基于颜色与缺陷检测的椪柑分级算法

以湘西椪柑为研究对象，提出了一种基于颜色与缺陷检测的椪柑自动分级算法.首先，对椪柑灰度图像进行阈值分割和孔洞填充得到二值图像；然后，将二值图像与椪柑彩色图像的R，G，B分量分别进行与运算，并将运算后的3幅单色图像进行合成，从而得到彩色图像的椪柑目标区域；接着，提取目标区域的黄色像素占比和缺陷面积作为色泽特征参数和缺陷特征参数；最后，利用决策树模型融合特征参数以进行椪柑分级判定.实验结果表明，对比基于单一特征的椪柑分级算法，新算法通过特征的互补提高了椪柑分级的准确率.

Ponkan Grading Algorithm Based on Color and Defect Detection

Taking Ponkan in Xiangxi as the research object,aponkan an grading algorithm based on color and defect detection is proposed.First,threshold segmentation and hole filling are used on the Ponkan gray image to obtain the corresponding binary image.Then,the AND operation is respectively performed between the binary image and the R,G,and B components of the Ponkan color image.After the AND operation,three monochrome images are obtained,which are combined to obtain the target area of the Ponkan color image.For the target area,the proportion of yellow pixels and the defect size are respectively extracted as the color feature parameter and defect feature parameter.Finally,decision tree model is adopted to fuse the feature parameters for grading the Ponkan.The experimental results show that,compared with grading algorithm based on a single feature,the proposed algorithm based on multiple features complementany can achieve higher accuracy.