基于BP神经网络的马铃薯遥感识别图像数据分析研究

Research on Image Recognition of Potato Remote Sensing Based on BP Neural Network for potatoes

运用遥感手段精确监测马铃薯种植面积是稳固马铃薯主粮化政策、维护国家粮食安全的必要保障。本文以吉林省长春市九台区纪家镇、兴隆镇为研究区,选用landsat8 OLI遥感数据,借助ENVI平台构建了基于BP神经网络的土地覆盖分类模型,应用于研究区的马铃薯等作物分类研究。以landsat8 OLI7个彩色波段作为输入,不断调节分类参数,最终确定了最优分类网络结构。结果显示,BP神经网络法马铃薯的分类生产者精度为94.22%。研究表明,BP神经网络分类方法是一种手段灵活、结果较准确的马铃薯遥感识别方法。

The efficient and accurate remote sensing means of potato acreage monitoring is vital to stabilize potato production as a staple food and maintain the necessary guarantee for national food security.In this paper,a land cover classification model based on BP neural network was constructed to detect potato and other crops in the area of Jijia Town and Xinglong Town,locating in Jiutai District,Changchun City,Jilin Province.Taking the seven color bands of landsat8OLI as input,the classification parameters are adjusted continuously,and the optimal classification network structure is finally determined.The results showed that the producer accuracy of potato was94.22%.The research shows that BP neural network classification method is a kind of potato remote sensing identification method with flexible measure and accurate classification result.