A comparison of vision-based tracking schemes for control of microbiorobots

Abstract

There has been significant recent interest in micro-nano robots operating in low Reynold's number fluidic environments. Even though recent works showed the success of controlling micro-nano robots, there are some limitations because of the tracking method. In this paper, we introduce and implement a feature-based tracking method (FTM). Scale invariant feature transform (SIFT) is a well-explored technique at much larger length scales for research fields regarding robotics and vision. Here, the technique is extensively investigated and optimized for microbiorobots (MBRs) in low Reynold's number environments. Also, we compare the FTM with the conventional tracking method for cells, which is known as the region-based tracking method (RTM). We clearly show that the FTM can track more accurate positions of the objects in comparison with the RTM in cases where objects are in close contact or overlapped. Also, we demonstrate that the FTM allows tracking microscopic objects even though illumination changes over time or portions of the object are occluded or outside the field of view.