Research Of Maize Yield Prediction System Based On RS And TSDM

The study of precision agriculture is the rapid development in the world, in China, the investments of precision agriculture is also in constant increase. Using remote sensing technique for crop growth for dynamic monitoring, crop planting area of crops were calculated and on production forecasts have become the research hotspots precision agriculture in China. This article is from remote sensing image used the state 863 project mentor presided over "corn accurate operation system research and application". Acquisition of remote sensing image generally can't directly use remote sensing image processing software; need to carry on the processing, extract the project information. This paper use of remote sensing image processing software for ENVI has strong ENVI, multi-spectral imaging function, can fully extracted image information and have rich complete projection packages, can support various projection type.This paper YuShuShi in JiLin province northeast corn as the research object, using remote sensing technology, image processing technology, data mining technology, the corn crop of planting area for yield prediction.(1) In remote sensing image processing, established the normalized vegetation index and the relation model between the maize yield. In the remote sensing measurement of corn task is to use old period similar phase spectrometry remote sensing images, after a series of remote sensing image processing after this, using principal component analysis selecting the main factors of decision maize yield. Extract normalized vegetation index and statistical output information based on the relation of corn yield, constructing normalization vegetation index with regression equation. To establish a team with high accuracy and operability of the corn production model test.(2) The principal component analysis of data dimension reduction, establish corn crop time series prediction model. From remote sensing image can gain mass data information, although after the principal component analysis method eliminates a small amount of influence maize yield factors, data quantity is very big still. Reflect the timing maize yield normalized vegetation index has very significant trend, the purpose is to find the analysis of a sequence of this trend, and use the trend of development of sequence make reasonable forecast. Based on maize yield is a random non-stationary sequence, so this paper summation auto-regressive sliding average model (ARIMA).(3) Yield prediction model of model matching and parameter estimation, identified the prediction precision. Sequential remote sensing image from the past for maize growth situation reflects the influence factors as research data on corn yield, through the time sequence of the autocorrelation function of the partial correlation function analysis, can judge ARIMA (p, d, q) model and maize yield sequence identification is better. Experimental application of model test data of six production, inspection results show that the precision of the 1,2,1) model ARIMA (ideal prediction effect, and the actual yield data, error precision control compared around 5%.