摘要
研究水稻病斑中的重叠病斑识别问题,针对诊断水稻图像信息和及早防治,传统的水稻病斑分割算法都是对病斑的像素进行直接的操作,容易造成像素信息的丢失,存在着识别准确性差问题,造成后期的识别率不高。为解决上述问题,提出了基于像素概率模型的水稻病斑分割方法。通过自适应病斑分割算法,用高斯混合模型描述每一像素的色彩分布情况,再以具有最大适应度值的子模型作为当前分布模型来描述每一病斑像素的特征,避免了对像素的直接操作。仿真结果表明,方法能够有效的分割大部分重叠水稻病斑特征,提高了识别准确性,取得了比较好的效果。
Research the overlap of rice lesion segmentation.The traditional rice lesion segmentation algorithm directly processes the lesion pixels,which easily leads to the loss of pixel information,a bad segmentation,and low recognition rate.For this problem,we propose a probabilistic model based on pixel rice lesion segmentation.By adaptive lesions segmentation algorithm,Gaussian mixture model is used to describe the color distribution of each pixel,and the maximum fitness value is as the current distribution model for sub-model to describe the characteristics of each lesion pixel,avoiding the pixel direct operation.Experiments show that this method can effectively separate most of the characteristics of overlapping lesion of rice and achieved better results
出处
《计算机仿真》
CSCD
北大核心
2011年第2期341-344,共4页
Computer Simulation
基金
国家863课题(2008AA10Z220)
河南省重大科技攻关计划(082102140004)
关键词
水稻病斑重叠
病斑分割
像素概率模型
Rice lesion overlap
Rice lesion segmentation
Probabilistic model based on pixel