Research And Application Of The Surface Object’s Classification Method Based On Remote Sensing Images For Typical Tempered Grassland

Grassland is the largest land ecosystem which plays a pivotal role on agriculture, animal husbandry and human survival and development in the process of energy flow and material cycle. At present, the situation of ecological environment is quite serious, such as grassland ecological degradation, grass production, massive sandstorm, soil erosion resulted from of long-term overgrazing, man-made destruction. So, it is very important to provide accurate, complete and rich scientific basis of the grass for the relevant departments. In the process of dynamic monitoring, remote sensing image classification is an essential and the key research content. In recent years, people started to analyze satellite remote sensing data and introduce the supervised classification and the artificial intelligence classification to the measurement of grassland biomass and some other factors of grassland. However,Typical tempered grassland of Inner Mongolia whose grassland area is large with simple landscape and a few land use types. However, the research on the classification of the surface material of Typical Tempered Grassland has not formed a practical classification method. So this paper focuses on remote sensing image classification method, and the experiments conducted on two typical tempered grassland of Hulunbuir and Xilinhot demonstrate the maximum likelihood classification method fits for the Typical Tempered Grassland.Specific contents are listed as below:(1) Preprocessing of the remote sensing image, including stripe image processing, bright line elimination, noise reduction processing, image filtering and image enhancement,etc.(2) Integrating the optimum bands of remote sensing image, bands 7,4,3 with R,G,B color by OIF method, according to the calculation of band’s standard deviation correlation coefficient.(3) Classification of the various remote sensing image includes ISODATA, K means of supervised classification, and the minimum distance, maximum likelihood classification method of unsupervised classification. Post-processing was conducted on the classification results including majority/minority analysis, clustering analysis, combination of classes and assigning different classes colors, and so on.(4) Utilization of classification accuracy evaluation and classification result, the confusion matrix method were used for the assessment of classification accuracy, and the land utilization was analyzed as well.The experimental results demonstrate that maximum likelihood classification method has the highest classification accuracy for remote sensing image of typical tempered grassland.