高级检索

    基于特征比的平面目标识别

    Planar Object Recognition Based on Characteristic Ratio

    • 摘要: 模式识别是人工智能研究领域的一项重要课题,对于目标物在射影变换、仿射变换下的识别,尤其是严重变形情况下的识别和匹配更是该领域的研究热点和难点.针对仿射变换下的平面目标识别问题,提出了一种新的几何特征不变量——特征比,并以此为基础构造了一种新的仿射不变图像特征描述符.该描述符通过构造一系列与目标图像相交的直线,将图像用一系列共线点的位置关系进行表示.进而,将点的位置关系转化为特征比并表示成一系列的特征比谱;最后,通过动态时间规整(dynamic time warping, DTW)算法比较特征比谱间的距离得出图像间的相关性,从而进行目标识别.实验表明,该算法不仅对严重的仿射变形有较高的识别率,对相似度较高的图像也有很好的区分效果.

       

      Abstract: Pattern recognition is an important research field in artificial intelligence. The stability of recognition under projection and affine transformation, especially under sever deformation has long been recognized as an important and difficult problem. In this paper, a novel geometry invariant named characteristic ratio(CHR) resistant to affine deformation is proposed. A new feature descriptor is also constructed based on this, which presents an image as the relationship of collinear points by a series of straight lines cross the image. Then, characteristic ratio is calculated by the position of collinear points, and the image is represented by a sequence of characteristic ratio spectra(CHRS). Finally, dynamic time warping (DTW) algorithm is employed to compare the similarity of spectra and to find the pixel level correspondence between two planar objects. It shows that the proposed method has not only good resistance to severe affine deformation, but also good discriminating ability to similar images.

       

    /

    返回文章
    返回