基于GF-1影像NDVI年度间相关分析的冬小麦面积变化监测

为实现区域冬小麦种植面积变化的快速监测,减少监测难度,提高监测效率和精度,该文提出一种基于年际NDVI相关关系的监测方法(relationship analysis of normal difference vegetation index,rNDVI)。选择河北省黄骅市、孟村县、海兴县3个县市为研究区,基于2014年4月14日、2017年4月26日两个时期的GF-1/WFV数据,基于rNDVI方法,通过将样本点两年度的NDVI值构建二维空间,采用最小二乘法拟合的方法获得不变地物点的上下包络线方程,进而得到冬小麦变化区域的监测阈值,提取冬小麦种植增加和减少区域,实现对研究区域的变化监测。结果表明,采用rNDVI算法总体精度分别为90.60%,Kappa系数为0.84,相比传统的先最大似然分类后再提取冬小麦种植变化区域的方法,总体精度与Kappa系数分别提高了6.6个百分点和16.7%。对冬小麦增加区域、冬小麦减少区域的变化监测结果进行分析,发现基于rNDVI的变化监测方法可以有效提高裸地、线状道路、破碎的冬小麦地块等区域的变化识别能力,提高监测精度。同时分别利用2014年3月1日和2017年3月12日、2014年5月17日与2017年5月20日两对GF-1/WFV数据进行基于rNDVI的冬小麦变化区域监测,结果表明3月份的监测精度较低,主要是由于3月份冬小麦长势尚不明显,5月份与4月份的总体精度相近,主要是由于5月份冬小麦NDVI已较高,易于识别。上述研究结果表明,基于rNDVI的冬小麦变化快速监测方法可以有效监测区域冬小麦种植面积的变化情况,算法简单高效,且能够在种植结构相对单一的冬小麦分布区域保持较高精度,能够满足农情遥感监测信息快速获取的需要。

Area change monitoring of winter wheat based on relationship analysis of GF-1 NDVI among different years

Abstract: In order to achieve fast monitoring of regional winter wheat area change, reduce monitoring difficulty, and improve monitoring efficiency and accuracy, the paper proposes a monitoring method based on the relationship analysis of normal difference vegetation index (rNDVI). By selecting 3 counties i.e. Huanghua, Mengcun, and Haixing County, Hebei Province as the study area, and by taking GF-1/WFV data of 2 dates i.e. April 14th, 2014 and April 26th, 2017, the paper conducted the monitoring in the study area by extracting the increased and decreased areas of winter wheat planting areas. Based on rNDVI, the paper built a two-dimensional space with two-year NDVI values of sample points, and thus obtained the monitoring threshold values of the changing areas of the winter wheat by employing the least squares fit method to obtain the upper and lower envelope equations of unchanged ground objects. The result shows that, the overall accuracy of rNDVI algorithm is 90.60%, with the Kappa coefficient of 0.84. Compared with the traditional method of making maximum likelihood classification and then extracting changed winter wheat planting areas, the overall accuracy of this method and its Kappa coefficient are improved by 6.6 percentages and 16.7% respectively. Analysis of the monitoring results on the change of the winter wheat increased area and decreased area shows that, the monitoring method based on rNDVI can effectively improve the identification ability on the change of the land areas such as bare land, linear roads, and fragmented winter wheat areas, and improve monitoring accuracy. The monitoring on the winter wheat changed area was conducted based on 2 pairs of GF-1/WFV data of March 1st, 2014 and March 12th, 2017, as well as May 17th, 2014 and May 20th, 2017. The result shows that the monitoring accuracy of March is relatively low, and the overall accuracies of May and April are close. The above study results show that, fast monitoring method of winter wheat change based on rNDVI can effectively monitor the change of the regional winter wheat planting area. The algorithm used in this method is simple and effective, and it can maintain relatively high accuracy for the sample planting structure of winter wheat planting area, and it can also meet the demand for fast acquisition of crop remote sensing monitoring information.