A significant fraction of captured satellite images are not usable for certain applications due to high cloud cover percentage. To address this problem, cloud detection and dynamic attitude correction algorithms are often used in tandem in order to increase the likelihood of capturing cloud-free images. Researchers from the STAMINA4Space Program and PhilSA developed MATA-Cloud, a software for evaluating the effectiveness of cloud detection and dynamic attitude correction algorithms for earth observation satellites. MATA-Cloud explores two key experiments. First is evaluating different image processing and machine learning-based approaches to detect cloud cover. The second is exploring dynamic attitude correction to minimize the effect of cloud cover on captured images. The cloud detection algorithms were evaluated based on their accuracy, latency, and memory consumption. The results show that Otsu, a traditional thresholding algorithm for image segmentation is the fastest at cloud detection. On the other hand, deep learning models are more accurate and adaptable to different datasets. MATA-Cloud was also demonstrated to be an effective testbed for evaluating dynamic attitude correction algorithms.

The paper was presented at the 35th Small Satellite Conference.