基于多时相高分一号影像水稻监测精度评价与修正——以德阳地区为例

Classification Accuracy Evaluation and Rejustment of Monitoring Rice Planting Area By Multi-temporal GF-1 Satellite Image:Taken Deyang District as Example

以2014年四川省德阳地区为研究区域,建立10个(500m×500m)样方作为训练区,同时期建立水稻验证点。提取2个时相高分一号的NDVI值,分析其变化特征,确定阈值;利用数字高程图(DEM)及坡度图,采用决策树分类方法,进行水稻遥感监测。以水稻地块样点作验证,评价高分一号数据在水稻识别方面的精度,最后利用样方测算的修正系数对遥感监测面积进行修正。结果表明,在类似德阳地块比较破碎的平原和丘陵区域,高分一号影像遥感识别水稻的用户精度可达92.3%,制图精度可达96.5%。以78%系数乘积修正该区域水稻遥感监测面积,得到更为准确的水稻播种面积。高分一号影像作为全新的高空间分辨率遥感数据,在水稻监测方面,可作为一种可靠的、免费的遥感影像替代源在更大区域中探讨使用。

Taken Deyang city in 2014 as studied area, 10 rice samples（500 m×500 m） were set up as training areas, meanwhile rice verification points were established during same period. Applied two adjacent periods muhispectral high spatial resolution image （2014/8/05 8 m） and medium-low spatial resolution image （2014/8/13 16 m） of GF-1 with ENVIS. 1 as platform, the changing characteristics were analyzed, and the thresholds of NDVI were determined. Based on digital elevation map （DEM） and slope map, the rice planting area with decision tree classification method was monitored. Using verification points to evaluate classification accuracy, the monitoring area by factors from the result of rice samples were corrected. The results showed that ： similar to the plains and hilly regions in Deyang, based on digital elevation map （DEM） and two periods images information of GF-1 satellite images, the classification user accuracy could be up to 92.3 %, and producer accuracy to 96.5 %. As 78 % multiplication coefficient amended monitoring area, more accurate rice area was obtained. It was concluded that GF-1 satellite image as a new kind of high spatial resolution data, applied to monitor planting rice in Deyang, achieved higher classification accuracy, which could be used as a reliable and free of charge image source in the other regions.