Analysis of factors influencing precise positioning estimates for 2D fiducial targets

Abstract

In the past decade, the rapid progress in image processing applications has made it increasingly crucial to obtain high-quality images. However, the enhancement of images through post-processing methods can sometimes be restricted by the capabilities of available cameras, such as resolution, sensor size, and FPS. Researchers have made significant efforts to enhance camera performance and maximize the effectiveness of post-processing, with improving image resolution being a particularly important focus area. This paper aims to study some factors affecting the 2D detection accuracy for fiducial target objects (circular shaped objects) when two modifications of Circular Hough transform (CHT) are used (namely Phase coding and two-stage methods). These factors include those related to the internal properties of Circular Hough transform algorithm such as search radius range, thresholding and colour intensity. Also, this paper analysis the influence of some factors connected to the nature of the object which involves the boundaries of the shape and the target’s speed. Therefore, Three 2D shapes with varying diameters are being used: a small shape with a radius of 1 cm, a medium shape with a radius of 2 cm, and a large shape with a radius of 3 cm. Moreover, the influence of occlusion problem on the overall detection success for the targets has also been highlighted. The results show that fiducial targets with circular shapes have a changeable effective resolution at slow motion speeds. Additionally, adjusting the internal parameters of the two CHTs resulted in a minor alteration in the detected pixels of the target within the image, even without any movement occurring. Consequently, there was no notable shift in the positioning data of the tracked object across the consecutive frames.