Extraction of Rice Cropping Area from High Resolution Remote Sensing Image Based on Sample Knowledge Mining

A new method of extracting rice cropping area from high resolution remote sensing image based on sample knowledge mining was proposed to aim at the fact that rice cropping area in high-resolution remote sensing images is actually mixed information of rice, soil, water, weeds and duckweed. Based on the spatial autocorrelation theory, this method makes use of the combined features of rice cropping area based on various basic units of spectrum and texture, to build an extraction strategy of rice cropping area: firstly, achieve image segmentation to obtain the basic unit of various mixed ground information. Then the rice basic unit types are determined by analyzing the basic unit types contained in the rice sample polygon, and the basic units of the corresponding types are all classified into the initial rice cropping area. Finally, eliminate the initial rice cropping area polygons which do not conform to the basic unit combination rule of rice cropping area learnt from the samples. The result of extracting rice cropping area shows that the method in this paper is precise and practical.