ABSTRACT
To verify the performance of crop classification and variety clustering based on electrophoretogram and investigate vegetable seed genetic relationships, three kinds of breeder seeds’ samples are collected. They are bell pepper, Chinese cabbage, and cucumber. Thirty varieties of each kind of crop have been collected. Standard electrophoretograms of them are prepared by the method of protein ultrathin isoelectric focusing electrophoresis. Then, the digital images of them are obtained by scanners. Using these images, a pattern recognition model of crop recognition based on principal component analysis (PCA) and support vector machine (SVM) is constructed. Test results obtained by the leaving-one method show that, more than 97% crops can be recognized correctly. In addition, through k-means clustering analysis, clustering trees of these three kinds of crops with different varieties are established and the paternities of partial varieties based on clustering tree are discussed. This study has positive significance to automatic seed inspection by computer based on electrophoresis and selection of breeding direction.
