各向异性光流法的目标边缘检测

光流法是在图像序列中对二维物体的运动矢量估计,用经典的HS算法计算时,目标与背景对比度小或有噪声将导致较大的虚警概率。而运用各向异性高斯核,运用非极大抑制条件,可以较好的抑制噪声,同时也保留了目标边缘,进而提高了梯度偏导的准确度。较好地解决了微分光流法的不足,从而提高了计算精度。最终对序列图像中动目标的边缘检测进行了方法改进,Gabor函数的替换以及实验表明算法的有效性。

Target Edge Detection of Optical Flow via Anisotropic Gaussian Kernel

Optical flow method is to estimate the movement of 2 d objects in image sequences.With classic HS algorithm,the smaller contrast between target and background or higher noise will lead to larger false-alarm probability.The use of anisotropic gaussian kernel,with the non-maximum suppression,can better suppress noise while retaining the target edge,and thus improving the accuracy of the gradient partial derivative.It is a better solution for the lack of differential optical flow method and improves the calculation accuracy.Finally,the edge detection method of moving targets in the sequence of images is improved,and the effectiveness of the algorithm is proved through the replacement of the Gabor function and experiments.