Surveying technical Using UAVs as a survey tool by Craig Vorster, Global Vision When searching for a cost-effective, accurate aerial survey solution that needs to be available on demand, an unmanned aerial vehicle (UAV) is the ideal platform. This article explains how unmanned aerial vehicles (UAVs) work as a survey platform providing details on the photogrammetry process, volume accuracies, time savings, and how they can contribute to mine safety. G lobal Vision, a provider of innovative survey solutions is increasingly being asked to provide accurate aerial and hydro surveys for clients working in the mining, environmental, and government sectors. The company manufactures pre-programmable stand out features of modern day photogrammetry is the capability of dense point cloud classification, this allows us to classify point cloud data into the following classes: ground points, low vegetation, medium vegetation, high vegetation, buildings, roads, water and many more. This is a powerful feature when generating DTMs (digital terrain models) or DSMs (digital surface models) How does the software do it? There are two options of dense point cloud classification: automatic division of all the points into two tri-rotor copters, which are equipped with DJI auto-pilot systems and a 24-megapixel camera (see Fig. 1). An on-board GNSS provides estimates of the image position, which are transmitted in real time, through a 900 MHz data link, to a ground laptop set up at a base in a safe zone of the survey site. After the flight, the images are processed by photogrammetry software packages. What is photogrammetry? Photogrammetry is the science of making measurements from photographs, especially for recovering the exact positions of surface points. Fig. 1: Tri-rotor copter with six propellers developed in-house. With photogrammetry software such as Agisoft PhotoScan Professional and/ or Pix4D, thousands of aerial images captured by lightweight UAV or aircraft can be converted into geo-referenced 2D mosaics, 3D surface models and point clouds. Survey-grade accuracy With photogrammetry up to centimetregrade, lidar-like 3D precision is achieved from lightweight compact cameras and photogrammetry software packages. Photogrammetry software packages allow advanced support of ground control points for optimal geo-location. Dense cloud classification Improvements in photogrammetry technology is happening at a rapid pace, one of the latest, 24 Fig. 2: Orthophoto. PositionIT – August 2014 SURVEYING technical classes – ground points and the rest, and manual selection of a group of points to be placed in a certain class from the standard list known for lidar data. Dense cloud point’s classification opens ways to customise the Build Mesh step: you can choose what type of objects within the scene you would like to be reconstructed and indicate the corresponding point class as source data for mesh generation. For example, mesh reconstruction based on ground points only, allows the exporting of the DTM as opposed to the DSM, based on the complete dense point cloud. The photogrammetry software uses the following parameters to control automatic ground point’s classification procedure: z Max angle (deg): Determines one of the conditions to be checked while testing a point as a ground one, i.e. it sets the limitation for an angle between the terrain model and the line to connect the point in question with a point from a ground class. In fact, this parameter determines the assumption for the maximum slope of the ground within the scene. z Max distance (m): Determines one of the conditions to be checked while testing a point as a ground one, i.e. it sets the limitation for a distance between the point in question and the terrain model. In fact, this parameter determines the assumption for the maximum variation of the ground elevation at a time. z Fig. 3: Point cloud of the digital terrain model. Fig. 4: Dense cloud classification. Cell size (m): Determines the size of the cells for the point cloud to be divided into as a preparatory step in ground point’s classification procedure. Cell size should be indicated with respect to the size of the largest area within the scene that does not contain any ground points, e.g. building or close forest. In case the result of the automatic dense cloud classification is not acceptable the procedure can be re-run using adjusted parameters (for example, if some on-ground objects like stones and small bushes were classified as ground points it is reasonable to reduce maximum angle and maximum distance parameter values). Dense cloud points may be classified manually, but the same workflow also allows users to reset the classification results for the selected dense cloud points. PositionIT – August 2014 Fig. 5: Point cloud. 25 SURVEYING technical Using UAVs for mining Frequently asked questions relating to accuracy assessment, time savings and safety regarding the UAVs for mining purposes include the following: z How long does the throughput from flight to final products of a UAV survey take? z What is the accuracy of the volume calculated from the digital surface model? z How can UAVs contribute to the safety of surveyors? Answers to these questions are provided below by detailing Global Vision’s experiences when working on various mines around South Africa. From flight to final products Global Vision began using Pix4D software in January 2013 and its abilities were examined by capturing the 66 ha of the Nooitgedacht mine, Fig. 6: UAV survey system. which is located 10 km outside the small town of Northam in Limpopo, South Africa. This narrow open-cast mine is currently being exploited by Andru Mining. During the aerial survey, 721 images were taken with a ground sample distance (GSD) of 2,24 cm, an along track overlap of 80% and an across track overlap of 40%. It took 160 minutes on a standard desktop PC to extract over 6,3-million key points, to calculate the values of their features and their sub pixel location and to assign to every key point its conjugates in the overlaps. The step resulted in 2,4-million 3D points suitable for the bundle block adjustment (BBA). The root mean square error (RMSE) after BBA was 0,16 pixel or 0,4 cm. To create a high density DSM 18-million key points were extracted which took about 250 minutes. These points were automatically filtered and interpolated to generate a DSM consisting of Fig. 7: UAV survey team. 6-million points with a GSD of 2,24 cm. The generation of the were measured by GNSS along the and 2,5 cm in planimetry. That is less borders of the Nooitgedacht site. Five than two times GSD, demonstrating GCPs were used in the BBA procedure the robustness of the BBA on consumer and two for verification purposes. grade cameras. For assessing UAV After BBA the residuals of the 5 GCPs DSM accuracy, a ground truth dataset mining information were generated. appeared to be less than 1 cm, with a of 3500 grid points with spacing 10 m RMSE of 0,83 cm. Comparison of the were collected by RTK GNSS across Accuracy assessment coordinates of the two verification GCPs the Nooitgedacht mine. Comparison For accurate geo-referencing purposes seven ground control points (GCPs) with those computed from the images of the heights of individual points of revealed a RMSE of 4,8 cm in height the GNSS DSM with the heights at the orthomosaic by combining the original images with the DSM took about 240 minutes. From these outputs reports on volume, mine planning, mine rehabilitation and other valuable 26 PositionIT – August 2014 SURVEYING technical same location in the UAV DSM revealed a RMSE of 4,9 cm, which is a similar value to the 4,8 cm obtained from the verification GCPs mentioned earlier and validates the accuracy of the DSM generation. Using Bentley InRoads the volumes of the UAV DSM and the GNSS DSM were computed and the difference was within 5%. Without doubt, the accuracy of individual GNSS points is higher than the accuracy of UAV points. However, the number of height points in the UAV DSM is nearly 2000 times larger than the number in the GNSS DSM. A spacing of a few centimetres instead of 10 m will handle surface fluctuations especially when shapes are complex resulting in a higher accuracy of volume computation even when the accuracy of the individual Fig. 8: Orthomosaic of Nooitgedacht mine. heights is less. Time savings Ruighoek is a South African chrome mine. To assess time and thus cost savings the UAV survey of their open pit mine was examined. The 3,2 km2 site has been surveyed many times by conventional means. A team consisting of two to four surveyors needs seven to ten days of field work for collecting 3D coordinates of terrain points on a 10 m grid using GNSS rovers. The in-office post-processing of the points collected takes one to three days. With a UAV the same area could be captured in one day by one operator carrying one UAV and one GNSS rover for collecting 10 to 20 GCPs along the border of the site. Processing of the 3216 images which is done automatically for 90% and the generation of end products requires PositionIT – August 2014 Fig. 9: 3D view of the digital surface model of Nooitgedacht mine. 27 SURVEYING technical the Riegl VUX-1. This lightweight, rugged and compact laser scanner, meets the challenges of emerging survey solutions by UAS/UAV/RPAS, gyrocopter, and ultra-light aircraft, both in measurement performance as well as in system integration. With regard to the specific restrictions and flight characteristics of UAS, the Riegl VUX-1 is designed to be mounted in any orientation and even under limited weight and space conditions. Modest in power consumption, the instrument requires only a single power supply. The entire data set of an acquisition campaign is stored onto an internal 240 GB SSD and/or provided as real-time line scan data via the integrated LAN-TCP/IP interface. It captures 10 mm survey-grade accuracy with a scan speed of up to 200 scans/ second and a measurement rate of up to 500 000 meas./sec (@ 550 kHz PRR & 330° FOV). The operating flight altitude of the Riegl VUX-1 is up to over 1000 ft. and has a field of view up to 330° allowing for practically unrestricted data acquisition, regular point pattern and perfectly parallel scan lines. The cutting edge technology provides echo signal digitisation, online waveform processing, multipletime-around processing and multiple target capability. Hydro survey systems Unmanned survey systems are not Fig. 10: Ruighoek mine. just available for aerial surveying, systems have also been developed for hydrographic surveying. Global Vision’s Hydro Survey System is a start to end shallow water hydrographic survey solution. It consists of a remote control survey boat, semi-autopilot system, ground station software and laptop, Trimble GPS system and SonarMite echo sounder system. The vessel is constructed from ultra-light, yet strong materials such as carbon fibre and aluminium, and the twin hull multi-powered vessel can withstand harsh hazardous toxic environments. Accurate shallow water hydrographic surveys can be achieved using this system. The vessel is remotely controlled to follow pre-determined grid and is equipped with a semi-autopilot system. The semi autopilot system is a combination of a variety of sensors and wireless data links. The data captured by the sensors is fed back to the ground station software and laptop, wirelessly and live. The vessel is equipped with a GPS and an echo sounder system that collects 3D surface points of the dam floor, whilst driving along the pre-programmed grid. With the 3D surface points, 3D model profiles of dams, contour plans, water volume reports and many more can be generated. Mining companies can use the solution for their tailings dam and shallow water survey requirements. Conclusion The company is continuing with their research and development programme into providing remotely controlled surveying technologies for both aerial and hydrographic surveying. Contact Craig Vorster, Global Vision, Tel 012 345-5330, [email protected] 28 PositionIT – August 2014