LiDAR Data Classification

Classification is the process whereby the acquired LiDARpoints are filtered, and those representing ground and above ground features (such as trees and buildings) are assigned to separate classes. LiDAR data can be classified into various categories including ground, vegetation, water body, and buildings. Typically, each LAS file is classified as bare-ground or not bare-ground according to the American Society for Photogrammetry and Remote Sensing (ASPRS) LAS format classification table (at a minimum):

1. Class 1 – Unclassified (non-ground)
2. Class 2 – Bare-earth Ground
3. Class 7 – Noise (low, high or manually identified)
4. Class 9 – Water (shots from water surface of oceans, lakes, rivers, or streams derivedfrom the breaklines generated from the intensity images)
5. Class 12 – Overlap

OverlapAn automated filtering process is first applied where various classes of points are separated. General parameters are set for terrain type (i.e. flat, rolling, hilly) and terrain cover (i.e. open/ non-vegetated, light vegetation, medium vegetation, heavy vegetation), along with other parameters that help fine-tune the automated classification. Vegetation and any other structures are initially separated using an automated process. While the automated classification process often classifies 80% or more of the undesirable above ground features, it also erroneously classifies objects such as natural terrain (hills, rock cuts), or man-made features that should be moved out of the ground Class 2. Therefore, a manual analysis using independent checks is performed to produce the final LiDAR point files. Supplementing automated terrain filtering, LiDAR technicians perform interactive processing to achieve reliable bare earth conditions. The resulting elevation accurately depicts the bare earth surface (Class 2). Class 12 (overlap) is used to classify overlap points that are not used in any other classes. These points are typically along the edge of the scan and are deemed to be unreliable or having poor accuracy and hence not to be used in the ground model. Breakline data are utilized to perform LiDAR classification for class 9 – water (see Section 6.4e). The manual classification is the most time-consuming and often the most expensive component of LiDAR processing. If application of the LiDAR data requires only bare-earth data, there is no need to request for additional classification of buildings, bridges, vegetation, etc. These data will be available in the “Unclassified” class (Class 1) and can be classified in the future if the need for these additional classes arises.