Computational stereoscopy for three-dimensional scour depth measurement in channels

Local channel scour is a problem of substantial importance in hydraulic engineering practice. Scour occurring around hydraulic structures has serious consequences in terms of damage to the structure and to human safety. The mechanisms of scour are complex and dynamic. This has resulted in most of the studies dealing with local scour to be experimental in nature. The amount of erosion has generally been quantified by noting the depth of scour at various locations in the bed. When the flow is unsteady or quasi-steady, determining the instantaneous depth is not a trivial task and involves a three-dimensional analysis of the scour field. Furthermore, the depth measurement has to be non-intrusive. The present study aims at developing a practical video image analysis for use in three-dimensional scour depth measurement. To acquire images, two cameras with similar specifications were mounted over the model of the eroded bed and the two images were simultaneously captured to simulate stereoscopic viewing. The computational scheme involved the use of an epipolar constraint and a relaxation technique to match corresponding points in the pair of stereo images. A correlation technique was developed to eliminate false matches. The depth of scour was determined using the parallax between the matched points. Following the successful application of the stereoscopy procedure to grayscale images, the scheme was applied to colour images using the pixel hue, saturation, and intensity values.