基于机器视觉的马铃薯质量和形状分选方法

马铃薯的质量和形状是机器视觉分级的2个重要特征和依据,为实现马铃薯质量与形状检测分级,该文提出了一种基于图像综合特征参数的分选方法。首先提取马铃薯俯视图的面积参数和侧视图的周长参数,通过回归分析建立马铃薯的质量检测模型,实现对马铃薯的质量分选;然后提取马铃薯俯视图像的6个不变矩参数,输入到已训练好的神经网络,完成对马铃薯形状分选。试验结果表明:该方法可以有效的检测马铃薯的质量并区分其形状,质量分选准确率为95.3%,薯形分选准确率为96%。可满足实际应用的要求。

Potato grading method of mass and shapes based on machine vision

In the machine vision technique of potato grading process, mass and shape are two important characteristics. In order to achieve potato grading with these two factors, a grading method of mass and shapes based on image characteristic parameters was put forward in this paper. After extracting parameters of top view area and side view perimeter, a potato mass grading model was constructed with stepwise regression analysis, and then four rounds of mass classification were completed with machine vision. To implement potato shape classification, six invariant moment parameters of vertical view were input the trained neural network. The potato grading experimental results showed that the precision ratio of potato mass grading was 95.3%, and the accuracy of potato shape grading was 96%. Therefore the potato grading results indicate that this kind of classification method can detect different mass of potatoes and distinguish 3 classes of potato shapes effectively, which meet the practical application requirement.