Identification Rice Varieties Based on K-means Clustering Algorithm and BP Neural Network

Abstract

BP neural network is characterized by adaptability and real-time learning, so it is used widely in classification. While the more complex samples classify, the lower the accuracy of BP neural network is. So a method was purposed to identify rice varieties that combined K-means clustering algorithm with BP neural network. The gray-scale mean value, aspect ratio and circularity, which were the three parameters that expressed rice, were extracted by image processing. The K-means clustering algorithm was used to classify the data based on the aforementioned three parameters and the classification result was entered into the BP neural network and trained to get the classifier. The overall results indicate that the method mentioned is more effective than using K-means clustering algorithm or BP neural network singly and the accuracy is up to 80%. Experimental results show that combination of K-means clustering algorithm and BP neural network is feasible for identifying rice varieties.