| 摘要: |
| 针对脐橙自动分级检测中存在正确识别率偏低、实时性不强的问题,提出一种综合特征提取方法:在对图像颜色模型进行转换后,用H分量图像提取脐橙的大小特征;S分量图像通过背景分割、边缘灰度补偿、整体亮度变换后提取脐橙的果面缺陷特征;采用R、G、R-G 3个分量的均值和标准差提取脐橙的颜色特征。以脐橙的大小特征、果面缺陷特征和颜色特征为支持向量机(Support vector machine,SVM)的试验输入向量,进行脐橙分级检测试验,以实现提高脐橙自动分级正确识别率和增强实时。试验结果表明:该SVM分类器对脐橙分级的正确识别率为91.5%,处理时间为160 ms,适合于实时环境下的分级检测。 |
| 关键词: 机器视觉 亮度补偿 支持向量机 脐橙 |
| DOI：10.11841/j.issn.1007-4333.2016.03.15 |
| 投稿时间：2015-03-19 |
| 基金项目:国家自然科学基金项目(61273282) |
|
| Online navel orange gradingdetection system based on machine vision |
|
HU Fa-huan1,2, DONG Zeng-wen1, KUANG Yi-shun2
|
| (1.School of Mechanical & Electrical Engineering, Nanchang University, Nanchang 330031, China;2.School of Mechanical & Electrical Engineering, Jiangxi University of Science and Technology, Ganzhou 341000, China) |
| Abstract: |
| Due to the low accuracy of recognition rate and poor performance in real time grading detection of navel orange, a method which extracting comprehensive characteristics was introduced after conversion of image color model.The size of navel orange is extracted by the H component image, the surface defect feature is extracted by background segmentation, edge gray compensation, overall brightness transform of the S component image, and the orange color feature is extracted according to the mean and deviation of R, G, R-G components.Then the size feature, surface defect feature and color feature of navel orange are imputed as vectors to support vector machine (SVM) for grade classification testing to realize high accuracy recognition rate and high real-time grading detection of navel orange.The results show that the correct recognition rate of the classifier is 91.5%, the processing time is 160 ms.proving that the classifier is suitable for grading-detection in real-time environment. |
| Key words: machine vision brightness compensation support vector machine navel orange |
