Identifying and analyzing footprints of buildings in aerial and satellite data is an important first step in many applications, including updating maps, modeling cities, analyzing urban growth and monitoring informal settlements.
But manually identifying and collecting information about buildings from single or stereo imagery is very tedious and costly. Automatic building detection methods can help solve this problem — especially deep learning-based classification methods.
Several DL-based classification and object detection methods have been implemented in ERDAS IMAGINE, which reduces the need for expert users and is highly accurate. The ease and flexibility of semantic segmentation make it a valuable tool in the process of building extraction.
To learn more, read the full white paper.
