Application of Naive Bayesian Classifier Based on Gaussian Distribution in Plant Recognition

摘要

This paper presented a kind of plant leaf classification method based on naive Bayesian classifier and leaf shape features. Firstly, estimate the class-conditional probability of the numeric features with the distribution function of Gaussian distribution, then reduce the deviation of probability statistics of the nominal features with Laplace smoothing. In the end, compare and analyse the influence of equal-width binning, equal-frequency binning, MDL discretization method and Gaussian probability estimation method on the classification efficiency. The experimental results on iris data set indicated that the correct classification rate of this method reached 95.3%, and the method was feasible and effective.