Cable tension monitoring through feature-based video image processing

As a key indicator of the structural performance of cable-stayed bridges, tensile forces in stay cables are required to be controlled for maintaining the structural integrity of bridges. In this paper, a non-contact vision-based system for cable tension monitoring is proposed. To measure the dynamic response of cables cost-effectively, a feature-based video image processing technique is developed. The Scale Invariant Feature Transform (SIFT) is adopted for the implementation of the feature-based methodology. Since the detected keypoints associated with the cable play a critical role in extracting the displacement time-history, a study on the feasibility of the feature-based detection algorithm is conducted under a variety of test scenarios within laboratory settings. The performance of the keypoint detector for tracking a vibrating cable is quantified based on a set of evaluation parameters. To extend the versatility of the keypoint detector within complex background scenarios, enhancement techniques are investigated as well. The analysis of the performance indicators demonstrates that the detector is capable of extracting sufficient dynamic information of a vibrating cable from a video image sequence. Subsequently, threshold-dependent image matching approaches are proposed, which optimize the functionality of the vision-based system under complex background conditions. The developed feature-based image processing technique is further integrated seamlessly with cable dynamic analysis for cable tension monitoring. Through experimental studies, the proposed non-contact vision-based system is validated for cable frequency identification as well as tensile force estimation.