normal rice seed as well as small rice seeds and large rice seeds are found. Computed threshold value in pixels of rice seeds is shown in Table III.
Particle classification: Classification identifies an unknown sample by comparing a set of its significant features to a set of features are compared that conceptually represent classes of known samples [14]. A particle classifier uses feature vectors to identify samples based on their shape. A color classifier uses color features to identify samples based on their color. Classification involves two phases: training and classifying. Training is a phase during which one teaches the machine vision software the types of samples one want to classify during the classifying phase. Training can be done on any number of samples to create a set of classes, which one later compare to unknown samples during the classifying phase. The classes are stored in a classifier file. Training might be a one-time process, or it might be an incremental process one repeat to add new samples to existing classes or to create several classes, thus broadening the scope of samples one want to classify. Classifying is a phase during which one’s custom machine vision application classifies an unknown sample in an inspection image into one of the classes one trained. The classifying phase classifies a sample according to how similar the sample features are to the same features of the trained samples. The need to classify is common in many machine vision applications. In this paper particle classification has been used three times to classify broken rice seeds, small rice seeds, normal rice seeds and large rice seeds. Fig. 5 shows particle classification step.