<p>Cross-section profile of roads is a critical design element which ensures surface drainage and provides stability to high-speed vehicular movement. Accurate assessment of cross-slope in operational period of road is necessary. Mobile LiDAR survey is observed to capture road profile data and researchers attempted to extract cross-section profile from this data. However, this is challenging when the road surface is rough, noise exists due to the presence of vegetation and vehicle trajectory is unavailable. In this work a novel approach for extracting the actual cross-section profile of highways, where noise due to vegetation exists and vehicle trajectory is not available, is proposed. The Lidar data is integrated with google earth image to extract the road centre line which can be used as a reference for cross section extraction in absence of vehicle trajectory. The cross-section estimation is then completed in four steps namely, cuboid segmentation for extracting point cloud of the region of interest, cross-section cleaning for noise removal, identification of points of inflection or <i>knots</i> using multivariate adaptive regression spline and estimation of cross-section profile. A two-stage filter, elevation based and best fit plane based on cost function, is proposed to remove noise from cross-section. The proposed cleaning de-noising is observed to be more effective in noise removal for locations with greater noise levels. Cross-slope extracted from the proposed approach is validated with cross-slope estimated from total-station survey for 20 sections of two 2-lane Indian highways and the difference is observed in the range of 0.13 to 0.21% with median value of 0.13%.</p>