Image georectification and feature tracking toolbox: ImGRAFT

Abstract

Abstract. The use of time-lapse camera systems is becoming an increasingly popular method for data acquisition. The camera setup is often cost-effective and simple, allowing for a large amount of data to be accumulated over a variety of environments for relatively minimal effort. The acquired data can, with the correct post-processing, result in a wide range of useful quantitative and qualitative information in remote and dangerous areas. The post-processing requires a significant amount of steps to transform images into meaningful data for quantitative analysis, such as velocity fields. To the best of our knowledge at present a complete, openly available package that encompasses georeferencing, georectification and feature tracking of terrestrial, oblique images is still absent. This study presents a complete, yet adaptable, opensource package developed in MATLAB, that addresses and combines each of these post-processing steps into one complete suite in the form of an "Image GeoRectification and Feature Tracking" (ImGRAFT: http://imgraft.glaciology.net) toolbox. The toolbox can also independently produce other useful outputs, such as viewsheds, georectified and orthorectified images. ImGRAFT is primarily focused on terrestrial oblique images, for which there are currently limited post-processing options available. In this study, we illustrate ImGRAFT for glaciological applications on a small outlet glacier Engabreen, Norway.

PyTrx: A Python toolbox for deriving velocities, surface areas and line measurements from oblique imagery in glacial environments

“…Feature-tracking can be conducted by creating and tracking exclusively between each image pair in an image set (e.g., Messerli and Grinsted, 2015), or by tracking continuously through the image set using the same pixel-intensity features (e.g., James et al, 2016). Whilst tracking continuously enables the calculation of cumulative feature displacements, tracking between each image pair reduces the risk of error propagation from false matches.…”

“…When tracking between an image pair, the two images have been referred to in many ways, such as 'image A' and 'image B' (e.g., Messerli and Grinsted, 2015), and the 'reference' and 'destination'/'search' images (e.g., Ahn and Box, 2010;James et al, 2016). The terms reference image and destination image will be used subsequently; the reference image being where 5 points are seeded or templates are defined, and the destination image being where a point/template is matched to.…”

“…Glacial features which provide good tracking points/templates, such as debris features, can evolve over time which introduce additional 20 displacements (e.g., How, 2013). This can be especially limiting for area-based tracking (e.g., Messerli and Grinsted, 2015).…”

“…Feature-based registration methods are more commonly used for glacial environments due to large variations in lighting and 30 glacier surface evolution over time, especially over long-duration sequences (e.g., Messerli and Grinsted, 2015;James et al, 2016). Feature-based registration aligns the images by tracking static feature points.…”

“…were the first toolboxes that were made publicly available and contain all the processes needed to obtain velocities from terrestrial time-lapse imagery in glacial environments (Messerli and Grinsted, 2015;James et al, 2016). Both are Matlab-based toolboxes using algorithms from the Computer Vision System toolbox, developed specifically for glaciological applications.…”