中国农业科学院农业信息研究所 中国农学会
Agricultural Information Institute of CAAS China Association of Agricultural Science Societies
En
Ch
中国农业科学院农业信息研究所 中国农学会 主办

AI检索

订阅

返回顶部
当前位置: 首页 > 文章 > 基于残差网络的遥感影像松材线虫病自动识别 林业工程学报 2022 (1) 185-191
Position: Home > Articles > Automatic identification of Bursaphelenchus xylophilus from remote sensing images using residual network Journal of Forestry Engineering 2022 (1) 185-191

基于残差网络的遥感影像松材线虫病自动识别

作  者：
周志达;李如仁;贲忠奇;何博洲;史岩岩;戴维序
单  位：
便利蜂商贸有限公司;辽宁工程技术大学测绘与地理科学学院;航天信德智图(北京)科技有限公司;沈阳建筑大学交通工程学院
关键词：
高分一号;Keras;残差网络;病虫害识别;松材线虫病
摘  要：
松材线虫病是针对松树的特殊疾病,具有前期发病特征隐蔽、传播范围广、致病速度快的特点,因此面对林业病虫害问题,对受灾区域染病树木进行高效识别和分类,监测其他区域的树木生长情况,并且根据受灾情况确定损失,进行保险理赔是十分重要的。针对染病树木识别准确率低、识别速度慢的问题,本研究利用遥感影像和残差网络相结合的方法,并对残差网络进行优化改进,最终实现了松材线虫病树木的识别,降低人工识别成本,减少识别错误。以湖北省宜昌市远安县嫘祖镇作为研究区域,利用5幅高分一号遥感影像及部分GPS染病树木点作为数据源;通过图像分割、图像增强和坐标转换等一系列预处理将原始数据制作成带有标签和坐标系结合的数据集;将数据集分别输入残差网络ResNet18、ResNet34、ResNet50、ResNet101、ResNet152模型中,根据模型在训练集和验证集上的F1值,结合模型在两个数据集上的准确率曲线及训练集的损失函数曲线,选择表现最好的模型,对其进行优化改进,提高识别精度。模型在Keras的深度学习平台下使用Python语言进行染病树木识别,研究结果表明,识别精度准确率达到87%,实现了松材线虫病树木识别,较适宜应用于林业病虫害问题诊断。
译  名：
Automatic identification of Bursaphelenchus xylophilus from remote sensing images using residual network
作  者：
ZHOU Zhida;LI Ruren;BEN Zhongqi;HE Bozhou;SHI Yanyan;DAI Weixu;School of Transportation Engineering, Shenyang Jianzhu University;Liaoning Technical University;Convenience Bee Trading Co.Ltd.;Aerospace Xinde Zhitu (Beijing) Technology Co.Ltd.;
关键词：
Gaofen No.1;;Keras;;residual network;;pest identification;;Bursaphelenchus xylophilus
摘  要：
Bursaphelenchus xylophilus is a special disease for pine trees. It has the characteristics of concealed early onset attributes, wide spreading range, and fast pathogenic speed. When forestry pests and diseases occur, it is the most important to effectively identify and classify diseased trees in the disaster area. In addition, it is very useful to monitor the growth of trees in other areas and determine the loss according to the disaster situation so that insurance claims can be made. Aiming at solving the problems of low recognition accuracy and slow recognition speed of diseased trees, this research used the method of combining remote sensing images and residual network, and optimized the residual network. This research realized the automatic identification of pine wood nematode diseased trees, reducing the cost of manual identification and minimizing the identification errors. Leizu Town, Yuan'an County, Yichang City, Hubei Province, was used as the research area, and five Gaofen No. 1 remote sensing images and some GPS-infected tree points were used as the data sources. Through a series of preprocessing such as image segmentation, image enhancement and coordinate conversion, the original data were imported into a data set with a combination of labels and a coordinate system. Finally, the data set was input into the five residual network models, i.e., ResNet18, ResNet34, ResNet50, ResNet101, and ResNet152, respectively. Based on the F1 values of the training set and the validation set of the model, being combined with the accuracy curve of the model on the two data sets and the loss function curve of the training set, the best performing model was selected. Using the model, recognition accuracy was optimized. Keras was selected as the deep learning platform for this research, and the model used Python language to identify diseased trees. The research results showed that the accuracy of the optimized model identification reached 87%, which realizes the automatic identification of pine wood nematode diseased trees, being more suitable for the diagnosis of forestry diseases and insect pests.

相似文章

导出引用
计量
文章访问数: 19
HTML全文浏览量: 0
PDF下载量: 8

所属期刊

推荐期刊

友情链接Friendly link
农业信息研究所Agricultural information institute 农业智能知识服务平台Agricultural intelligent knowledge service platform 农业学术服务平台Agricultural Knowledge Service System 中国知网CNKI 维普网VIP 万方WANFANG DATA
联系我们Contact us

电话Tell：010-82109652-808

邮箱E-mail：agrijournal@caas.cn

服务Service
关于我们About us 意见建议Opinions and suggest
版权说明Copyright
单位：中国农业科学院农业信息研究所 中国农学会 © 2017中国农业科学院农业信息研究所 ICP备案号：京ICP备09089781号-46