Lichti, Derek D.Jones, Kent Douglas2024-02-292024-02-292024-02-15Jones, K. D. (2024). Automatic registration of imagery to mobile LiDAR maps (Doctoral thesis, University of Calgary, Calgary, Canada). Retrieved from https://prism.ucalgary.ca.https://hdl.handle.net/1880/11818110.11575/PRISM/43025Mapping in 3D that records geospatial data from platform-mounted sensors with digital twinning supports maintenance and future planning of civil infrastructure. Three-dimensional mapping is efficiently performed with a Mobile Mapping System (MMS). This research demonstrates camera-only registration of subsequently captured images to an MMS point-cloud for updating MMS datasets. The research resolves key issues with inherent resolution differences between MMS laser scanner point-clouds and camera images by bridging differences between MMS point-clouds and camera images using a synthetic camera image (SCI). SCI are used to determine the approximate pose or coarse register the camera image to the MMS point cloud. The SCI coarse registration precision is maximized by generating surfaces, interpolating intensity values, and reducing noise with a median filter. The SCI is processed with a median filter to remove salt-and-pepper noise from the generation methods while preserving edges. Edgeboxes are adapted to find similar features in both SCIs and camera images. These features are then passed through layers of a convolutional neural network to provide a feature descriptor for coarse registration. Real camera images (RCI) are processed to mitigate resolution differences with the SCIs. The RCI is downsampled to align with the spatial resolution of the SCIs. Robust features are used to register the RCI to the SCIs. SIFT is used for fine registration between RCIs and SCIs generated from dense point-clouds. Landmark features are used for registration of RCIs to SCIs generated from MMS point-clouds. The edgebox parameters require tuning to detect the same features in two disparate image sets. The fourth layer of AlexNet was found to provide the most ideal feature descriptor for registration between RCIs and SCIs. The approximate location of the RCI using SCIs as interpreters between RCI and MMS point-cloud detect scenes at a precision of 97% when changes are less than 20%, and foliage does not exceed 20% of the camera image. This novel application of landmark features aligns with camera-to-camera place recognition precision. The focal length and IOPs do not influence the precision of the registration because the registration precision does not change when different cameras capture the real images.enUniversity of Calgary graduate students retain copyright ownership and moral rights for their thesis. You may use this material in any way that is permitted by the Copyright Act or through licensing that has been assigned to the document. For uses that are not allowable under copyright legislation or licensing, you are required to seek permission.Mobile Mapping SystemsSensor FusionTwin ModelsRegistrationEngineeringdoctoral thesis