Implementing High Dynamic Range (HDR) Photography to Improve 3D Laser Scanning Point Cloud Visualization and Segmentation

Graduation date: 2016Access restricted to the OSU Community, at author's request, from September 2, 2015 - September 2, 2016This research explores several novel approaches to improve visualization and segmentation\ud of point clouds acquired with 3D laser scanning. 3D laser scanning is used\ud in a wide variety of applications including surveying and mapping, transportation asset\ud management, facilities management, building information modeling, crime scene investigations,\ud cultural heritage and geologic instigations. Although most 3D laser scanners can\ud co-acquire photographs during the scanning process, this information is mostly being used\ud to color the point cloud for visualization purposes, rather than utilizing it as a quantitative\ud source of information to enrich point cloud processing.\ud Two key limitations of most digital cameras are their restrictions in capturing the\ud full luminance range of the scene and a full field of view. These limitations result in overexposed/\ud underexposed pixels and lead to inconsistencies between adjacent photographs\ud when linking them with scans. High Dynamic Range (HDR) photography is a promising\ud solution to these problems.\ud HDR combines multiple images with varying exposure levels to cover the dynamic\ud range in the scene. However, HDR presents unique challenges of its own. Inconsistencies\ud in object positions between images can result in ghosting effects in the composite image. Additionally, it can still be difficult to properly map the color tones to ensure details are\ud properly displayed in regions of extreme lighting.\ud This research presents novel methods to address these challenges so that point cloud\ud data can be segmented more effectively with the aid of the HDR photographs. First,\ud a new method is developed to detect these moving objects in digital images without\ud requiring a camera response function, supervision of objects in the scene, or selection\ud of a reference image. The proposed approach yields effective results with a Matthews\ud correlation coefficient of 0.82 on HDR photographs from both indoor and outdoor scenes.\ud Second, a new tone mapping operator is invented with the intent of preserving\ud detail in both dark and bright areas of the images. It utilizes geostatistical variography to\ud reproduce optimal color values. The approach proved to be highly versatile and worked\ud consistently well in producing realistic images for a variety of scene types.\ud Third, this research contains two novel approaches for point cloud segmentation\ud utilizing computer vision algorithms. The first approach applies a superpixel algorithm\ud for clustering the colorimetric data. Then, each superpixel is classified as either "ground",\ud "wall", or "vegetation" based on its corresponding normal vector. The proposed method\ud effectively generates accurate segmentation results when compared to ground truth. The\ud second approach uses both colorimetric and geometric data to segment lidar point clouds.\ud The fixed angular acquisition pattern of the scan permits creation of 2D structured\ud panoramic image maps (PIMPs) representing various subsets of the data including normal,\ud intensity, range, and RGB color. An image segmentation algorithm is applied to\ud each PIMP, and then the union of the segmented PIMPs is mapped back to the 3D point\ud cloud. Compared with RANSAC, the proposed approach has significantly higher perceptual\ud segmentation and efficiency. The use of HDR photographs and corrected laser\ud intensity significantly improves the segmentation by the average MCC value of 0.87