COMBINED MULTIPLE CLASSIFIED DATASETS CLASSIFICATION APPROACH FOR POINT CLOUD LIDAR DATA

Abstract. Airborne Laser scanners using the Light Detection And Ranging (LiDAR) technology is a powerful tool for 3D data acquisition that records the backscattered energy as well. LiDAR has been successfully used in various applications including 3D modelling, feature extraction, and land cover information extraction. Airborne LiDAR data are usually acquired from different flight trajectories producing data in different strips with significant overlapped areas. Combining these data is required to get benefit of the multiple strips’ data that acquired from different trajectories. This paper introduces an approach called CMCD “Combined Multiple Classified Datasets” to maximize the benefits of the multiple LiDAR strips’ data in land cover information extraction. This approach relies on classifying each strip data then combining the results based on the a posteriori probability of each class of the classified data and the position of the classified points.

Two datasets from different overlapped areas are selected to test the proposed CMCD approach; both are captured from different flight trajectories. A comparison has been conducted between the CMCD results and the results of the common merging data approaches. The results indicated that the classification accuracy of the proposed CMCD approach has improved the classification accuracy of the merged data-layers by 6% and 10% for the two datasets.