摘要
机器视觉在苹果质量自动检测中应用,由于苹果梗蒂和缺陷在颜色和形状上相似,容易混淆。梗蒂的识别是检测中的难点,极大的影响苹果质量等级识别中的识别效率。提出了一种基于纹理分析的多种分类器识别比较的方法,采用经典的阈值分割算法实现类似梗蒂的目标区域分割,提取纹理、统计特征,通过SFFS特征选择,然后采用多种类型的分类器进行梗蒂识别。经过多种分类器实验结果比较,发现支撑向量具有比较好的识别效果,梗和蒂的识别率达到了95%和96%,只有13%的缺陷被误分为梗蒂。
During the inspection of the apple,stem - calyx concavities of apples are easily confused with blemishs. Apple Stem - Calyx Identification is very important for quality grading of apples.In this thesis,a new method based on texture analysis and multi - classification is proposed to identify blemish and stem of Fuji apples.The designed and implemented system is based on segmentation of stem - calyxes and blemishes from fruit images,feature extraction/selection and classification steps are taken to recognize apple stem - calyx and blemishes.In comparison with multi - classification mechanism,we found that the classification based on support vector machines(SVM) used in recognizing stem -calyx resulted in a better effect than others.In stem and calyx databases 95%and 96%of SC are correctly recognized,and only 13%of the blemishes were classified as SC.
出处
《计算机仿真》
CSCD
北大核心
2010年第7期210-213,共4页
Computer Simulation
基金
浙江省教育厅科研项目(Y200804724)
关键词
机器视觉
图像处理
特征选择
分类器
模式识别
Machine vision
Image processing
Feature selection
Classification
Pattern recognition