Building and Infrastructure Defect Detection and Visualization Using Drone and Deep Learning Technologies

This paper presents an accurate and stable method for object defect detection visualization on building infrastructural facilities. uses drones cameras to collect three-dimensional (3D) point clouds via photogrammetry, orthographic or arbitrary views of the target objects generate feature images points’ spectral, elevation, normal features. U-Net is implemented in pixelwise segmentation using multiple images. was validated four applications, including on-site path detection, pavement cracking highway slope facade window detection. The comparative experimental results confirmed that with features has a better performance than separately each single feature. developed can implement different shapes, striped objects, thin recurring regularly shaped bulky which will improve accuracy efficiency inspection, assessment, management buildings