Conditional Random Fields for Airborne Lidar Point Cloud Classification in Urban Area

Over the past decades, urban growth has been known as a worldwide phenomenon that includes widening process and expanding pattern. While the cities are changing rapidly, their quantitative analysis as well as decision making in urban planning can benefit from two-dimensional (2D) and three-dimensional (3D) digital models. The recent developments in imaging and non-imaging sensor technologies, such as airborne Light Detection and Ranging (LiDAR) system, lead to a huge amount of remotely sensed data which can be employed to produce 2D/3D models. Although much of the previous researches have investigated on the performance improvement of the traditional data analyzing techniques, recently, more recent attention has focused on using probabilistic graphical models. However, less attention has paid to Conditional Random Field (CRF) method for the classification of the LiDAR point cloud dataset. Moreover, most researchers investigating CRF have utilized cameras or LiDAR point cloud; therefore, this paper adopted CRF model to employ both data sources. The methods were evaluated using ISPRS benchmark datasets for Vaihingen dataset on urban classification and 3D building reconstruction. The evaluation of this research shows that the performance of CRF model with an overall accuracy of 89.06% and kappa value of 0.84 is higher than other techniques to classify the employed LiDAR point cloud dataset.