图像散布图和小波多分辨分析的模具异物检测

为了实现成型生产时模具的自动保护,提出基于散布图模具异物检测方法.采集模具监控图像,建立以灰度为变量的散布图,分析两幅图像对应像素灰度的概率分布及回归关系,实现了检测的光照无关性.基于B样条进行回归线构造,采用散点离差统计直方图获得图像间的匹配置信区间,从而得出异物对应的失配像素集.采用小波多分辨分析的方法,通过对两幅图像分别进行同阶小波分解,获得了消除边缘细节的逼近图像,使得后续的异物检测排除了对比图像之间错位的影响,解决了检测中的图像几何偏差问题.通过应用验证了该方法的良好效果.

Inspection of foreign substances in mould using image scattergrams and multi|resolution analysis

A method for detecting foreign substances in mould based on scattergrams was presented to protect moulds automatically during moulding production. In order to make the algorithm illumination invariant, a gray level scattergram was plotted, and the regression relation between the pair of compared images was derived by analyzing the probability distribution of the corresponding pixel gray in the two images. The regression curve was constructed using B-spline, and the width of confidence interval was obtained by deviation statistical histogram to detect the false matching pixels representing the foreign substance. Multi-resolution analysis(MRA) was used before inspection in order to solve the problem of the geometric deviation between two images. Wavelet decomposition of the two images was conducted to give an approximate image, which eliminated the edge details from the original images. The method was verified by several images．