A Visual Contour-based Detection Method for Orthodontic Archwire during Robotic Bending

With the development of robot technology, the bending of orthodontic archwire is becoming automated. Online visual measurement of archwire parameters during bending is beneficial for guaranteeing high robotic bending precision. The traditional feature point detection method is unsuitable for the visual detection of the archwire due to the lack of texture features. Aimed to realize the visual detection of this kind of tiny curved object, this paper proposes a measurement method based on contour positioning and binocular vision. Firstly, the stereo correction was performed on the original archwire image. Then the thinning algorithm was used to extract the archwire skeleton. A multi-layer contour detection method was developed to extract the bending point of the archwire. Lastly, the normalized cross-correlation coefficient is used to obtain parallax information and complete the parameter measurement. The experimental results showed that the detection method successfully extracted all feature points while the SIFT algorithm and literature [15] method did not make it. Maximum measurement error not more than 0.4mm, which is practicable for justifying the bending quality and close-loop controlling the robotic bending precision.