Multiple Plane Detection Method from Range Data of Digital Imaging System for Moving Robot Applications - Visual Computing: Scientific Visualization and Imaging Systems - page 202

Fig. 11.9 Comparison experiments for an outdoor planar surface. a Original left, right image,
and range image obtained from the stereo sensor; b Plane segment result of conventional RHT
and peaks in Hough parameter space, in which each peak presents a plane; c Plane segment result
of proposed RHT and peaks in parameter space
plane due to the effect of noises. While the conventional RHT method detects nine
planes on an uneven flat and sidewalk flat surface. In the proposed method, the
sampled data on an uneven flat surface is filtered by planarity evaluation while the
conventional RHT method votes the parameters of sampled data on an uneven
surface into the Hough parameter space, even though it is not a plane.
Figure 11.10 and Table 11.3 shows the experiment where there is an obstacle in
the heading direction of the robot. The proposed method detected two planes of a
flat surface and obstacle, while the conventional RHT method detected nine planes
because of the influence of noises. Figure 11.10 and Table 11.3 don't include the
parameter values of the 7-9th planes, because these are not the meaningful planes.
The range data of stereo camera is easily affected by many external factors such as
lighting, shape, texture, and so on. In the proposed method, sampled data that
includes errors is filtered by planarity evaluation, but in the conventional RHT