AN EDGE-FEATURE-DESCRIPTION-BASED SCHEME COMBINED WITH SUPPORT VECTOR MACHINES FOR THE DETECTION OF VORTEX-INDUCED VIBRATION

摘要

Vortex-induced vibration (VIV) has been studied over the past decades because vibration may cause serious damages to structures such as bridges, pipelines, skyscrapers, and airplanes. Detecting VIV has long been a challenge. For example, when attempting to detect VIV, inspectors might fail to use multiple hot-wire or hot-film probes during concurrently measurement of the whole flow field. This study proposes a novel vision-based method for detecting the wake patterns of VIV; it employs an edge-feature-description (EFD)-based scheme with a multiclassifier of support vector machines (SVMs). The proposed hybrid EFD/SVM method detects and adaptively segments wake-pattern images for effective classification of the patterns. The VIV can be detected on the basis of the classification results. The experimental results demonstrated that the proposed method can effectively detect the VIV using a vision-based algorithm, which incorporates hybrid EFD/SVM for classifying wake-pattern images. The method applies a nondestructive scheme for detecting the VIV and achieves high recognition rates when classifying the wake patterns of VIV.