Visual measurement of orientation using ceiling features

A system for determining orientation by exploiting straight line features commonly found on indoor ceilings is described. A method of random sampling of minimal sets to extract straight line primitives from binary images of edge feature points is used. Orientation of the lines extracted gives the relative orientation of a local camera reference frame. Input to the random sampling based line extraction stage is supplied by an image processing stage. The image processing stage applies a morphological gradient to grey scale ceiling images and automatically adjusts a threshold to produce the required binary images. Feasibility of the proposed system was tested using a suite of ceiling images and the results are reported