基于彩色信息的尺度不变特征变换图像特征点提取与匹配

针对将彩色图像转化为灰度图像后再进行特征点提取与匹配会丢失彩色信息，可能导致误匹配这一问题，采用一种基于彩色信息的SIFT特征点提取及匹配算法(CSIFT)以实现彩色图像的特征点提取及匹配，结合目标的颜色特征与几何特征，以颜色不变量作为输入图像,再提取特征点并描述特征点周围的信息,通过最近邻匹配法求出图像间的匹配对。实验将该算法应用于视觉里程计中，对相机所拍的相邻两帧图像进行了特征点提取及匹配，并比较了该算法与传统SIFT算法的差异。实验结果表明该算法是有效的。

Image feature extraction and matching of color-based scale-invariant feature transform

That image feature extraction and matching will be done after the color image is transformed into gray one may cause color information loss and lead to wrong matching. In order to solve such problem, an image feature extraction and matching method named CSIFT (Color-based Scale-Invariant Feature Transform) was used to realize color image feature extraction and matching. Combining the color features with the geometrical features of the objects, feature points were extracted and neighbor information around these points was described using color invariant value as input information. Then the points between two images were matched using the nearest neighbor method. The algorithms were applied to vision-odometer, feature extraction and matching with two adjacent frames from camera were operated and compared with the SIFT algorithm in the experiment. The result shows that the algorithm is effective.