基于Landsat的棉田各生育期面积提取算法优选:以新疆阿拉尔垦区为例  

Optimization of cotton field area extraction algorithm in different growth stages based on Landsat:a case study of Alar reclamation area in Xinjiang

作　　者：鲁新新 张丽 郝梦洁 阿迪力·亚森 蒋青松[1] LU Xinxin;ZHANG Li;HAO Mengjie;ADILI·Yasen;JIANG Qingsong(College of Information Engineering,Tarim University,Alar,Xinjiang 843300)

机构地区：[1]塔里木大学信息工程学院,新疆阿拉尔843300

出　　处：《塔里木大学学报》2021年第1期98-105,共8页Journal of Tarim University

基　　金：中央高校专项基金中农-塔大联合项目“基于近地传感-模型耦合的南疆农田土壤盐分监测与三维制图”(ZNLH201904);国家自然科学基金项目“盐分对南疆土壤有机质高光谱特征与定量反演的影响及方法研究”(41361048)。

摘　　要：为了获得不同生育期棉田面积提取的最优算法,本研究以南疆阿拉尔垦区Landsat多时相遥感图像数据为数据源,利用ENVI对遥感影像进行数据预处理,并对研究区掩膜提取后,分别利用最大似然、神经网络、随机森林和支持向量机等算法并结合参数调节手段,分别对四个不同生育期的棉田面积进行提取,并对精度误差加以比较。结果显示,最大似然算法在棉花苗期、蕾期和吐絮期可以获得最好的计算精度,误差比分别为:0.56%、5.75%和8.97%;神经网络在花龄期有较好的提取效果,误差比为7.85%。该研究可为南疆地区不同生育期阶段棉田面积提取算法的优选提供理论参考。In order to obtain the optimal algorithm of cotton field area extraction in different growth periods,this study took the Landsat multi temporal remote sensing images of Alar Reclamation Area in southern Xinjiang as the data source,used ENVI to preprocess the remote sensing images,and extracted the mask of the study area,then used maximum likelihood,neural network,random forest and support vector machine combined with the means of parameter adjustment Four different growth stages of cotton field area were extracted,and the accuracy error was compared.The results show that the maximum likelihood algorithm can obtain the best accuracy in the cotton seedling stage,bud stage and boll opening stage,and the error ratios is 0.56%,5.75%and 8.97%respectively;the neural network has a good extraction effect in the flowering stage,and the error ratio is 7.85%.This study can provide a theoretical reference for the optimization of cotton field area extraction algorithm at different growth stages in southern Xinjiang.

关 键 词：遥感 面积提取算法 优选 精度比较 

分 类 号：TP753[自动化与计算机技术—检测技术与自动化装置]