High-precision automated processing of sequential images for high-speed videogrammetric measurement

High-speed videogrammetric measurement is widely used in fields of structural health monitoring. However, it is difficult to perform efficient and accurate automation for the copious sequential images data. This paper proposes a novel high-precision automated sequential images processing method for high-speed videogrammetric measurement. First, a precision circular marker detection network model is proposed to detect circular marker, which can effectively address recognition challenges associated with small, dense, and deformed markers in complex scenarios with a precision of 99.37 and recall of 98.86. Second, a circular center tracker, utilizing Kalman filtering and the Hungarian matching algorithm, is presented to achieve highly automated sub-pixel tracking of marker centers with an root mean square error of 0.092 pixel. At last, a global confidence optimization matching strategy is put forward to attain precise automated stereo images matching with an accuracy of 94.48%. The results show that the proposed method can significantly advance the intelligence of high-speed videogrammetric measurement.