Study of fluid edge detection and tracking method in glass flume based on image processing technology. (October 2017)

• Record Type: Journal Article
• Authors: Du, Hai; Li, Muguo; Meng, Juan
• Is Part Of: Advances in engineering software. Volume 112(2017)
• Page Start: 117
• Page End: 123
• Publication Date: 2017-10
• Subjects: Oceanographic techniques -- Fluid flow measurement -- Image edge detection -- Image segmentation -- Image motion analysis; Computer-aided engineering -- Periodicals; Engineering -- Computer programs -- Periodicals; Engineering -- Software -- Periodicals; Periodicals; 620.0028553
• DOI: 10.1016/j.advengsoft.2017.04.007
• Languages: English
• ISSNs: 0965-9978

Abstract

Highlights: A new fluid level detection method base on image processing technology is proposed and an image segmentation technology with shifting-window is used to track fluid level changes instead of traditional image edge detection method. The interrogation window in proposed method can change its position adaptively with fluid motion, which can flexibly track fluid level changes and can ensure that the amount of data to be analyzed always remains stable. The proposed method is not only fit for edge detection of water level in a still image, but also can be used in image sequence to track fluid level dynamically. The propose method can be easily used to achieve one and more fluid edges in one image and is also effective to detect edge of water with a certain transparency. Abstract: Research on changes in the fluid edge of a wave flume is important for experimental hydrodynamics. However, disturbances often occur because of the presence of sensors. To solve this problem, a new grey-scale image processing method for fluid edge analysis is presented here. By fusing methods combining image gradients and image segmentation with shifting-window technology and with concepts derived from experimental fluid mechanics, the proposed method can overcome many of the inherent challenges of fluid-edge measurement. First, the geodesic distance is modified to obtain a class curve. Second, an edge position is determined by the inflection point of the class curve related to the gradient peak distribution. Next, the position of the interrogation window is relocated with reference to neighbors or to previous results, and the current edge position can be calculated according to the predicted value. During the computation, the interrogation window can change its position adaptively with fluid motion, ensuring that the amount of data to be analyzed always remains stable. A model combining the class curve and gradient curve can improve the validity of edge identification. Finally, the performance of the proposed method has been evaluated using images in a glass flume. The results show that the proposed method for studying the fluid edge is effective and robust.