The Research Of Extracted Tobacco Planting Area Based On Object-Oriented Classification Method

Remote sensing technology has long been a core part of space technology, it has advantage of high amount of information, macroscopic view, objectivity and effectiveness, makes it become the irreplaceable monitoring and management technical means in agricultural, forestry, land and resources sectors. Accurate estimates of crop planting area related to the production forecast, the market price and other important practical livelihood issues. The traditional area estimation method is time-consuming, labor-intensive, consumption funds, and the result not very arccurate because of affected by human factors, can’t meet the management needs. Extracted crops planting area with remote sensing always one of the important issues of agriculture remote sensing, has carried out a wide range of staple crops, with impressive results. In recent years, extracted cash crops planting area with remote sensing has been gradually attention and has become a hot topic. The tobacco industry has significant economic effects of tobacco planting area of remote sensing extract worthy of further study.This article is based on object-oriented classification method, used of ENVI4.8and ENVI EX4.8software, explored the Luobin town tobacco area of Liangshan Prefecture, Sichuan Province extraction method from the point of view of the remote sensing applications. According to the special nature of the cultivation of tobacco and the crop phenology in study area, Optimum Temporal Remote Sensing is the tobacco mulching period at May11, the data include HR camera resolution of2.36m image and PMS camera resolution is10m multispectral data came from ZY-102C. Image preprocessing, cut out remote sensing images of the study area. Using the HSV fusion, Brovey fusion, PCA fusion CN fusion, Gram-Schmidt fusion and PAN fusion method fusion HR image and PMS image, by subjective evaluation and objective evaluation of the optimal fusion method is PAN Fusion. Then use object-oriented software ENVI EX4.8carry on image segmentation. Determine the segmentation parameters is40, merge parameter is90by comparative analysis of the preview window. Conduct Feature selection analysis and selection, establish the rules of tobacco extracted, the Spectral rules exclude building land, bare land and some woodland by chose attribute of the mean and standard deviation of the infrared light, the Texture rules can exclude other cultivated land, the Shape rules exclude road information which similar with tobacco field and small block by chose attribute of ELONGATION, RECT_FIT and AREA. Eventually get tobacco fields distribution of Luobin town and export the final image, estimates the tobacco planting area is7.5thousand, larger than actual situation in value. In order to prove the effectiveness of this method, in this study compared traditional supervised classification Support Vector Machine and Segmentation Support Vector Machine supervised classification with Object-oriented classification in Classification accuracy. Were obtained classification accuracy and Kappa coefficient:85.59%and0.57,90.63%and0.71,93.74%and0.81. It is all evident that the object-oriented classification method can effectively extract the tobacco planting area of Luobin town.