利用Radarsat-2数据基于比值检测的水稻制图  

Rice mapping based on ratio method using Radarsat-2 SAR data

作　　者：化国强[1,2] 李晨 杨沈斌[4] 葛道阔[1] 李秉柏[1] 

机构地区：[1]江苏省农业科学院农业经济与信息研究所,江苏南京210014 [2]南京市测绘勘察研究院有限公司,江苏南京210004 [3]江苏省镇江高等职业技术学校,江苏镇江212016 [4]南京信息工程大学应用气象学院,江苏南京210044

出　　处：《江苏农业学报》2012年第6期1451-1458,共8页Jiangsu Journal of Agricultural Sciences

基　　金：国家自然科学基金项目(40901238);公益性行业气象科研专项(GYHY201106027;GYHY201006027)

摘　　要：利用2010年获取的2景Radarsat-2全极化SAR数据,探讨水稻及其他地物在不同时相上的后向散射特征差异,并提出在雷达遥感图像上基于比值检测理论的水稻分类制图方法。通过分析研究区内水稻、茭白等主要地物在2个时相上不同极化方式下的后向散射特征,利用比值检测理论建立水稻及其他地物像元概率分布及分布函数并计算出主要地物的分类阈值,在剔除茭白的影响下,完成水稻的分类制图。水稻和茭白作为水生作物具有明显的雷达图像表现特征;在HH/VV比值图像中,分类阈值1.393对水稻的识别最好,制图精度达到92.64%,总误差概率为7.74%。研究结果表明,Radarsat-2数据在水稻监测中具有丰富多样的技术手段以及更为广泛的应用潜力。The difference of backscattering characteristic between rice and other surface object was discussed and the rice mapping based on ratio method was proposed on radar remote sensing image,using Radarsat-2 full-polarimetric SAR image data at August,2010.The radar backscattering characteristic of bi-temporal and multi-polarimetric surface object were analyzed.The results showed that the underlying surface environment of land crops and aquatic crops led to the differences in backscattering characteristic in rice,water bamboo,maize and soybean,etc.The multi-temporal speckle noise was analyzed by inherent backscattering mechanism of aquatic plants,which used to achieve rice mapping.The optimal classification threshold of main surface object was acquired by distribution function of image pixel.The probability distributions and distribution functions of image pixel of rice and other object based on the radio method were established.The rice pixel distribution curve was fitted with the band ratio,and the classification threshold between rice and no-rice area was acquired,being 1.393.By the same way,the classification threshold of water bamboo was 2.03.Excluding the interference of water bamboo,the classification threshold of rice mapping could achieve the best recognition in HH/VV ratio image,with accuracy being 92.64% and probability of error（PE） being 7.74.The results showed that the rice mapping method based on HH/VV ratio change detection had better classification accuracy for agriculture on 1∶ 100 000 rice mapping.The radarsat-2 data had rich variety of technical means and more extensive application potential in rice area monitoring.

关 键 词：分类阈值 后向散射 水稻制图 雷达遥感 

分 类 号：S127[农业科学—农业基础科学]