The Study Of The Crop Extraction Model Based On NDVI Time Series

The extraction of crop information is the main research content of agricultural remote sensing.The remote sensing classification method based on NDVI time series data can give full play to the characteristics of short cycle,fast speed and strong macro.However,there are some problems in the previous research,such as low resolution of data source,single classification method and so on.Therefore,this paper using high spatial resolution and high temporal resolution advantage GF-1 as data sources to construct multi temporal NDVI time series curve,from the perspective of NDVI time series similarity analysis,the traditional extraction model is improved.In the traditional model,the spatial vector extraction model and the coordinate transformation extraction model is simple and feasible,and the overall accuracy is higher than 75%,study on large area crop classification which can be applied to agricultural remote sensing,But the extraction capacity of certain ground objects(such as minor crops)is limited;The result of the curve integral extraction model is affected by the similar integral value of different NDVI time series curves,the parameters of the extracted model are overlapped greatly,so the reliability of the model is low;In order to improve the extraction accuracy of the model,the similarity analysis conditions of the three extraction models are strengthened,and a new improved model is constructed.The improved model breaks through the limitations of a single model,reduces misclassification error and improves classification accuracy.However,there are still some defects in the model(the correction of the leakage error can not be completed),there is still room for further improvement.Based on the experimental results,the overall improvement of the accuracy of the improved model is 92.42%,reached a high accuracy,verify the feasibility and applicability of the research lies in the accurate classification of large area,many crops,has a certain application value and promote the significance.At the same time,this study explored the potential application of GF-1 data and provided a new approach for crop extraction in multitemporal NDVI time series.