Portal de Periódicos da CAPES

Evaluating the Accuracy of iPhone Lidar Sensor for Building Façades Conservation

2024; Springer International Publishing; Linguagem: Inglês

10.1007/978-3-031-48715-6_31

2522-8722

Sahar F. Abbas, Fanar M. Abed,

Advanced Optical Sensing Technologies

Apple company recently started to include lidar sensor in iPhone 12, 13 Pro, and iPad Pro system devices. Lidar was introduced by Apple to improve camera focusing and support augmented reality (AR) applications. However, Apple lidar attracted app developers and users to start scan indoor and outdoor environments within the specified range configurations. Although Apple has not revealed the type of the 3D scanner used, some scholars believe Apple's lidar is based on direct time of flight (DToF) lidar. It is a mobile handheld scanner that provides greater flexibility and control when scanning close objects, and therefore, more affordable and much cheaper than TLS devices in very close ranges. This study aims to assess the positional accuracy of the new lidar sensor in the iPhone 12 Pro Max in outdoor environment, particularly for scanning building façades. Sitescape app was used for scanning objects using this device where it is primarily concerned with extracting 3D point clouds following specific scanning pattern. The study is carried on the façade of a selected building, where a number of artificial targets are affixed and observed using three different laser-based remote sensing techniques (total station, iPhone lidar, and TLS). The research study is divided into two stages where total station measurements used as reference ground-truth dataset. The first stage compares mobile DToF to static TLS sensor in terms of precision. The data are collected in the same range and environmental conditions to analyze the level of accuracy. The result showed that the accuracy of the iPhone lidar and TLS are comparable. The RMSE of TLS data was 3.44 mm, while the iPhone lidar delivered 4.89 mm. In the second stage, the sensor accuracy was tested over various ranges starting from 0.25 to 5 m to show range dependency on the positional accuracy. The accuracy shows convergent results; however, errors increased slightly as range increased. The data acquisition time was generally short and directly proportional to the scene size. Results are later analyzed and studied for future conservation applications.