Figure 4.6 : A shape that will cause the partitioning algorithm to fail.
first marker pixel that would cast a ’shadow’ in the direction of projection and once found ceases to look further as any further pixels found would not change the result, i.e., that there is a shadow.
There are cases that could cause the algorithm to fail. Consider the example in Figure 4.6. In this example the projections in both horizontal and vertical directions create continuous ’shadows’, however, the ROI contains distinct markers. Beacons can be designed so that these shapes do not occur (or occur infrequently). lso, as partitioning only occurs in the segmentation phase, once a ROI is created it will not need to be partitioned again while it
is being tracked.
ROI size adaptation algorithm
s the camera module moves, markers will appear to grow and shrink due to the changing
distance between the camera and the markers. The ROIs that surround these markers must also grow and shrink so that each marker is surrounded entirely by its ROI and so ROIs do not overlap. The latter could happen if the camera moves away from a beacon. Figure 4.7a shows ROIs surrounding two markers. s the distance from the camera to the markers increases, an image pixel covers a greater physical distance on the marker. Therefore, the ROIs will cover a greater area around the markers and could overlap with an adjacent ROI as shown in Figure 4.7b. If the overlap is great enough to include part of the adjacent marker the centroid will be biased and could affect the pose estimate generated by the
The diameter of the marker within the ROI can be measured to determine whether the size of the ROI needs to be increased or decreased. Calculating the diameter of the marker can be achieved by using the formula for the area of a circle, i.e., A = πr2. Criteria are needed to determine whether a marker image is too large compared with the size of its ROI or too small. ROI is too small if it means that motion that could be encountered would cause the marker image to be lost, i.e., leave the ROI. Equations for the size of a ROI were developed in Chapter 3. This algorithm uses Equation 3.23 and a tolerance value T in
lgorithm 1 illustrates the algorithm operation. In this code S
L is the actual