首页|Otsu和形态学相结合的人参叶斑图像分割系统

Otsu和形态学相结合的人参叶斑图像分割系统

扫码查看
原文链接
  • 万方数据
  • 维普
针对现场采集的人参图像中背景复杂、不能实时诊断人参叶斑病害严重程度的问题，提出Otsu和形态学相结合的人参叶斑图像分割系统。通过HSV空间模型和Otsu算法，对图像中叶面周围复杂背景的特征进行分割，得到初步分割复杂背景的结果，采用形态学算法，实现复杂背景的精细分割，提取人参叶面目标图像，根据颜色特征信息，将健康绿色叶面与病斑进行分割，实现病斑图像提取，依据分割结果，计算健康叶面、病斑部分的比例，对比预置值，进行人参叶斑病害严重程度分析。结果表明:该方法的病斑覆盖率结果与实际测量值相比误差＜5％，目标和背景判定准确率平均结果分别为91。99％和91。53％。样本图像处理最高耗时10。27s，满足人参现场监测诊断叶斑病害严重程度的要求，也可为其他作物叶斑分割和叶斑病害严重程度诊断提供参考。
Ginseng Leaf Spot Image Segmentation System Based on Otsu and Morphology Algorithm
Aiming at the problems of complex background,low extraction accuracy and difficulty in diagnosis of ginseng leaf spot,a ginseng leaf spot image segmentation system based on Otsu and mor-phology algorithm was proposed.By using the HSV spatial model and Otsu algorithm,the features of the complex background around the leaf surface in the image were segmented,and the preliminary re-sults of segmenting the complex background were obtained.By using morphological algorithms,pre-cise segmentation of complex backgrounds was achieved and target images of ginseng leaves were ex-tracted.Based on the color feature information,the healthy green leaf surface was segmented from the diseased spots,and the extraction of diseased spot images was achieved.Based on the segmenta-tion results,the proportion of healthy leaves and diseased spots was calculated,and the severity of ginseng leaf spot disease was determined by comparing the preset values.The experimental results show that the error between the diseased spot coverage results of this method and the actual measure-ment values is less than 5％,and the average accuracy of target and background determination is 91.99％and 91.53％,respectively.The maximum processing time for sample images is 10.27 sec-onds,which meets the requirements for on-site monitoring and diagnosis of leaf spot disease severity in ginseng.It can also provide reference for leaf spot segmentation and diagnosis of leaf spot disease severity in other crops.

artificial intelligenceimage segmentation systemOtsuginsengleaf spotcomplex background segmentation

刘媛媛、孙嘉慧、王跃勇、于海业

展开 >

吉林农业大学信息技术学院,长春 130118

吉林农业大学工程技术学院,长春 130118

吉林大学工程仿生教育部重点实验室,长春 130025

人工智能 图像分割系统 Otsu 人参 叶斑 复杂背景分割

国家自然科学基金项目吉林省科技发展计划项目吉林省教育厅科学技术项目吉林省发改委创新资金项目吉林大学工程仿生教育部重点实验室开放基金项目

4200125620220402023GHJJKH20190927KJ2019C054K201706

2024

10.13327/j.jjlau.2020.5794

吉林农业大学学报
吉林农业大学

吉林农业大学学报

CSTPCD北大核心
影响因子：1.014
ISSN：1000-5684
年,卷(期)：2024.46(4)