Study On The Artificial Grassland Extraction From Remote Sensing Image Based On Mathematical Morphology

A vast area of grassland resource distributes in China, the grassland is not only the material basis for development of national economy, but also the natural barrier of protecting land bio-environment. At the same time the grassland is also a relatively fragile ecosystem, and the natural grassland resource faces such problems as area decreased, the quality declined, the contradiction of grass resource and livestock sharpened, and so on,only relying on natural grassland has not reached the requirements of developing animal husbandry and maintaining ecological balance. And artificial grass has the important status and role in the modern agricultural production for its ecological and economic function, so it is imperative to develop artificial grassland moderately. But how to build a reasonable artificial grassland extraction model, which used for access to the geographical position,area, as well as the distribution range of artificial grass area accurately and duly, has important practical significance and theoretical value for the expansion and construction of artificial grassland.Remote sensing monitoring with the features of large-scale of the area, small-scale of the time and grasping the situation of grass resource in real-time, could provide strong support for the dynamic monitoring and analysis management of the grass resource.Mathematical morphology is based on the method of set theory, can be used to analyze and process data through expansion, corrosion, opening operation, closing operation and so on,thus obtaining remote sensing gradient image with small noise clear boundary, which achieve the effect of edge detection of remote sensing image. At the same time, the reasonable selection of structure element in mathematical morphology method is a powerful tool for the quantitative description of the geometry object, it can be used to analyze the geometrical characteristics and structure form of image, thus achieving the purpose of extracting image target information.Taking Taipusi Banner, Inner Mongolia as an example, the artificial grassland extraction model based on mathematical morphology was built. Firstly, the remote sensing image was preprocessed by radiometric calibration and atmospheric correction, etc. Secondly, through the analysis of the geometric feature of artificial grassland, using the method of mathematical morphology to do edge detection of image,the edge gradient map was obtained and binarization processed. Thirdly, the artificial grassland extraction model based on the circular ring structure element in binary image was built, the circular artificial grassland with regular shape was extracted.Comparing with the actual number and area of artificial grassland, verified that the extraction model has higher precision. At the same time, for the part of artificial grassland with irregular shape, this article selected the BP neural network. The image data corresponding to the artificial grassland with regular shape and irregular shape was regarded as the training set, and the whole image data was regarded as the test set, the artificial grassland in the whole image was extracted through training, testing and simulating, which makes the algorithm precision reach the higher demand.The artificial grassland extraction model based on mathematical morphology and BP neural network reasonably extracted the artificial grassland in Taipusi County, the extraction accuracy of quantitative is higher, reached 83.56%, the extraction accuracy of area is higher, reached 83.56%, and the applicability of this model is wide, can be used in different areas, different features with similar geometric features.