White-paper-DT18-PPK
15Pages

White paper^ __ Jean Francois Aumont (1), Bastien Mancini (2). (1) Image Processing Manager, Delair-Tech, (2) Managing Director, Delair-Tech. achieving centimeter accuracy using drone-acquired data for mapping and topography purposes.

Vector and Payload DT18-3Bands PPK ... 4 DT26X-3BandsX PPK ... 4 GNSS/IMU APX-15 ... 5 Test field Comparison Method Global View Example of acquisition & processing Results Linear Flights ... 13 Cover Zones ... 13 Auteurs: Jean François Aumont / Bastien Mancini

The purpose of this document is to analyse the benefits of a precision GNSS/IMU combination on the accuracy of products obtained by photogrammetry (DSM, DTM, orthophotography, stereoscopic restitution, 3D model) with Delair-Tech UAVs. High accuracy makes it possible to work without Ground Control Points (GCPs), thus reducing overall costs. The following case study was carried out from December 2015 to March 2016 with our DT18-PPK & DT26XPPK models. Auteurs: Jean François Aumont / Bastien Mancini

2. Vector and Payload For this study, DT18-3bands PPK and DT26X-3bands XL PPK UAVs were used. 2 kg 2h of autonomy 1.8 m wingspan 1.2 m length 60 km/h cruising speed The payload consists of a precision IMU/GNSS Applanix APX-15 and an industrial grade camera, with the following specifications : ◊ ◊ ◊ ◊ ◊ ◊ Sensor installed in landscape mode in the drone’s direction of flight 3.45 microns (physical size of the pixels), ensuring good image quality 2448 x 2048 pixels Fixed, stable lens, with 12mm focal length. 1 s maximum frame rate At 150 m, coverage on the ground of 105 m x 88 m and a pixel on...

The payload it carries is the DT3Bands-XL PPK. This consists of a Sony Alpha 7R coupled with an Applanix APX-15 GNSS/IMU, with the following characteristics: ◊ 7392 x 4920 pixels ◊ Pixel pitch 4.9µm ◊ Focal length : 35mm ◊ 1.4s maximum frame rate ◊ At 150 m ASFC, coverage on the ground of 155 m x 103 m and a pixel on the ground of 2.1 cm 2.3 GNSS/IMU APX-15 Applanix has been recognized in the aerial mapping field for many years, and its APX-15 UAV model was chosen. The position and attitude characteristics, once the data have been post-processed with POSPac software, are specified as: ◊ 2...

3. Test fields Several test fields were used, and for each of them, control points were measured on the ground with a Trimble R6 (post-processed using the fixed base shifts) guaranteeing accuracy of the checkpoints to within 2-3 cm. Figure 3 – Example of one field with control points 4. Flights Two types of flights were tested: ◊ Cover zones: because they contain several axes, the whole system is more constrained. In order to resolve ambiguities (camera, lens, distortion and principal point of autocollimation, as well as the angles related to the set-up between the Applanix map and the...

5. Processing After the flight, the Applanix navigation data were post-processed using POSPac software in order to integrate the lever arms and calculate the precise trajectory (X,Y,Z,O,P,K). Two bases were used in order to evaluate the influence of the distance on corrections and results : ◊ A local base station (located at the Ground Control Station, therefore close to the flight area). Here a Trimble R6 receiver was used. ◊ A network station, in this case the CSTN beacon, accessible from https://applanixsmartbase. com and located 40km from the flight area (CSTN = Castanet-Tolosan, cf....

Two photogrammetry softwares applications were used: Pix4D and UASMaster. Auteurs: Jean François Aumont / Bastien Mancini

The usual process was used, with no GCPs. From Positions (X,Y,Z), Angles (O,P,K) and Pictures, knowing the theoretical camera parameters (focal length, sensor size, principal point, etc…), the aerial triangulation process estimates the “real” positions of the camera (X,Y,Z) as well as its “real” exterior orientations (O,P,K ). The photogrammetry software then estimates the DSM and generates a projection of the image points on the DSM to produce the orthophoto. At the end of the process, every point of each image has unique (XG,YG,ZG) coordinates. Camera parameters (from design)...

6. Comparison Method In the following tables we present the differences (RMS) between: ◊ Exterior Orientations of the camera as calculated by POSPac (X,Y,Z,O,P,K) and as estimated by the Photogrammetry software (X, Y, Z, O, P, K). This gives (ΔX, ΔY, ΔZ, ΔO, ΔP, ΔK) ◊ Coordinates (XG,YG,ZG) of specific points on the ground and their coordinates measured on the field by an independent method, in our case : a Trimble R6 receiver (Xref,Yref,Zref). These reference points are called “CheckPoints”. They should be distinguished from Ground Control Points (GCPs) in that they are not used during the...

7. Global view SD CARD Raw Data GNSS & IMU Lever arms (from design) Photogrammetry Softwares (Pix4D or UAS Master) Independant Trimble R6 GNSS Ground Receiver Camera Parameters (from Design) (Xref Yref Zref) X, Y, Z, O, P, K Camera parameters (estimated) The differences are calculated between the “green” and “blue” pyramids as seen here in Pix4D The differences are calculated between the “real” coordinates of the Point as measured on the ground (center of the target) and the coordinates as seen in the reprojection from the photogrammetry software (here : Pix4D) Auteurs: Jean François

8. Example of Acquisition & Processing For flight Number 1, acquisition took place on December 18th, 2015 in good weather conditions. 161 images were acquired. The raw data were processed in POSPac in order to get exterior orientations. The dataset was processed in Pix4D and UASMaster, without GCPs. With UASMaster, the following results were obtained in comparison with checkpoints measured on the ground by an independent GNSS receiver (Trimble R6): Check point errors For this dataset, with UASMaster, the accuracy (RMS) in X,Y,Z on the ground is therefore : With Pix4D, the following results...