Effective LiDAR data classification by row data and parameter analysis framework

Airborne Light Detection and Ranging (Li-DAR) technology has provided an accurate and efficient way to obtain topographic information in three dimensions. The objective of this study was to investigate the effects of ground and non-ground classification errors in different terrain categories to obtain a high accuracy terrain class. For this purpose, the first priority was to clean the outliers point completely. Adaptive Triangulation Irregular Network(ATIN) method was used in both processes. For implementation, a heterogeneous and mountainous terrain covering the provinces of Artvin, Borçka and Ardanuç in Eastern Anatolia Region of Turkey has been selected as the study area. In this area, 11 different sites were identified in 4 study areas in different terrains. Here, RDAF approach was proposed which improves the performance of the ATIN method. As a result of the process, the performance of the method was examined by calculating the errors of indicating inlier points incorrectly as outlier (Type-I) and indicating outlier points incorrectly as inlier (Type-II). In addition, different results were obtained for different terrain classes with different iteration angle parameters in the Earth classification. In general, there was an increase in comparison with naked land in both Type-I and the Type-II error percentages in areas with detailed objects on the surface and with dense surface coverage, areas with sparse vegetation, as well as artificial objects with complex structure. Type-II error percentages were determined to be lower for tall objects like electric poles, long trees etc., in comparison with other samples.