Three Photogrammetric Software Are Compared Using Four UAV Datasets

Abstract

Photogrammetric software utilizing Structure from Motion (SfM) have been extensively used with drone-acquired imagery. SfM software, such as Agisoft PhotoScan and Pix4D, use computer vision techniques to solve for aero-triangulation. This study attempts to compare these two software with traditional software such as Trimble Inpho using a dense ground control field. Since AgiSoft and Pix4D show increased robustness in their analyses and provide visually appealing three-dimensional (3D) models, they are used by many. Nonetheless, the details of these software, predominantly through statistical analyses of results, are not well understood. On the other hand, careful selection of initial settings combined with a rigorous check and analysis of the output are prerequisites for obtaining the best possible results in photogrammetry. For the companies which are producing photogrammetric products, bringing Agisoft and Pix4D into play might result in significant cost savings over time. Nevertheless, to make this major move, these software must be vetted. Thus, in this project, these three software (Agisoft, Pix4D and Trimble Inpho) were analyzed using four Unmanned Aerial Vehicle (UAV) datasets, and the results were compared in terms of precision and accuracy. It is found out that Inpho results were the least precise and accurate among these software. This is because Inpho is not sensitive to the biases created by reverse strips. Hence, the same data sets were processed without reverse strips and the errors got smaller, and systematic bias was removed. In addition, Inpho results deteriorated with flying height. Additionally, 90 m real-time network solutions showed the same performance as 90 m solutions without real-time network solutions.