Image Processing for Structural Health Monitoring: The Resilience of Computer Vision-Based Monitoring Systems and Their Measurement

In today’s Information Age, the use of emerging technologies and techniques, such as machine learning, has made structural health monitoring systems reliable, and their measurements accurate and recoverable. Computer vision (CV)-based systems offer remote measurement of structures’ responses (e.g., vertical deflections, accelerations) or damage patterns on their surface (e.g., cracks). In comparison to measurements collected with contact sensors (e.g., displacement sensors and accelerometers), CV-based measurements, such as from cameras, can be fairly easily recovered in case of sensor malfunctions, or easily predicted using novel measurement interpretation techniques (e.g., machine learning algorithms). The quality of the measurements depends on environmental and camera-specific factors. The replacement of CV-based systems is effortless once they become defective due to their resilient dimension. The adaptable and sustainable nature of the smart, sophisticated CV-based systems, which also include measurement interpretation tools, emphasizes their suitability for monitoring various external stressors, hazards, and risks due to climate change. This chapter (i) defines and explains the resilience dimension of the CV-based systems and their measurements, and (ii) informs the quantification of the resilience curves for structures being monitored (i.e., by updating the fragility functions and enriching the recovery function library).