基于计算机视觉的大米外观品质检测  

Rice Outer-quality Inspection Based on Computer Vision

作　　者：吴彦红[1] 刘木华[1] 杨君[1] 郑华东[1] 

机构地区：[1]江西农业大学工学院

出　　处：《农业机械学报》2007年第7期107-111,共5页Transactions of the Chinese Society for Agricultural Machinery

基　　金：江西省自然科学基金资助项目(项目编号:0230019)

摘　　要：开发了一套基于计算机视觉技术的稻谷品质检测系统,采用灰度变换、自动阈值分割、区域标记等方法从采集的稻米群体图像中提取单体米粒图像,对单体米粒的裂纹、垩白特征进行了统计和检测方法研究。提取了米粒的面积、周长等10个特征参数作为整精米检测特征,并进行了主成分分析,确定了判别整精米的优化阈值。检测试验结果表明:裂纹米粒识别的准确率为96.41%;垩白米粒识别的准确率为94.79%;整精米识别的准确率为96.20%。The quality of rice is the main factor that affects the market price of rice. Today, the detecting and grading of rice are mainly carried out by manual ways, which is time-consuming and toilful, and even easily leads to improper judgment. A detecting system of rice quality based on computer vision was developed in this paper. The methods that segmenting single kernel from mass rice image using gray transformation, automatic threshold segmentation, and region marking were discussed. In order to detect the head rice ratio, ten parameters were selected from the profile of rice kernels, such as the area and perimeter of rice kernel, the two axes of the equivalent oval, the inspection of the profile of rice kernel and head rice rate were discussed after using the principal components of the profile parameters of rice kernel. The results of detecting experiments on five varieties of rice indicated that the accurate ratio of detecting fissure is about 96.41%, the accurate ratio of chalkiness detecting is about 94.79%, and the correct ratio of detecting head rice is about 96.20%.

关 键 词：大米 品质 检测 计算机视觉 

分 类 号：S511[农业科学—作物学] TP391.4[自动化与计算机技术—计算机应用技术]