Scanning of Three-Dimensional Objects by Photogrammetry Methods Using LiDAR and Mobile Computing

Abstract

Three-dimensional models of objects are widely used in various fields, including science, construction, healthcare, and entertainment, making the task of creating such models highly relevant. The primary goal of this paper is to develop a mobile application which uses photogrammetry methods for capturing real-world three-dimensional objects or environments. The main advantage of photogrammetry is low hardware requirements with relatively high accuracy of the models obtained. Additionally, the application utilizes smartphone’s LiDAR sensor to enhance capture quality, especially for low-textured objects. The LiDAR sensor allows for precise distance measurement between the device and the object, which is crucial for accurately capturing the object’s size and position. To build 3D model of the object from a series of images, the application uses Object Capture API, available on iOS, iPadOS, and macOS operating systems. This API fully leverages the build-in GPU and Neural Processing Unit to build and tessellate the point cloud and generate the polygonal mesh model. The application was developed for iPhones that support LiDAR sensor, using Swift programming language, SwiftUI for the user interface, and RealityKit for Object Capture API. The app features three modes: object capturing, model reconstruction and model preview. To simplify the process of capturing the object, the app can automatically take photos of the object, provide the user with guidance and recommendations on optimal lighting conditions, camera positioning, and a preview of the point cloud. Once object capture is complete, the application transitions to the reconstruction mode, which uses captured photos and point cloud data. This process involves image alignment, detailed point cloud generation, polygonal mesh model generation, texture and normal map generation, and model optimization. After reconstruction, the user can obtain model in the USDZ file format and preview it using built-in system tools. Following these steps, two test models of a drum and a garden statue were built with satisfactory quality and accuracy, while maintaining the original size of the scanned objects. The resulting three-dimensional polygonal models can be successfully exported to various 3D editors and programs.