Image processing for a Moon lander


Remotely navigating a lunar module in the descent stage of a Moon landing can incur a critical loss in precision. Control signals transmitted from Earth are not received instantaneously by the lander due to the distance they must travel, preventing real-time operation of the lander. Image processing can be used to give the lander the ability to analyse a lunar surface, detect hazards and guide itself to a safe location.

A simulation of an image processing based autonomous landing system was developed to experiment with hazard detection algorithms. Images are retrieved in sequence from the Planet and Asteroid Natural scene Generation Utility (PANGU) developed by the University of Dundeeā€™s Space Technology Centre. Three different algorithms were incorporated which utilise techniques such as dynamic thresholding, morphological edge extraction and rotation invariant pair matching to identify craters and other hazardous terrain. Physics such as gravity and the effect of engine thrusters on the lander were also modelled to create a more realistic simulation.

An evaluation mode runs multiple simulations with different algorithms and sun elevations and their results are compared to surface height data to compute a hazard detection rate. The results are then output to a CSV file to ease analysis.


Jordan Sewell