X hits on this document

PDF document

A Prototype Optical Tracking System Investigation and Development - page 53 / 170





53 / 170


lgorithm design


Figure 4.2 : A ROI is created when a bright pixel is detected at a grid point. The grid is moved every frame to allow stationary markers to be found. The numbers in the boxes identify the ROI.



Figure 4.3 (a) A ROI can contain multiple markers. This can occur when a ROI is created around a group of markers that are far away from the camera. (b) Partitioning a ROI can solve this problem.


Multiple marker detection

During the segmentation phase, the initial marker detection algorithm can create ROIs that contain multiple markers. This occurs when the size of the ROI is large compared with the size of the markers’ images (Figure 4.3a) and can happen when a beacon is far from the camera. In this case the default ROI size will cause multiple markers to be encompassed

by one ROI. this situation.

n algorithm that can split the ROI into smaller regions was developed for

Pixel connectivity and projection approaches were considered.

pixel connectivity ap-

proach is suggested by Shortis et al. 1994 [47]. Every pixel in a ROI is traversed to find groups of connected pixels that are above a threshold. Each group forms a ROI. This method was discarded as it requires each pixel to be tested. Compared with the projec-

tion approach described below it is likely to be computationally expensive.

The projection approach is to look for lines that can be drawn between the markers in the ROI. Since a ROI in this software is defined using vertical and horizontal lines only, a method of partitioning can be developed using only vertical and horizontal lines. Figure 4.3b shows an example of partitioning a ROI containing three markers using two vertical lines.

Document info
Document views300
Page views300
Page last viewedFri Oct 28 19:55:53 UTC 2016