一种利用形态相似性的水稻信息提取算法  

A Rice Information Extraction Algorithm Using Morphological Similarity Measurement

作　　者：高永康 王腊红 陈家赢[1] 李建民 GAO Yongkang;WANG Lahong;CHEN Jiaying;LI Jianmin(College of Resources and Environment,Huazhong Agricultural University,Wuhan 430070,China)

机构地区：[1]华中农业大学资源与环境学院,武汉430070

出　　处：《遥感信息》2020年第2期76-86,共11页Remote Sensing Information

基　　金：国家重点研发计划项目(2016YFD0800907);国家自然科学基金项目(41101192)。

摘　　要：针对一般相似性算法对曲线形态欠缺考虑或形态敏感性较差的问题,提出了一种新的曲线形态相似性度量算法。使用该算法计算未知像元NDVI时序曲线与标准NDVI时序曲线的相似性,并基于特征曲线的相似性实现地物的识别。算法通过特征距离加权,对曲线偏移距离进行了差异性放大,相较于欧式距离、动态时间弯曲距离等,提高了对曲线形态的敏感性,降低了算法的复杂性。对该算法的有效性检验显示,5种地物用户精度平均为98%,制图精度平均为92%,表现了较高的可靠性。基于该方法对仙桃市2005-2015年间的水稻种植信息提取。结果显示,该区域单季稻分布基本稳定,总体制图精度为93.29%,总体面积精度为80.57%,达到较高水平。A new curve shape similarity measurement algorithm is proposed to solve the problem that the general similarity algorithm lacks consideration of the curve shape or its shape sensitivity is poor,which calculates the similarity between the unknown pixel NDVI time series curve and the standard NDVI time series curve,and realizes the object recognition based on the similarity of the characteristic curve.The algorithm enlarges the difference of the offset distance of the curve by weighting the characteristic distance,comparing with the European distance,dynamic time bending distance and so on.It improves the sensitivity of curve shape and reduces the complexity of the algorithm.The validity test of the algorithm showed that the average user accuracy of the five objects was 98%,and the average mapping accuracy was 92%,which indicated high reliability.Based on this method,the information of single rice planting in Xiantao City from 2005 to 2015 was extracted.The results showed that the distribution of single-cropping rice in this area was basically stable,the overall mapping accuracy was 93.29%,and the overall area accuracy was 80.57%.

关 键 词：MODIS NDVI 时序曲线 形态相似性 单季稻 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]