Building detection from aerial images using invariant color features and shadow information

Detection of Landing Areas for Unmanned Aerial Vehicles

“…Roof detection is widely used for the detection of buildings, houses and residential areas [2] [12] [13]. Methods based on the colour of roofs, presence of shadows, or shape detection have been used in the detection of roofs and houses.…”

“…Methods based on the colour of roofs, presence of shadows, or shape detection have been used in the detection of roofs and houses. In our system, we will use a colour based system that is used in [13].…”

Automatic Vegetation Identification and Building Detection from a Single Nadir Aerial Image

“…Sirmacek et al in [10] have developed a feature and area based approach employing color invariant features [11] and shadow information for building detection from aerial imagery. Shadows are detected by Otsu thresholding [12] a blue color invariant image and red building roof tops are detected by thresholding the red color invariant image.…”

“…Some will make limiting assumptions which may reduce the ability of the algorithm when executed for data sets other than what were presented in the associated paper. Examples of such limiting assumptions would be buildings only having specific root top colors [8][9][10], buildings only existing at right angle corners and/or buildings having parallel sides [7,10]. Rather than saying all buildings have parallel sides or right angle corners, a more general assumption would be that buildings tend to have convex hull rooftop sections.…”

“…Finally, the results should be tested on several hundred or even thousands of buildings of varying sizes, shapes, orientations, roof top colors, and roof top textures. Several of the current related papers in the literature [8][9][10]13] only present results from a couple of hundred buildings. While the semi-automatic/supervised approaches [9,13,14] tend to have good accuracy, they rely on manually extracting features from 1/4 to as much as 1/2 the data set.…”