基于深度学习的苹果叶面病害缺素识别系统  被引量:1

Apple leaf disease deficiency based on deep learning

在线阅读下载全文

作  者:郑淋萍 李万益 李南健 黄灿敏 林祎琦 陈津毅 邝芸 郭小芸 Zheng Linping;Li Wanyi;Li Nanjian;Huang Canmin;Lin Yiqi;Chen Jinyi;Kuang Yun;Guo Xiaoyun(School of Computer Science,Guangdong University of Education,Guangzhou 510303,China)

机构地区:[1]广东第二师范学院计算机学院,广州510303

出  处:《现代计算机》2023年第24期26-32,共7页Modern Computer

基  金:广东省大学生创新训练计划资助项目国家级大学生创新训练计划资助项目(S202314278042);广东省教育科学规划课题(2022GXJK073、2023GXJK125);广东省高等学校党的建设研究会2022年党建研究课题(2022BK024);广州市哲学社会科学发展“十四五”规划共建课题(2023GZGJ171);广州市基础与应用基础研究项目(202002030232);广东第二师范学院高等教育教学改革项目(2022jxgg33);广东第二师范学院高层次人才培养特殊支持计划项目(2022年优秀青年教师培养对象:李万益);广东第二师范学院思政专项研究项目(2022SZZX13);广东第二师范学院2023年校内科研项目(学生工作专项)(2023XSGZ05)。

摘  要:针对叶面病害对苹果单产水平的提升产生负面影响的问题,将斑点落叶病、花叶病、褐斑病、灰斑病和锈病等五种作为重点。主要研究内容及结果如下:在苹果叶面存在差异以及图像环境不一的影响下,对病害图像进行去噪操作后,通过对关注主体的图像增强及Ostu法进行彩色图像分割,得到病斑特征与背景灰度的强度比图像。记下分割后的二值化图像的像素灰度值为0的坐标,并使原图在对应位置的像素灰度值为0。通过对卷积神经网络的特征提取,得出颜色、形状、纹理三种病斑特征参数。运用数学方法提取病斑的H方差后,得到H-S直方图为区分病斑的主要参照,采用一定的权重对特征参数建立病叶特征判别函数。根据判别函数对苹果叶片病害进行有效的统计学归类,建立Swin Transformer识别模型,将图像从测试数据集精准分类到特定的叶面病害类型。对卷积神经网络模型进行改进后,训练获得AlexNet-F苹果病害识别模型。采用Python代码运行程序,使用Python语言的Django框架,将苹果叶面病害识别模型及知识图谱作为核心,融合前端、数据库技术来开发系统。结果表明,上述分类操作能够对重点的五种苹果叶面病害图高效识别,满足项目作为识别系统的需要。In response to the negative impact of leaf diseases on the improvement of apple yield,five key diseases were identified,including leaf spot disease,mosaic disease,brown spot disease,gray spot disease,and rust disease.The main research contents and results are as follows:Under the influence of the difference of apple leaf surface and the different image environment,after the denoising operation of the disease image,the intensity ratio image of the disease spot feature and the background gray level is obtained by image enhancement and Ostu method of color image segmentation.Write down the coordinates where the pixel gray value of the binary image after segmentation is 0,and make the pixel gray value of the original image at the corresponding position be 0.Through the feature extraction of convolutional neural network,three characteristic parameters of color,shape and texture were obtained.The H‑S histogram was obtained as the main reference for distinguishing diseased spots,and the characteristic discriminant function of diseased leaves was established with certain weights for the characteristic parameters.According to the discriminant function,the apple leaf diseases are effectively classified statistically,and a Swin Transformer recognition model is established to accurately classify images from the test data set to specific leaf disease types.After improving the convolutional neural network model,AlexNet‑F apple disease recognition model was obtained.Use Python code to run the program,use the Django framework of Python language,take the identification model of apple leaf disease and knowledge graph as the core,and integrate front‑end and database technology to develop the system.The results show that the above classification operation can effectively identify the key 5 kinds of apple leaf surface disease maps,and meet the needs of the project as an identification system.

关 键 词:苹果叶面病害 二值化图像 卷积神经网络 Swin Transformer识别模型 

分 类 号:TP18[自动化与计算机技术—控制理论与控制工程] TP391.41[自动化与计算机技术—控制科学与工程] S436.611.1[农业科学—农业昆虫与害虫防治]

 

参考文献:

正在载入数据...

 

二级参考文献:

正在载入数据...

 

耦合文献:

正在载入数据...

 

引证文献:

正在载入数据...

 

二级引证文献:

正在载入数据...

 

同被引文献:

正在载入数据...

 

相关期刊文献:

正在载入数据...

相关的主题
相关的作者对象
相关的机构对象