基于机器视觉的大米垩白米的检测方法

Detection Method of Chalk Rice Based on Machine Vision

通过分析2种米粒的不同的直方图分布表现来区分普通米与垩白米,并进而对检测出来的垩白米粒采用改进的最大类间方差法来进一步分析其垩白率、垩白度等信息。试验采用了12组样本,每组样本60粒,按照不同的放置方向及位置,各采集10幅图像,共120幅图像。对所有图像进行检测,检测结果表明本研究提出的方法能快速,准确地区分垩白米粒及正常米粒,并且可以进一步较准确地检测垩白米粒的垩白度,相较于传统的OTSU算法,垩白米粒的检测准确率平均提高了8%左右,垩白度的检测准确率平均提高了6%左右。

In this paper,a new detection algorithm is proposed by analyzing the histogram character of the two kinds of rice to get the distinguish information. Furthermore,we can then get the information of chalky grain rate and chalk degree by the improved OTSU algorithm. The experiment uses 12 samples,and each sample includes 60 grains of rice. Each grain of rice is set by different angle and position,and 10 photos are taken for each sample（ totally 120 images of different rice samples）. Experiment result indicates that our algorithm can detect chalk rice more quickly and efficiently,and furthermore the degree of chalkiness also can be calculated more correctly. Compared to traditional OTSU algorithm,the accuracy of our algorithm in detecting rate of chalk rice is 8% higher than OTSU on average,and the accuracy of our algorithm in detecting degree of chalkiness is 6% higher than OTSU on average.