preliminary retrieval.Secondly,the SIFT features can be extracted from the retrieval results which is similar to the reference images.They can be matched with the SIFT features in the object area which is cut out from the reference image.Finally,the preliminary retrieval results can be sorted again according to the matching results of SIFT features.The object location and migration information returns to the control interface as feedback.The experimental results show that this method has good adaptability of object brightness changes,position translation,scaling and shape changes.By using this method the object can be retrieved accurately and it has high practicability.