Fast factorised back-projection for reconstruction of tomographic images from bistatic radar measurements

Abstract

The compelling evidence for climate change has increased the importance that we understand our environment and can properly model it to get factual prognoses. The carbon cycle is one part, of which there exists a large uncertainty in the land biomass, primarily forests. To improve our estimation abilities the BorealScat project is currently running, which consists of a tower based radar measuring the response from a forest. The data is used to reconstruct a vertical cross section image of the forest (tomographic image), thereby having another way to estimate biomass. The issue is that the processing of acquired data by the radar takes more time than the current measurement rate. Therefore this thesis investigates the possibility to implement a different reconstruction algorithm to decrease computation time. The tower’s full bistatic antenna geometry is simulated as well as a simpler monostatic approximation. For each of these a Global Back-Projection is implemented and used as a reference. The Fast factorised back-projection (FFBP) algorithm is implemented in the monostatic case, and partially implemented for the bistatic case (called BiFFBP). Simulated data and measured data were used for development and benchmarking of the algorithms. A speedup fo 50% - 70% is achieved in the monostatic approximation case, which is of the same order as the theoretical maximum; four times speedup . The partial results for the bistatic case are demonstrated, however, time constraints hindered the completion of the implementation. The conclusion is that the implementation of FFBP and BiFFBP show potential to help reduce the computational burden of the BorealScat project.