基于ExG因子的水稻病斑分割技术  

ExG Factor-Based Spot Segmentation Technology for Rice

作　　者：朱盼盼 张正华[1] 郭丽瑞 闫雪纯 张珂元 ZHU Panpan;ZHANG Zhenghua;GUO Lirui;YAN Xuechun;ZHANG Keyuan(College of Information Engineering(School of Artificial Intelligence),Yangzhou University,Yangzhou Jiangsu 225127,China)

机构地区：[1]扬州大学信息工程学院(人工智能学院),江苏扬州225127

出　　处：《信息与电脑》2022年第9期73-76,共4页Information & Computer

基　　金：江苏省大学生创新创业训练计划项目(项目编号:202111117062Y);扬州大学学科特区学科交叉课题“智能化稻麦生产精准变量喷洒关键技术研究与平台研发”(项目编号:yzuxk202008)。

摘　　要：为了监测水稻病害等级并及时做好预防,本文提出一种基于超绿超红算法(ExG+ExR)结合最大类间方差法(Nobuyuki Otsu,OTSU)分割水稻病斑区域。首先,针对传统阈值算法分割效果不佳的缺陷,利用加权平均值进行灰度处理,通过直方图均衡化改善图像质量,并且以自适应中值滤波保护图像细节,从而实现对图像的降噪及增强处理;其次,通过ExG+ExR结合OTSU依次从背景分割出水稻叶片和病斑区域;最后,本文提出的方法与复印称重法和KNN算法比较,在800×300图像中平均绝对准确率高达98.28%,并且只有1.53%的平均绝对误差率。对比结果表明,该算法能够有效分割水稻的病斑区域,为有效识别病害种类及等级奠定了基础。In order to monitor rice disease levels and prevent them in a timely manner,this paper proposes a super green and super red algorithm(ExG+ExR)combined with the maximum inter-class variance method Nobuyuki Otsu(OTSU)to segment rice disease areas.Firstly,to address the defects of poor segmentation effect of traditional thresholding algorithm,we use weighted average for grayscale processing,improve image quality by histogram equalization,and protect image details with adaptive median filtering,so as to achieve noise reduction and enhancement processing of the image;secondly,the rice leaf and disease spot areas are sequentially segmented from the background by ExG+ExR combined with OTSU;finally,the method proposed in this paper is compared with copy-weighing method and KNN algorithm,the average absolute accuracy in 800×300 images is as high as 98.28%,and only 1.53%of the average absolute error rate.The comparison results show that the algorithm can effectively segment the disease spot areas of rice,which lays the foundation for effective identification of disease types and levels.

关 键 词：水稻病斑 灰度处理 直方图均衡化 中值滤波 ExG+ExR算法 OTSU 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]