A Novel Vision Sensor for Remote Measurement of Bridge Displacement

This project was aimed at developing a camera-based computer vision sensor system for accurate remote measurement of multipoint bridge displacements in outdoor environments. The vision sensor can track the movement of natural targets on the structure from a convenient remote location without accessing the bridge to install sensors, thus significantly reducing the cost of the equipment and operation as compared to conventional sensors. In addition, the measurement points can be easily altered in post processing of the video. However, outdoor conditions such as changes in illumination and background, heat haze-induced image distortions, and camera vibration can cause significant measurement errors. To overcome these challenges, this project developed innovative algorithms and methods including (1) a gradient-based template matching algorithm for robust tracking of “natural” markers (such as rivets on a bridge), integrated with a subpixel technique for enhancing image resolution required for multipoint measurement; (2) a heat haze filtering technique based on detection and statistical characterization of heat haze-induced image distortions; and (3) a practical vibration cancellation method based on simultaneous measurement of displacements of the target structure and a stationary point. In addition, this project proposed a practical calibration method to convert image pixel displacements into physical displacements. The project was executed in two stages. Stage 1 focused on developing the algorithms, techniques and methods to systematically address all sources of environmental noise that deteriorate image quality and measurement accuracy. Extensive laboratory tests were carried out to validate the effectiveness of the developed algorithms and techniques using simulated environmental noise including low-lighting, shadowing, heat haze, and change in illuminating light and background conditions. In Stage 2, the algorithms developed in Stage 1 were integrated into a software package and field performance evaluation tests were carried in three bridges including two long-span steel bridges, the Manhattan Bridge and the Williamsburg Bridge, and a short-span concrete bridge, the Jamboree Bridge. The remote, real-time and multipoint measurement capabilities of the vision sensor system developed in this study were further validated in presence of various sources of field environmental noise including heat haze and camera vibration. In addition, this project demonstrated, through a laboratory experiment, the use of the vision sensor system for structural dynamic tests, modal analysis, and damage detection, by simultaneously measuring a dense array of displacements at extremely low cost. In the future, the system can be further developed for permanent installation at bridge sites to enable long-term continuous monitoring of structural integrity and safety. The final report is available.