摘要
研究玉米病斑识别问题,提高识别准确率。人们知道玉米可能会同时出现多种病斑,当多种病斑重叠生长时病斑颜色将发生改变,传统的高斯模型识别算法只根据图像颜色特征进行识别,无法识别重叠病斑,造成识别准确率不高。为此提出一种模糊识别技术,首先根据学习集获取病斑的中心,基于同一种病斑纹理特征向量不突变的规律,引入模糊均值对图像特征向量遍历,得到病斑连通图即完成病斑的识别,避免了只对颜色特征的依赖。实验证明,这种方法能够使玉米重叠病斑准确识别,取得了满意的效果。
Research the problem of corn disease spot identify to improve identification accuracy. Corn may appear many kinds of disease spots, and when various disease spots grow overlap, the disease spot colors will change. The traditional Gaussian model identification algorithm identifies corn disease spots according to the characteristics of image colors, wihieh is unable to identify overlap disease spots and causes the low identification accuracy. This paper presented a fuzzy recognition algorithm. First, the disease spots center was obtained from the training set. Based on the regularity that vein feature vector is not mutated for the same disease spot, through traversing the image feature vector with fuzzy mean, connected graph of disease spot was acquired and the disease spot recognition was completed. Experiments show that the method can identify the corn overlap disease spots accurately and achieve satisfactory resuits.
出处
《计算机仿真》
CSCD
北大核心
2012年第5期251-253,286,共4页
Computer Simulation
关键词
模糊识别
玉米病斑
特征向量
The fuzzy recognition
Corn disease spot
Feature vector