A Robust Circular Marker Detection Method Based on Revised Tiny-yolov3

Circular markers were often used in computer vision measurement due to their obvious geometric features and high positioning accuracy. However, traditional methods were susceptible to external interference when locating the circular marker, which could lead to location failure and affect the actual measurement. For this reason, a robust circular marker detection method based on revised tiny-yolov3 was proposed in this paper. Specifically, we improved the network structure of tiny-yolov3 and added the geometric constraint of the object boundary box in the loss function of the algorithm to improve the positioning accuracy of circular markers. We also performed a large number of experiments on actual datasets. The results have shown that our method could accurately locate the region where the circular marker is located, and the precision reaches 98.93%, which was more than 8 percentage points higher than the state-of-the-art methods. The robustness is significantly improved.