Electronic Thesis and Dissertation Repository

Thesis Format



Master of Engineering Science


Electrical and Computer Engineering

Collaborative Specialization

Planetary Science and Exploration


McIsaac, Kenneth


Science autonomy onboard spacecraft can optimize image return by prioritizing downlink of meaningful data. Martian polygonally cracked ground is actively studied by planetary geologists and may be indicative of subsurface water. Filtering images containing these polygonal features can be used as a case study for science autonomy and to reduce the overhead associated with parsing through Martian surface images. This thesis demonstrates the use of deep learning techniques in the classification of Martian polygonally patterned ground from HiRISE images. Three tasks are considered, a binary classification to identify images containing polygons, multiclass classification distinguishing different polygon types and semantic segmentation of polygon regions. Due to time and resource constraints, transfer learning is employed on state-of-the-art deep learning networks. Convolutional neural network model architectures are compared for the binary and multiclass scenarios. UNet was used for semantic segmentation. Overall, the models show promising results as a first filter method.

Summary for Lay Audience

Planetary spacecraft house instruments that collect large amounts of data for the purpose of scientific investigation. Once collected in space, this data is sent back to Earth for study. One restriction in the continued collection of data from these spacecraft is limited communication with Earth. This limitation is pushing the need for spacecraft to make decisions about what is meaningful data without human input. This automatic filtering of scientifically meaningful information from large datasets is referred to as scientific autonomy. It can be used to reduce time and resource intensive data parsing on Earth. Cracks on the surface of Mars can form interconnected networks. These networks are called polygonally patterned ground and are an area of active research for planetary geologists. In order to study these cracks, geologists must parse through collections of Martian images, which can be inefficient and time consuming. This work investigates methods for filtering images containing polygonally patterned ground using deep learning, a subset of machine learning. Three tasks are performed including determining if images contain polygons, what type of polygon and where in the image the polygon cracks are found. The model trained in identifying images as polygon or non polygon achieves good results. When considering different types of polygons, the deep learning model is able to correctly distinguish between polygon types but suffers between what is and what is not considered polygon in the experiment. Finally, in determining where in the image polygons are found, preliminary results are promising. This work shows that the use of deep learning for data filtering of spacecraft images is promising and further investigation into its use is merited.