尺度空域相关性在机械零件形貌检测中的应用

Utilization of Scale-domain Correlation on Edge Detection of Image

为了能准确快速的提取机械零件的边缘信息,根据小波分析多尺度空间的相关性,在模极大值边缘检测的基础上提出一种新的图像边缘检测方法,此种方法不需要对零件的视觉图像进行预处理就能较精确地检测出图像的边缘,在抑制噪声的同时提高边缘的定位精度,克服了直接对图像进行模极大值边缘检测造成的误差。结果表明,使用该方法对边缘进行检测可获得满意的结果。

In order to find the edge information of parts accurately, the paper proposes a new method about image edge detection according to the correlation in the multi-scale space of wavelet analysis, this method can detect the edge of image exactly without preprocessing. Doing the correlation calculation between the adjacent scale wavelet coefficients can improves the position accurately of the edge and reduces the noise at the same time. The new method can also overcome the inaccuracy brought by the method of edge detection with modulus maxima. The experiment results show that this new method is very satisfactory.