Forest roads are essential for forest management, forest harvesting, wood transportation, recreation, education, research, and forest protection. To meet these needs, forest road networks have been constructed in the northern forests of Iran. Forest road mapping especially over large and mountainous areas is time-consuming and expensive. Today, remote sensing data can be considered as an important tool for forest roads extraction. Therefore, in this research, LiDAR data and UltraCam images were applied in order to extract hidden forest roads. At the first step, noise points in the point cloud data were removed. Then, according to the Central Limit Theorem (CLT), the third statistical moment (amount of skewness) of the data was calculated and the non-ground points were eliminated. At this next step, a number of non-ground points were identified as ground points. In order to eliminate these errors, slope-based algorithm with a radius of 10 meters and a slope of 22 degrees was applied on the points obtained from the first step, these points were eventually removed and the ground points were extracted. Then, extracted ground points were converted to grid. Then the grid was converted to polygon based on the pixel density, by using the DTM as well as UltraCam aerial images, polyglots that were not related to the road were removed. Until this stage, the output was the roads that were not hidden by the forest canopy. Therefore, the hidden parts of the roads were extracted by applying slope-based algorithm with the radius of 10 meters and 65 degrees slope on the whole LiDAR points and interpolating the results by spline interpolation method. By connecting and modifying the polygons, 3m wide dirt roads and 2m wide skidding roads were extracted. The results are evaluated by comparing to manually acquired road data. The quality measures completeness, correctness and quality were 82%, 86% and 72%, respectively.