Remote-sensing Classification Method of County-level Agricultural Crops Using Time-series NDVI
zhang rongqun,wang shengan,gao wanlin,sun weijian,wang jianlun,niu lingan
2015-01-01
Abstract:Getting all kinds of crop planting area information accurately is the Agricultural Information Management Department’s main responsibility in order to master the basis of crop production information in an efficient manner. A remote sensing classification method was used based on time-series NDVI that is gathered by using Landsat8 satellite equipped with remote sensing technology. This remote sensing technology possessed a short cycle,performed its analysis in a very speedy manner and used a strong microscope to closely analyze the area it has been assigned to. Based on the analysis of the time-series spectrum character curve,crop type identification and acreage extraction can be effectively achieved.This helped to overcome the confusing agricultural crops classification problem caused by "same object with different spectra"and "foreign body with spectrum"by using a single temporary remote sensing image. In order to accurately ascertain the planting area for the various kinds of crops for providing technical support, the best NDVI threshold range for the crops was studied and the various crop classification rules were explored. The Quzhou County,Hebei Province was taken as the study area,and a distribution map of the study area was made based on this information which was gathered in 2014.Throughout five time phases of Landsat satellite data gathered in 2014,a study on the classification of remote sensing for planting area of winter wheat,summer maize,spring corn,cotton,and millet in the study area was conducted. Classification results can be shown for 2014 with all kinds of crops in the study area,respectively: winter wheat is 27 776. 61 hm2,summer corn is 27 776. 61 hm2,spring corn is2 582. 73 hm2,cotton is 6 485. 94 hm2,and millet is 277. 65 hm2. Using the Kappa coefficient and statistical data to verify the accuracy of this classification,the result shows that the winter wheat,summer corn,spring corn,cotton and millet can be effectively identified,with an overall classification accuracy of 89. 166 7%,along with a Kappa coefficient of 0. 857 4. Compared with the statistical data,the relative margin of error for individual crops is as follows: winter wheat-0. 80%,summer corn-0. 32%,spring corn-3. 15%,cotton-2. 71%,millet 4. 12%. This paper proves that mass crop planting areas can be precisely obtained from analyzing the time-series data of remote sensing images with a medium spatial resolution. It also proves that this method can provide a technical basis for using remote sensing to investigate crop planting areas at a county level.