Machine Learning Kibera, Kenya
Machine learning has been a core component of spatial analysis in Geographic information systems. Tools to operate geospatial ML models such as ArcGIS Pro and Picterra have been applied to solve historic classification problems. With these applications, one can use vector machine algorithms to create land-cover classification layers on images captured by drones. At the highest level, deep learning, a type of machine learning, has enabled user to creates training samples, for example by drawing polygons over building footprints, and the computer model learns from these training samples and scans the rest of the image to identify similar features.
With the increased availability of satellite and drone data, and the improvement of technology and processing power, we wanted to use deep learning methodologies to extract building footprints in the informal settlement of Kibera. Using Picterra, the process of developing and running deep learning models was very simple (no coding was required). Picterra provided tools to capture training data, train various models, inference and then derive the needed footprints after processing.
Another tool one can leverage to extract this data is ArcGIS Image Analyst for ArcGIS Pro. Though more advanced with some coding knowledge required, it provides tools for advanced image interpretation, exploitation, and geospatial analysis on an array of imagery modalities. One can automate and speed up workflows such as feature extraction, image classification, multidimensional analysis, and change detection with a robust set of image-based machine and deep learning tools.
Used procedures and technologies
ArcGIS Pro
Machine Learning
Picterra
#MACHINE LEARNING - #URBAN PLANNING - #URBAN MANAGEMENT