“…While the sensor-based SHM often requires expensive data acquisition systems [ 8 , 9 , 10 , 11 , 12 , 13 ], the vision-based approach requires only simple and low-cost setup and operation. Besides this, vision sensors are non-contact, while traditional sensors must be contacted to a target structure, causing degradations in the sensor’s quality under environmental effects [ 14 , 15 ]. Furthermore, the structural information obtained from a vision sensor is not affected by environmental variation, while the signals obtained from a contacted sensor can be significantly affected by environmental changes, which is the main cause of false detections [ 16 , 17 , 18 , 19 ].…”