Abstract: A new image feature and variety method of maize seed recognition based on Support Vector Machine（SVM） and Genetic Algorithm（GA） is proposed.Firstly，this method optimizes the image feature of maize seed by genetic algorithm.Then，it uses multi-class SVM—decision binary tree to recognize the maize seed.This classification method distributes classifier to each node to constitute multi-class SVM，which reduces the number of SVM classifier and duplicates training samples.The experiment shows that this method can select the suitable maize seed image feature and recognize precisely the maize seeds.