Rice blast recognition based on principal component analysis and neural network

Based on principal component analysis and back propagation neural network (PCA-BP), a rice blast recognition method was proposed to solve the problems of low accuracy, inefficiency and subjectivity of artificial recognition of rice blast. First, image of harvested lesion was processed, with 6 color features, 10 morphological features, and 5 texture features of each lesion were extracted. Secondly, stepwise regression analysis was used to analyze the correlation between the characteristic parameters. The results showed a linear correlation. Then, the principal component analysis (PCA) method was used to reduce the dimension linearly, to map 21 features into 6 comprehensive features as input parameters. Finally a 6-11-4, 3-layer back propagation (BP) neural network identification model was constructed for the classification and recognition of the lesion. The experimental results show that the average recognition rate of rice blast based on principal component analysis and BP neural network is 95.83%, which is 7.5% higher than the average recognition rate using BP neural network and 2.5% higher than the existing SVM method with high accuracy in identifying rice blast. It can identify rice blast quickly and effectively.