基于多项扩展式局部方向张量的仿射不变兴趣点检测算子

An affine invariant interest point detector based on polynomial local orientation tensor

针对基于立体视觉的移动机器人导航中摄像机视角及光线亮度的变化会使所获得的场景环境图像发生一定的仿射变换，为了选取稳定的图像局部特征点，提出了一种具有仿射不变的兴趣点检测算法——多项局部方向张量（polynomial local orientation tensors，PLOT）算子。PLOT算子是基于图像多项扩展式局部方向张量的兴趣点检测算法，它将图像的每一像素点以多项扩展式的方式展开建立局部图像信号模型，求其对应的局部方向张量，搜索局部方向张量最小特征值的局部邻域最大值，获得对图像兴趣点的初始检测定位，并应用仿射递归算法实现兴趣点及对应特征区域的最终准确检测与定位。基于重复率准则的兴趣点检测实验表明。

Considering that in vision-based transformation because of the changes mobile robot navigation, the scene images captured may undergo a general affine of camera＇ s visual angle and light brightness, this paper proposes the PLOT （polynomial local orientation tensor） detector, a novel method for detecting scale invariant and rotation interest points to obtain the invariant local features. The detector is based on the local orientation tensor, which is constructed from the polynomial expansion of the image signal. The initial interest points are detected by local maxima search for the small eigenvalues of the orientation tensor. Then an iterative procedure is used to allow the initial points to converge to affine invariant points and regions. The results of the experiment on interest point detection based on the repeatability criteria show the PLOT detector＇s strong performance in different rotations when camera＇s viewpoints and illumination change.