面向对象和多分类器的棉花出苗信息快速提取方法

Rapid extraction method of cotton seedling information with object oriented and multi-classifier

为了及时准确地获取棉花苗期的出苗信息,提出了采用面向对象的分类思想,充分利用了图像的纹理、形状和颜色等信息。采用支持向量机、朴素贝叶斯、K最邻近、决策树及随机森林分类器对无人机获取的可见光影像进行棉花出苗信息的快速提取,将目视解译的Shape文件与5种分类器的分类结果进行对比及精度分析。结果表明,最佳机器学习分类器为支持向量机,研究区棉花出苗信息提取总体精度达97.47%。研究可为无人机在农业苗情信息诊断中的应用提供参考,为精准化管理提供支持,达到丰收增产的目的。

In order to obtain the information of the emergence of cotton seedlings in a timely and accurate manner,an object-oriented classification method was proposed to make full use of the texture,shape and color of the image.Support vector machine,naive bayes,K nearest neighbor,decision tree and random forest classifier were used to extract the information of cotton emergence from the UAV image.The visual interpretation of Shape file was compared with the classification results of five classifiers and the accuracy was analyzed.The results show that the best machine learning classifier is support vector machine(SVM)the extraction accuracy of cotton seedling emergence information is 97.47%.This study can provide a reference for the application of UAV in the diagnosis of agricultural seedling situation information,and provide support for the precision management,so as to achieve the goal of harvest and yield increase.