KNN-Spectral Regression LDA for Insect Recognition

Insect recognition is the basis of crop pest and disease control. Traditional insect recognition methods are time-consuming and hard-labor. Automatic machine insect recognition can solve the problem. In this paper, spectral regression LDA is used to reduce high dimension spaces of insects images, and get insect feature subspace. Then coefficient vector in the subspace is taken as the input of KNN algorithm. Finally the unrecognized insects are classified and recognized. The accurate recognition rate of KNN-Spectral regression LDA is 90%, which is better than that of PCA and run length matrix.