基于多峰分布最大类间方差的垩白米粒检测方法

采用一种基于多峰分布最大类间方差(Otsu)方法检测垩白米粒。首先利用垩白米粒图像的灰度分布具有多峰分布这一特性,应用数理统计方法进行多峰分布度测度。然后根据垩白米粒图像中垩白部分具有白色不透明这一物理特征将垩白米粒图像的灰度级分成3类,通过使用最大类间方差法确定两个分割阈值,并且检测出大米中垩白部分面积。由于使用了区域控制分割,使得最大类间方差阈值的确定具有自适应性和准确性特点。实验结果表明,本文采用的方法对垩白米粒检测正确度为97.4%,效果好于基于形态学分水岭的检测算法。

Detection of chalky rice based on multi-peak distribution otsu algorithm

A detection method of chalky rice based on multi-peak distribution otsu algorithm is proposed.Firstly,the multi-peak distribution of the chalky rice gray imges is measured by mathematical and physical statistics.Then the gray levels of the chalky rice image are classified into three parts according to the chalky kernel features.Two segmentation thresholds are determined by the otsu algorithm,and the area of the chalky region is detected.Because the method of otsu can exactly segment different regions by using the gray-level histogram algorithm,the threshold of the chalky region can be adaptive.The experimental results indicate that the detection accuracy of the proposed method is 97.4%,which is better than the morphological watershed method.