Cotton Estimation Model Based On Time-series Remote Sensing Images

In the development of nature and human society production and life,land is very important,and many things are closely related to the land.Since the 20 th century,with the rapid development of society and economy,people's demand for land resources has increased gradually,and the contradiction between human society and land has become more serious.At the same time,the development of food and its security in recent years is also related to the survival and development of human society,and the land is the most basic element of food production.Arable land is the essence of the land,and it is the most basic and irreplaceable data of agricultural production.Under the background of scarce land resources,precision agriculture technology is an inevitable trend of agricultural development.Precision agriculture can be dynamic,real-time,macro,to monitor the quality of soil and vegetation information quantitatively and vegetation is the floorboard of the surface of plant community,including crops,also is connected to the water,atmosphere and soil,vegetation changes to represent the change of land cover.Since the 1970 s,when the LACIE program,an estimated crop of arable land implemented in the United States,has begun,the prediction of crop yields has been a major concern of national governments.Eventually crop yield is the main purpose of the farmland cultivation,and evaluation of land resource productivity and direct index of the farmers' income.In addition,crop yield is one of the most concern of the state and farmers.The timely and effective forecasting and forecasting of crop yield can not only serve the national government's macro decision-making,but also provide guidance for farmers' grain storage and grain trading.But plenty of factors can affect the final output of crop,simple production forecast and cannot satisfy the requirement of the high precision,accurate production prediction and forecasting will be helpful to the country's grain policy formulation and the adjustment of grain prices,and the development of rural economy and foreign trade of grain.Therefore,it is very important to establish a good and executable crop yield prediction model.This paper was based on California's San Joaquin Valley region in the western United States,and selected the Sheely Farm plot as the research area,two cotton with time series Landsat_5TM,Landsat_7ETM remote sensing image data sources for research.And this paper combined remote sensing images with variable yield machinery for production of field survey data,through the optimization modeling method,to obtain universal higher cotton yield estimation model.In this paper,we studied mainly divided into three parts,the first part: obtain the non-synthesis,contain accurate phase information of vegetation index?VI?time series data,and build a precise mathematical model,to describe the VI in the process of crop growth real change rule;the second part: considering that it is the most difficult to get a lot phase images at the same time,so this paper analyzed time-series VI curve,and reconstructed the time-series VI curve by the mathematical model,finally this paper used a few times VI to build the entire time series curve,and the fitting accuracy of the reconstruction model is evaluated;the third part: this paper extracted time-series curve characteristic parameters,and correlation analysis with cotton yield,to determine the best yield factor.This paper established the estimation model and do the model precision evaluation,and analyze the operability of the reconstruction time series method of mathematical model.Finally,conclusions was drawn as follows.?1?This paper was through the analysis of time-series VI curve,aiming to extract the characteristic parameters of the curve,and analyze the correlation coefficient between single VI and multi-VI and cotton yield,and the optimal estimation phase is determined.When the real Landsat vegetation index and the reconstructed vegetation index were used,the maximum period of vegetation index and yield correlation coefficient was the best.As for single yield estimation model,when took NDVI on the DOY 198?July 14?as the input variable the model had a high accuracy;for the multi-phases cotton yield estimation,when took the average NDVI during the blooming period,it was the optimal model;?2?In this paper four kinds of mathematical model were tested into simulate the time-series VI curve,and finally this paper found the Extreme model had the highest model accuracy;for the reconstructed model,the result of EVI is better than NDVI,the trend of time-series EVI curve is more consistent with the real time-series curve;?3?After the reconstruction of time-series VI curve by mathematical model,the accuracy of yield estimation model is improved,compared with that of pre-reconstructed vegetation index and output.At the same time this paper also reconstructed the time-series VI curve throughout the growing season using VI of few phases in the study area during the period of cotton,and it was found that when used five phases,the evaluation effect was also good,and the yield estimation model had a high accuracy.The result of this paper also solves the problem of high spatial resolution image,such as the low acquisition rate of Landsat time.To sum up,through this study,the time-series VI curve can be reconstructed by using mathematical model,and the time-series VI curve can be obtained by day.The accuracy of the yield estimation model can be improved.In this paper,the results can provide theoretical foundation for the domestic crop yield estimation.