Published in Scientific Papers. Series E. Land Reclamation, Earth Observation & Surveying, Environmental Engineering, Vol. V
Written by Gabriel POPESCU, Octavian Laurentiu BALOTA, Daniela IORDAN
The paper presents how the use of multispectral LiDAR intensity data for classification has high potentials to increase land classification accuracy. Traditionally, classification of LiDAR data refers to the separation of terrain from other objects based on elevations (range data). Up to about 70% of overall accuracy can be achieved using intensity data only. Land classification accuracy, of about 80%, can be achieved by incorporating both the geometric and radiometric record of LiDAR data. Range and scan/incidence angle have prominent effect on the radiometric correction of intensity data. Radiometric correction of LiDAR intensity data is required for potential use of the LiDAR intensity in land cover classification and radiometric correction can be achieved day or night with similar good results. Current research involves the use of image segmentation and object oriented classification techniques to improve the classification results. The increased number of wavelengths in a sensor has the effect of increasing the information content that can be derived from the target surface and allowing surveying professionals to address many more applications using a single-sensor solution. The complementary information of multispectral LiDAR data may greatly improve the classification performance, especially in the complex urban areas. Use of a minimum of three intensity images from a multi-wavelength laser scanner and 3D information included in the digital surface model (DSM) has the potential for land cover and land use classification. Over 90% of overall accuracy is achieved via using multispectral LiDAR point clouds for 3D land cover classification.
[Read full article] [Citation]
