Mobile plant species classification: A low computational aproach

In this paper a reduced shape and color feature extraction method is proposed for a mobile device based plant classification system. For scientists, botanists, farmers, and others plant identification is a useful and important task. The original image captured is reduced to similar aspect ratio which does not affect the shape information but reduces the computation cost nearly up to half of the total cost. The algorithm first calculates the geometric feature and then polar Fourier transform and trained using k-NN classifier. Then two nearest classes were selected on the basics of smallest distance which is further rectified by the color features using a decision tree. The algorithm proves to be better in performance compared to other already existing algorithms.