Damage inspection of concrete structures is necessary to prevent disasters and ensure the safety of premises such as buildings, sidewalks, dams and bridges. Cracks are among the most prominent damages in such structures. In this research, a computer vision and machine learning-based solution for identifying and modeling cracks in concrete structures from high-resolution images captured by a stereo camera is proposed. First, using deep learning-based semantic segmentation networks trained on a custom-dataset, crack pixels are identified. Moreover, techniques for improving the accuracy of such networks are developed and evaluated. Second, modifications are applied to the stereo camera’s calibration model to ensure accurate parameter estimation. Finally, two 3D reconstruction methods are proposed, one of which is based on detecting the dominant structural plane surrounding the crack, while the second method focuses on matching the crack pixels across two images. As a result, a 3D model of cracks is produced, from which the cracks' size and other geometric characteristics can be deduced. The solution proposed in this thesis can be used by professionals to regularly inspect concrete structures and make timely maintenance decisions.