Implementation and Evaluation of Vision-Based Sensor Image Compression for Close-Range Photogrammetry and Structural Health Monitoring.

Much research is still underway to achieve long-term and real-time monitoring using data from vision-based sensors. A major challenge is handling and processing enormous amount of data and images for either image storage, data transfer, or image analysis. To help address this challenge, this study explores and proposes image compression techniques using non-adaptive linear interpolation and wavelet transform algorithms. The effect and implication of image compression are investigated in the close-range photogrammetry as well as in realistic structural health monitoring applications. For this purpose, images and results from three different laboratory experiments and three different structures are utilized. The first experiment uses optical targets attached to a sliding bar that is displaced by a standard one-inch steel block. The effect of image compression in the photogrammetry is discussed and the monitoring accuracy is assessed by comparing the one-inch value with the measurement from the optical targets. The second application is a continuous static test of a small-scale rigid structure, and the last application is from a seismic shake table test of a full-scale 3-story building tested at E-Defense in Japan. These tests aimed at assessing the static and dynamic response measurement accuracy of vision-based sensors when images are highly compressed. The results show successful and promising application of image compression for photogrammetry and structural health monitoring. The study also identifies best methods and algorithms where effective compression ratios up to 20 times, with respect to original data size, can be applied and still maintain displacement measurement accuracy.