This study semi-automatically detects road lines using object-based classification methods from orthophoto produced from UAV images. Three studies were carried out on images with visible and infrared wavelengths. The study's major aim is to examine the effect of spectral bands and different areas on classification accuracy. First, orthophotos were produced from UAV images of different areas and an object-based classification algorithm was used to detect roads. Then, statistical comparisons were made and, user accuracy in the three study regions ranged from 85 to 91%. Overall accuracies were calculated between 0.819 and 0.889, and the results were within the confidence interval.