Classification of Full-Waveform LiDAR Data in Urban Areas by Combining Physical and Geometrical Features

In the last two decade the use of Aerial Laser Scanner (ALS) or LiDAR (Light Detection and Ranging) sensor in geomatics engineering and surveying application has augmented significantly. The main reason of the mentioned phenomenon is the reliability and accuracy of the data obtained by LiDAR sensors. The output of LiDAR is unclassified 3D point cloud. Classification of the LiDAR point clouds in different and distinguished classes is the first step in applying such data in different geomatics applications. The purpose of this article is to classify Full- Waveform LiDAR data with the compilation of geometric and physical parameters of each point in the point cloud. First of all the geometrical parameter  is extracted from raw 3D coordinate of the points. This geometrical parameter is the calculation of the relational association of the point in the construction of a plane with the help of 3D Hough transform. The feature vector also includes physical features that exclusively belong to full-waveform LiDAR. These features are amplitude of the pulse, width of the pulse and the number of the returned pulse. After the construction of the feature vector for each point, the next step is to classify the point cloud into three classes; bare earth, building and vegetation with the utilization of Support Vector Machines classification method. The final step is accuracy assessment of the classification method. The results are promising; 81.04% Overall Accuracy, 0.69 Kappa Coefficient and 79.21% Average Accuracy. Keywords:Full-waveform LiDAR, Support Vector Machine Classifier, Point Cloud Processing, 3D Hough Transform, Urban Areas.