The Study of Forest Types Using Remote Sensing Classification of Daxinganling
|Northeast Forestry University
|texture information semi-covariance function classification of remotesensing
Forest resource information is an important foundation of the state information resources, an important basis for the forestry construction of the various decision-making, and Forest Resource Management and monitoring is an important task of the industrial forest areas. Especially a series retrieval and access of forest information are determined according to the type of forest. As the recognition accuracy of the traditional forestry type remote sensing recognition technology is low and the workload is heavy, the remote sensing data availability is reduced.In order to improve the precision of classification of coniferous, broadleaf, conifer-broadleaf mixed forest by TM images for large areas, the ranomness and spatial distribution of the pixel values of the TM data were considered using geostatistics methods. Taking the Daxinganling forest farm as the research region, the texture information of TM images were calculated using semi-covariance functions with a window of9x9pixels and increment of4pixel for all directions. At the same time, the absolute semi-covariance was selected based on analyzing the factors of affecting the texture information. Then the forest classification results were obtained using the maximum likelihood method of the classical classifier with the texture information combined the original spectral information and normalized differential vegetation index (NDVI). Results show that the accuracy of identifying forest type of the maximum likelihood method with the texture information is72.47%, and the kappa coefficient is0.6738. It gains the purpose of differing the mixed coniferous and is also a feasible forestry type identification technology with the goals that not only reducing the workload but also improving the recognition accuracy especially used in larger studying area.