FORMAP: A SIMPLE AND FAST PHOTOGRAMMETRY FRAMEWORK FOR DIRECT GEO-REFERENCING SYSTEMS

The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XLII-2/W17, 2019 6th International Workshop LowCost 3D – Sensors, Algorithms, Applications, 2–3 December 2019, Strasbourg, France FORMAP : A SIMPLE AND FAST PHOTOGRAMMETRY FRAMEWORK FOR DIRECT GEO-REFERENCING SYSTEMS G. Truong Nguyen1, ∗, N. Seube2 1 Geown France, 13 Avenue de l’Europe, 31520 Ramonville St-Agne, FRANCE, 2 3051 rue du Plateau, J7V 8P2 Vaudreuil-Dorion, Québec, CANADA - Commission II KEY WORDS: Photogrammetry, Direct georeferencing, Orthophoto, UAV mapping ABSTRACT: This paper presents FORMap (Fast Ortho Mapping) a simple, automatic, fast and accurate commercial photogrammetry processing software for Unmanned Aerial Vehicles (UAV) imagery equiped with Direct Georeferencing (DG) technology. DG technique allows user to directly geo-reference the acquisition without the use of Ground Control Points (GCP) by providing image external orientation (EO) parameters in a mapping frame. However, it requires a sensor of relatively high quality to provide an accurate EO with each image shot, which is somehow limited by the light weight of UAV payloads. FORMap makes use of EO information delivered by DG as an a priori information to accelerate its photogrammetric processing. We present the functionalities and some application of FORMap in the field of UAV mapping. We evaluate its accuracy and its robustness on several datasets. Test result shows that FORMap is robust for 3D scene reconstruction despite of inaccuracies of DG input data. It is also faster than standard digital photogrammetry solution based on SfM (Structure from Motion) approach and can provide orthophotos and dense point cloud in quasi real-time. 1. INTRODUCTION 1.1 Direct geo-referencing in photogrammetry Low-cost UAV (Unmanned Aerial Vehicles) mapping systems can provide aerial images with external orientation (EO) parameters thanks to their integrated GNSS (Global Navigation Satellite System) and IMU (Inertial Measurements Unit) sensor. These systems — called direct geo-referencing (DG) — have a major advantage in determining the coordinates of ground points by direct triangulation of bundles. Using DG information, we can geo-reference the scene without the use of GCPs (Ground Control Point). However, inaccurate EO parameters given by small UAV platform could be a problem if we use them directly for 3D scene restitution. Aerial triangulation (AT) can then be used to enhance the precision of EO determination, especially in the case of non-post processed of navigation data. The combination between AT, DG and optimized strategies in tie-points extraction, dense matching and orthophoto rectification significantly accelerates the photogrammetric processing time. The development of UAV (Unmanned Aerial Vehicles) technology in the recent years brings us a large choice of drones with DG technology (Eisenbeiß, 2009), from the professional grade such as Microdrones systems (Mian et al., 2015, Pérez et al., 2013, Sauerbier et al., 2011) to a popular low-cost solution such as DJI Phantom platform (Taddia et al., 2019, Peppa et al., 2019). However, the accuracy of geo-referencing depends on the quality of integrated sensors, on the time synchronization and finally on boresight and lever arm calibration. (Gabrlik et al., 2016). DG data can be embedded into image metadata (by geotagging). Exchangeable Image File Format (EXIF) (Association ∗ Corresponding author et al., 2010) standard is widely used for this purpose. It is extended by Adobe with Extensible Metadata Platform (XMP) standard to provide more flexibility to add user-defined tags (Ball, Darlington, 2007, Specification, 2005) (e.g. embedding image’s attitude, gimbal orientation). Most of commercial digital photogrammetry solutions such as Pix4D or Agisoft automatically extract these data from image metadata for their processing needs. Software can get image position (latitude, longitude, altitude) and attitude (roll, pitch, yaw) and use of them as an absolute EO parameters for georeferencing purposes or to optimize the image network geometry. Post processing of DG navigation data can enhance the precision of absolute geo-referencing (Rabah et al., 2018). With a rigorous lever-arm calibration between GNSS and the camera optical center, absolute geo-referencing could reach a centimetric precision (Daakir, 2017), even in some difficult overlapping configurations such as the one found in corridor mapping applications (Zhou et al., 2018). 1.2 Aim of the paper This paper aims at introducing FORMap R : a commercial photogrammetry solution designed for UAV acquisition with DG technology. FORMap is developed by Geown, and focus on the simplicity of use and high processing speed. From a direct geo-referencing image dataset, without any user interaction, FORMap computes 3D dense point cloud and geo-referenced orthophoto. FORMap accelerate tie-points extraction process and bundle block adjustment by exploiting a priori EO data. Dense point cloud computation is accelerated based on initial depth estimated from a sparse tie-point surface model. Test results show that FORMap response time is compatible with quasi real-time applications of UAVs. This contribution has been peer-reviewed. https://doi.org/10.5194/isprs-archives-XLII-2-W17-385-2019 | © Authors 2019. CC BY 4.0 License. 385 The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XLII-2/W17, 2019 6th International Workshop LowCost 3D – Sensors, Algorithms, Applications, 2–3 December 2019, Strasbourg, France • Refined image EO parameters, • Refined camera IO parameters, • 3D dense point cloud, • Geo-referenced orthophoto without the need of GCPs. 2.2 Figure 1. A Microdrones md4-1000DG system with Sony RX1-RII camera and APX-15 direct geo-referencing sensor. UAV photogrammetry applications such as first response in disaster scenario, validity of data usability, quick preview on the field or a sample large scale survey require a fast map/3D models with less user interaction, at the price of degrading slightly the absolute direct georeferencing accuracy and quality of final result. FORMap is designed for these purposes. The next section presents the photogrammetry data processing done within FORMap. The last section describes some test flight mission from several professional UAV manufacturers, from the data acquisition protocol to the processing workflow with FORMap, and the comparison of accuracy and processing time with Pix4D. Processing steps FORMap does not employ a SfM (Structure from Motion) approach, which computes image EO and IO in a common 3D coordinate frame, by an incremental reconstruction process starts from an initial image pairwise or triplet (Bianco et al., 2018, Rupnik et al., 2017), or by a hierarchical approach (Toldo et al., 2015). FORMap achieves a robust BBA (Bundle Block Adjustment) by using initial EO parameters (...truncated)