InSARTrac: Monitoring surface displacements in 3D by combining Computer Vision and InSAR

InSARTrac, Monitoring surface displacements in 3D by combining Computer Vision and InSAR
Interferometric Synthetic Aperture Radar Feature Tracking (InSARTrac) is a highly accurate three-dimensional (3D) surface displacement monitoring method arising from the combination of an Interferometric Synthetic Aperture Radar (InSAR) and Feature Tracking (FT). InSAR acquires one-dimensional (1D) displacement measurements with sub-millimeter accuracy independent of weather and illumination by analyzing microwaves with advanced mathematical methods. FT is a computer vision method to track displacements of image features in image series with sub-pixel accuracy. By merging both systems, InSARTrac acquires high accuracy 3D vectors without requiring signalized areas.
Surface displacement vectors in 3D are far superior to 2D or 1D vectors, since they allow the unbiased acquisition of total displacements and give the displacement direction. The displacement direction can provide a deeper insight into movement mechanisms, such as defining the intersection line of two discontinuity sets in a wedge slide or by allowing to split the vector into individual components, such as melting and ice flow of a glacier. Commonly, Total Stations are used to acquire 3D displacement measurements, but they only measure discrete points. In difference, pointcloud based measurements generally do not require assumptions or interpolations within the monitored area because of their high point density, but they require a computationally expensive pointcloud registration, if they are used for monitoring. Hence, they are sparsely applied for high temporal resolution monitorings. InSARTrac provides such areal measurements with high repeatability and a temporal resolution faster than ten minutes.
A research program was developed to verify InSARTrac's principle, including prototype tests in controlled laboratory experiments, controlled field experiments, and uncontrolled field experiments. Under these conditions, the system's accuracy, applicability, and hardware performance were evaluated. The controlled tests included measurements of a displaced target using a micrometer-resolution translation stage. These were executed at ranges of 13 and 150~m. The uncontrolled experiments were conducted at the Pasterze and the Mölltal Glaciers in Austria for having continuous and, to some extent, predictable surface displacements. At the Mölltal glacier, two setup positions were employed to evaluate the system accuracy by applying vector comparison.
The controlled experiments confirmed InSARTrac's working principle and achieved accuracies in the micrometer range. The controlled and uncontrolled outdoor tests revealed an InSAR accuracy of about 0.1~mm and an FT accuracy of about 0.2~pixels (prototype: 4.6~ppm), being similar to modern Total Stations. In the uncontrolled tests, the 3D vectors were resolved into individual components, such as melting, ice flow, and pure surface displacements. The high temporal resolution also allowed correlations between the displacements and atmospheric conditions, such as temperature and global radiation.
This proof-of-concept research utilized a single camera and the fiel of view was therefore limited. InSARTrac's field of view could be increased using multi camera systems or a robotic system to reorient the camera to monitor multiple positions, theoretically allowing 360° monitoring. Combined with the InSAR monitoring, InSARTrac facilitates the definition of 3D movement characteristics of entire or individual parts of displacing masses and correlations to external influences, such as temperature, precipitation, or geologic boundary conditions. InSARTrac is considered a novel monitoring system that synergistically combines recent advantages in computer vision and terrestrial InSAR technology.