This dissertation thesis presents two approaches to forest roads extraction (object-oriented classification and edge detection). Data derived from mobile (MLS) and airborne (ALS) laser scanning and from aerial photogrammetry were used as a based material for both extraction methods. Data was collected on roads of interest located in the land register of the boroughs of Vigľaš and Detva. Digital terrain models from ALS and MLS were generated by the Inverse distance weighted method. Besides lasers technologies, ortophomosaics generated from digital aerial images in red-green-blue (RGB) and infrared channels (CIR) were used for extraction. DMT MLS and ortophotomosaics were generated with a pixel size of 0,15 m, DMT ALS with pixel size of 0,30 mThe first step of object-oriented classification was multiresolution segmentation. Two classes (road and non-road) were defined for segment assignation, as part of the classification process. A set of rules (height of object from normalised digital surface model, width, and length-width ratio) was defined as a main part of the road classification. Other method of forest roads extraction was edge detection based on the Canny algorithm. For comprehensive assessment of extraction methods, a quality index was created using a combination of indicators (overall accuracy, KHAT index, completeness, redundancy, and root mean square error). Besides extraction methods were also evalutaded type of road surface (asphalt, concrete, basalt cobble and other).The best results of quality extraction were obtained from laser scanning data in both methods.DMT MLS and DMT ALS acquired a value of 0,85 in object-oriented classification. The results of the edge detection method were 0,87 in the case of DMT MLS, and 0,60 for DMT ALS. The poorest results were from object-oriented classification with RGB (0,60) and edge detection with CIR (0,37).In addition to the evaluation of road extraction methods, we also verified the vertical accuracy of DMT from laser scanning data and digital aerial images (DAI). Digital terrain models were generated by IDW interpolation with a pixel size of 0,5 m (ALS, DAI), 0,1 m (MLS) for vertical accuracy analysis. Elevations extracted from DMT were compared with terrestrial data. Subsequently statistically analysis were performed. The lowest result of root-mean-square error mz =0,04 m were from ALS with asphalt road surface. The lowest result of the DMT MLS was mz = 0,08 m in case of basalt cobble road surfaceThe highest result of of root-mean-square error mz = 0,28 m was in case of digital aerial images.Methods of extraction presented in this dissertation have the potential to be used in creating and maintaining forest road inventories, which are necessary for effective forest planning and management.