Abstract
High resolution and high accuracy terrain mapping can provide height change detection for studies of erosion, subsidence or land slip. A UAV flying at a low altitude above the ground, with a compact camera, acquires images with resolution appropriate for these change detections. However, there may be situations where different approaches may be needed, either because higher resolution is required or the operation of a drone is not possible. Pole photogrammetry, where a camera is mounted on a pole, pointing to the ground, is an alternative. This paper describes a very simple system of this kind, created for topographic change detection, based on an action camera. These cameras have high quality and very flexible image capture. Although radial distortion is normally high, it can be treated in an auto-calibration process. The system is composed by a light aluminium pole, 4 meters long, with a 12 megapixel GoPro camera. Average ground sampling distance at the image centre is 2.3 mm. The user moves along a path, taking successive photos, with a time lapse of 0.5 or 1 second, and adjusting the speed in order to have an appropriate overlap, with enough redundancy for 3D coordinate extraction. Marked ground control points are surveyed with GNSS for precise georeferencing of the DSM and orthoimage that are created by structure from motion processing software. An average vertical accuracy of 1 cm could be achieved, which is enough for many applications, for example for soil erosion. The GNSS survey in RTK mode with permanent stations is now very fast (5 seconds per point), which results, together with the image collection, in a very fast field work. If an improved accuracy is needed, since image resolution is 1/4 cm, it can be achieved using a total station for the control point survey, although the field work time increases.
Highlights
High resolution and high accuracy terrain mapping can provide data for the detection of height changes due to several reasons, as erosion, or mass movements (James and Robson, 2012)
The processing methodology designated as Structure from Motion (SfM) is essentially a method of automatic triangulation, of images not necessarily acquired in a regular manner, as in conventional photogrammetry
The following step is the generation of a dense point cloud, by multi-view stereo matching, which leads to a digital surface model (DSM) and an ortho-rectified image mosaic
Summary
High resolution and high accuracy terrain mapping can provide data for the detection of height changes due to several reasons, as erosion, or mass movements (James and Robson, 2012). Provided that ground control exists with that positional accuracy, current processing software, based on SfM algorithms, will allow for the extraction of very accurate and detailed surface models. Images can be acquired in time lapse mode, with rates of one or more images per second, for thousands of continuous images (Digital Photography Review, 2014). Another interesting aspect of action cameras is the large field view, which can cover large areas, when compared to other compact cameras. Ground control points can be surveyed with centimetric accuracy in very short times of a few seconds. Structure from motion processing was done with Photoscan (Agisoft, 2016), in order to achieve a digital surface model (DSM) and an ortho-mosaic
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
