A comprehensive alleviation technique for optical-turbulence-induced errors in vision-based displacement measurement

Computer vision-based displacement measurement techniques can be affected by the optical turbulence in the field in hot weather. The optical turbulence can be observed naturally when viewing a thick layer of heated air. The nonuniform density distribution of the air due to heat creates spatial variations in refraction indexes, resulting in distortions in video images and measurement errors in displacements obtained from distorted videos. The effect of optical turbulence on vision-based displacement sensing is first illustrated in the field test conducted in hot weather. Then, the statistical characteristics of the optical-turbulence-induced measurement errors are analyzed and confirmed. And a comprehensive optical-turbulence error alleviation technique, which has two steps, is designed based on the statistical studies. A previously developed high-performance multitarget vision-based displacement measurement technique is utilized in the first step to track multiple targets. Distortions of multiple targets are estimated simultaneously continuously. Targets with the least distortions in each frame are identified and their displacements are extracted as the primary displacement. To further alleviate errors in the primary displacement, an adaptive optical-turbulence error filter is formulated in the second step based on the statistical characteristics of the optical-turbulence errors. Validations of the optical-turbulence error alleviation technique are performed in both laboratory and field tests. Field tests are conducted on the Williamsburg Bridge in the natural environment. After alleviating optical-turbulence and camera vibration errors, frequency domain analyses are conducted on vibrational displacements of the Williamsburg Bridge in response to train passing.