An Integrated Machine Learning Approach for Automatic Highway Extraction from Airborne LiDAR Data and Orthophotos 45 À complexity of separation of roads from other ground points 46 À with the similar intensity value (Gong et al. 2010). In 47 À addition, LiDAR intensity values are affected by several 48 À factors such as surface reflectance, transmitted power, 49 À atmospheric attenuation, and incidence angle and range 50 À distance (Coren and Sterzai 2006). Apart from that, roads 51 À have missing data due to above obstacles (e.g., trees and 52 À vehicles), noise data (e.g., road markings), and different 53 À types of materials (e.g., asphalt and concrete). Therefore, 54 À incorporating color information from aerial photos is critical 55 À for accurate road extraction (Gong et al. 2010). 56 À Machine learning (ML) is a subfield of computer science 57 À and artificial intelligence based on the biological learning 58 À process. ML explores the study, design, and construction of 59 À algorithms to learn from the past and make predictions on a 60 À new set of data (Lary et al. 2015). ML covers main areas 61 À such as data mining, statistics, and software applications. It 62 À is a collection of a variety of algorithms (e.g., neural net-63 À works, support vector machines, self-organizing map, deci-64 À sion trees, logistic regressions, genetic programming, etc.). 65 À ML is an efficient empirical method for both regression and 66 À classification of nonlinear systems (Lary et al. 2015). Several 67 À methods based on ML were proposed for remote sensing 68 À applications and mostly for image classification (Butenuth 69 À