基于ICM的局部不变特征提取方法

A Novel Method based on ICM for Local Invariant Feature Extraction

提出了一种新的基于交叉皮层模型(intersecting cortical model,ICM)的局部不变特征提取方法。首先将k-means聚类后的Harris角点作为关键点,接着以关键点为中心构造特征区域,最后采用ICM图标信号对特征区域进行描述作为局部不变特征。实验结果表明:该方法所获取的局部不变特征具有很好的可重复性,并且当目标处于复杂背景以及遮挡条件下具有较好的识别性能。

This paper presents a novel method based on ICM （intersecting cortical model） for local invariant feature extraction. Firstly, the key points were extraceted by Harris operator and they were clustered by k-means algorithm. And the clustered points were called as key points. Then, the key points were used as center to extract their feature regions. Lastly, the ICM values of these feature regions are calculated as local invariant features. Finally, the experimental results showed that our method has good performance in stability to the repeatability and good recognition performance in the condition of complex background and partially occluded.