基于贝叶斯决策的果树害虫图像分割研究

Study on the Image Segmentation of Fruit Tree Pests Based on Bayes Decision

由于果树害虫图像中会存在背景区域,而用户关注的是图像中的害虫,需要把害虫从背景区域中分割出来。针对这种情况,提出基于贝叶斯决策的果树害虫图像分割方法。算法首先将图像中的像素分为前景和背景两类,获取两类的先验概率,取得图像中蓝色通道的灰度图,统计每个灰度级中像素出现的个数,计算出正态分布函数的参数,得到类条件概率密度函数。利用贝叶斯公式计算后验概率从而确定每个像素在两类中的归属。经过试验,可以有效的把前景区域分割出来。

Due to the existence of background region in the fruit tree pest image,it needs to split the regions of pests from the background image while users are really interested in the pest region.In view of this,this paper proposed an algorithm of image segmentation for fruit tree pests based on Bayes Decision.This algorithm divided the pixels in the image into two categories first,including foreground and background.Then it got the prior probabilities of two categories and the blue channel of the image,and counted the frequency of the gray value of each pixel,and then calculated the parameters of normal distribution function to obtain the class conditional probability density function.At last,it calculated the posterior probabilities by using Bayes formula to determine the ownership of each pixel in the two categories.The satisfied results were gotten after experiment.