Detection of windows and doors from thermal images by grouping geometrical features

In recent years, very high energy consumption is the major problem of the big cities. Most of the energy of the cities are disbursed to warm and cool buildings. Thus, detecting heat leakages on building walls is a new research problem. In this study, we propose a novel system to detect thermal leakages automatically from thermal camera images. To this end, we use sequential thermal images of buildings. First, we start with fusing thermal image sequences to obtain rectified building facade with higher resolution. Then, we detect L-shaped features using a set of steerable filters. We use L-shaped features and perceptual organization rules to detect windows and doors from rectified thermal image. After eliminating detected doors and windows from building facade, we detect problematic regions. One of the advantage of proposed system is that, it can also be used to detect building damages automatically even in night time. Therefore using proposed system, it may be possible to detect thermal leakages and also damages by only using images taken from a vehicle which is moving around interested buildings.