Research Of Cotton Yield Estimation Based On Vegetation Index And Temperature Factor

Xinjiang is the main production area of cotton in china. The change of cotton's sown area and output influences cotton's output and stock situation of nation immediately. Monitors Xinjiang cotton growing trend scientifically and forecasts its output accurately, may provides prompt accurate cotton production state of agricultural production for all levels of the government, and it has the significant significance carries on the agricultural decision-making to the government.The research take Kalamy as an example, take the remote sensing material as the foundation, through extraction cotton normalization vegetation index, union temperature factor, establish multi-dimensional linear regression estimation model of cotton per unit area yield ,by time enhance the precision of field estimation,and provide the scientific basis for big area output estimation of cotton in Xinjiang.After analysis, altogether 10 NDVI factors has the good relevance with the cotton output, moreover the time interval is quite centralized ,mainly in late June to late July as well as early September. And there are 8 factors in late June to late July's 8 factors, is being related with the cotton output; 2 factors in early September's is inverse correlation with the cotton output. Altogether 5 temperature factors is remarkable relevance with the cotton per unit area yield's, and all is inverse correlation, moreover the same as NDVI factors, the time interval is quite centralized, concentrates in May and mid-August.The study using force introduction law, establish Multi-dimensional linear regression equation based on July NDVI accumulation value accounts for the main period of duration NDVI accumulation value percentage and average temperature in mid-August ten-day period : Y=-1624.061+11855.483XN10-40.679XT1. The value of F is 13.297, and the significance level is 0.017.Carrying Multi-dimensional linear regression model and the single factor regression model on the contrast examination. After examination, single factor linear regression model based on factor N10 , has estimation relative error between -8.6﹪and 13.3﹪, index regression model based on factor N10 has relative error between -7﹪and 11.8﹪. Single factor linear regression model based on factor T1 has estimation relative error between -13.4﹪and 10.7﹪. The multi-dimensional linear regression model based on factor N10 and factor T1 has estimation relative error between - 5.9﹪and 9.9﹪.We may see the multi-dimensional linear regression model's estimation precision is higher, and the model stability is better, it has certain feasibility.