Abstract
Recently, Lidar data becomes popular data source for creation of spatial database. Lidar intensity image reveal information about land cover in the sort wave infrared spectral range with high spatial resolution can be used in combination with other Lidar data such as DEM, DSM, simultaneously take airphoto and high resolution satellite image for land cover classification at sub meter resolution. The article introduce common methods for classification namely: decision tree; object oriented; maximum likelihood, nearest neighborhood, neuron network. These methods are experimented with 4 different combination of above mentioned input data sources for test area in Bac Giang city. Classification evaluation using error matrix shows that in the best case we can distinguish 8 land cover classes: grass and bush; high tree; dry vacant land; wet vacant land; asphalt road; tiled roof; concrete roof with accuracy around 85-90% when applying neuron network method using all the available input data. The research result opens up opportunity to quick establishment of land cover layer from Lidar data to add to large scale geodatabase.
PDF (Tiếng Việt)
| Download: 117
Downloads
Download data is not yet available.