Accurate extraction and reconstruction of route corridor features from geospatial data is a prerequisite to effective management of road networks for engineering, safety and environmental applications. High quality road geometry and road side features can now be extracted from dense point cloud LiDAR data, recorded by modern day Mobile Mapping Systems. This valuable route network information is gaining the attention of road safety and maintenance engineers. Road points are needed to be correctly identified, classified and extracted from LiDAR data before reconstructing intrinsic road geometry and road-side infrastructure. In this paper, we present a method to automatically extract the road from terrestrial based mobile laser scanning system using the GVF (Gradient Vector Flow) snake model. A snake is an energy minimizing spline that moves towards the desired feature or object boundary under the influence of internal forces within the curve itself and external GVF forces derived typically from 2D imaging data by minimizing certain energy such as edges or high frequency information. In our novel method, we initialise the snake contours over point cloud data based on the trajectory information produced by the MMS navigation sub-system. The internal energy term provided to the snake contour is based on adjusting the intrinsic properties of the curve, such as elasticity and bending, whilst the GVF energy and constraint energy terms are derived from the LiDAR point cloud attributes. Our method primarily differs from the traditional snake models in initialisation and in deriving the energy terms from the 3D LiDAR data.