Computer vision applied to wave flume measurements
Introduction
Physical model tests are a fundamental part of many Ocean and Coastal Engineering projects, hence the importance of measuring techniques in the wave flume. In the case of physical model tests of a floating body, such as an oil containment boom, the conventional approach requires a relatively complex deployment of sensors in contact with the moving medium to record its motions. In addition to the complexity and cost of this type of deployment, it usually imposes unwanted boundary conditions in the form of curtailments of the motion. These laboratory effects may jeopardise the representativity of the physical model tests.
The techniques to perform precise and reliable measurements of the environment by the use of images are called videometrics (Ma et al., 2004); among other advantages, they are normally non-intrusive, so that their effect on the environment tends to be minimum or nil. In this work, a new computer vision system for wave flume measurements is presented and put to use in physical model tests of a floating oil boom. The system determines the motion of the floating body (displacements and roll) without sensors of any kind in contact with the physical model; thus, the laboratory effects brought about by conventional measuring techniques are avoided. The computer vision system also records the motion of the free surface, with two main advantages relative to conventional wave gauges. First, it allows the determination of free surface motion with the video camera's field of view—unlike conventional wave probes, which only give the free surface displacement at a point. Second, the computer vision system is non-intrusive: no elements are placed within the flow domain.
A number of methods have been put forward to measure the free surface in a wave flume by means of computer vision (e.g. Erikson and Hanson, 2005; García et al., 2003; Flores et al., 1998; Javidi and Psaltis, 1999; Canning et al., 1998; Zhang, 1996; Bonmarin, 1989). Lee and Kwon (2003) applied wavelet transforms, a computationally intensive task which precludes real-time operation. Other methods resorted to active contour models—elastic curves that dynamically deform within an image to match any shape of an object (Neuenschwander et al., 1997; Xu and Prince, 1998; Duncan et al., 1999 Yao and Wu, 2004). In particular, Yao and Wu (2004); use a high-speed camera in combination with an argon-ion laser, and dye the water in the flume with sodium uranine fluorescein; their experimental setup is expensive, and in addition their algorithm requires user intervention to determine the desired air–water boundary. For their part, Hequan et al. (2004) make use of light refraction for the measurement of wave surface elevation by filming a section of the wave flume bed painted with black and white stripes; it is an inexpensive, non-intrusive method, but of limited accuracy. Hilsenstein (2005) used thermographic image sequences of the water surface for the 3D reconstruction of the water surface. Most of the above methods have one point in common—they do not allow the presence of objects or other fluids in the water. For this reason, they cannot be used for wave flume tests of a floating physical model, such as an oil boom.
Relative to previous applications of computer vision to wave flume measurements, the system presented in this work differs in four respects. First, it is the only method that detects the displacements of a floating body as well as those of the adjacent free surface. Second, the new system does not require colouring of the water with a fluorescent dye – nor, for that matter, with any kind of dye – unlike many of the previous systems (e.g. Erikson and Hanson, 2005; Yao and Wu, 2004). Third, the processing of the video frames and the determination of the displacements of interest, during a test, is concurrent with the test itself, so that no post-processing is necessary; in other words, the system works “in real time”—unlike most previous methods. It goes without saying that to achieve this goal, the computational tasks had to be streamlined insofar as possible. For example, the application of wavelet transforms to determine the wave profile (Lee and Kwon, 2003) was precluded, because of its high computational cost. Finally, whereas some of the previous applications required the use of expensive, very specific hardware, such as an argon-ion laser (Yao and Wu, 2004), only ordinary laboratory hardware is required in the present method (a video camera and a conventional PC).
A fundamental question when dealing with a measuring technique is its accuracy. This has been analysed by comparing the free surface measurements carried out with the computer vision system to those performed with a conventional wave gauge. An excellent agreement was found. In this context, it is of interest to recall that some of the recent applications of computer vision to the determination of free surface motions have led to less accurate results (e.g. Hequan et al., 2004).
This article is structured as follows. First, the experimental setup is described. This is followed by the particulars of the computer vision system (Section 3). Next, the validation results obtained in 19 tests, including both regular and irregular wave tests, are shown (Section 4). Finally, conclusions are drawn.
Section snippets
Experimental setup
The oil boom tests were carried out in the new wave flume of the University of Santiago de Compostela, a 20 m long, 0.65 m wide and 0.90 m high flume with a piston-type wave paddle equipped with an active wave absorption control system. A dissipation ramp was installed near the downwave end of the flume to minimise wave reflection; it has a constant slope of 1:10, and culminates at the end wall with a height of 55 cm above the flume bed (Fig. 1).
The physical model represents the section of an oil
Computer vision system
Computer vision systems are artificial systems that obtain information from images (Shapiro and Stockman, 2001). The image data may take on many forms, such as a photograph, a video sequence, views from multiple cameras, etc. The present system uses images from a digital video camera and a set of processing algorithms; while the camera records a video sequence of the flume section of interest – the model section – at a rate of 25 frames per second, the algorithms process the images to determine
Results
For validation, the measurements obtained with the artificial vision system were compared with those obtained by means of a conventional wave gauge in a series of tests with both regular and irregular waves. The wave gauge used, type 202 of the Danish Hydraulic Institute, is based on the principle of conductivity between parallel electrodes; its technical specifications are: resolution, <1 mm; effective measurement length, 58.50 cm; linearity, <±1.5% of the effective measurement length; and
Conclusions
A new computer vision system was developed for use in wave flume tests, which measures the displacements of the free surface (waves) and – whenever necessary – those of a floating model. The system was validated based on free surface measurements carried out with a conventional wave gauge in regular and irregular wave tests; the agreement between both measurements was found to be excellent.
The new system presents a number of advantages relative to conventional measuring techniques. First, it is
Acknowledgments
The authors wish to express their gratitude to three anonymous reviewers, whose comments contributed greatly to improving the manuscript and figures. This research was funded by the European Union, within the framework of the Interreg IIIC programme. (Ref. IIIA-PROLIT-SP1E194/03).
References (26)
- et al.
A method to extract wave tank data using video imagery and its comparison to conventional data collection technique
Computers & Geosciences
(2005) - et al.
Wave profile measurement by wavelet transform
Ocean Engineering
(2003) - Bendat, J.S., Piersol, A.G., 2000. Random data: analysis and measurement procedures. John Wiley, New...
Practical Handbook on Image Processing for Scientific Applications
(1997)Geometric properties of deep-water breaking waves
Journal of Fluid Mechanics
(1989)- Canning, P., She, K., Morfett, J., 1998. Application of Video Imaging to Water Wave Analysis. Experimental Studies:...
- et al.
Deinterlacing—an overview
Proceedings of the IEEE
(1998) - et al.
Gentle spilling breakers: crest profile evolution
Journal of Fluid Mechanics
(1999) - Flores, H., Andreatta, A., Llona, G., Saavedra, I., 1998. Measurements of oil spill spreading in a wave tank using...
- García, J., Herranz, D., Negro, V., Varela, O., Flores, J., 2003. Tratamiento por color y video. VII Jornadas Españolas...
Wave measurement based on light refraction
Acta Oceanologica Sinica
Cited by (15)
A novel and simple passive absorption system for wave-current flumes
2023, Alexandria Engineering JournalAn algorithm for tracking drifters dispersion induced by wave turbulence using optical cameras
2021, Computers and GeosciencesCitation Excerpt :The motions of the water surface along the section are detected exploring the high contrast along the interface in the digital images. In some instances, the displacement of a floating body can be determined as well, using colored markers to simplify its detection on the image applying Gaussian differences (Iglesias et al., 2009). With a similar approach, Erikson and Hanson (2005) reported satisfactory results to extract the interface between the sand bottom and the water, as well as the interface between the water and air.
Water elevation measurements using binary image analysis for 2D hydrodynamic experiments
2018, Ocean EngineeringCitation Excerpt :In addition, wave probes may provide false water elevation measurements in cases where two or more water surfaces are observed, such as in the case of air trapped due to a plunging wave (Yao and Wu, 2005). The problems described above have motivated researchers to apply image processing and analysis methods to measure water elevations or related fluid parameters in tests conducted in laboratories (Erikson and Hanson, 2005; Iglesias et al., 2009; Wang et al., 2012; Zhang, 1996; Yao and Wu, 2005; Kim et al., 2011; Lee and Kwon, 2003). The advantages of image-based methods are that they are non-intrusive and allow spatial and temporal measurements at desired virtual wave probe positions.
Genetic programming and floating boom performance
2015, Ocean EngineeringCitation Excerpt :The models, consisting on three typical elements: a cylindrical float, a vertical skirt and a weight acting as ballast, were allowed to move freely in presence of both waves and currents, as move the prototypes. In order to determine the time evolution of the effective draft during the tests, a Computer Vision System was used to record the free surface position and the motions of the boom models in a nonintrusive manner (Iglesias et al., 2009a). The boom performance in each test was characterized using a statistic of the effective draft time records, the significant effective draft, which will be the target variable for the GP.
Field evaluation of optical-based three-dimensional dynamic motion measurement system with multiple targets for a floating structure
2013, Ocean EngineeringCitation Excerpt :For flume tank tests, several researches were carried out to improve the measurement techniques using optical system. Iglesias et al. (2009) applied the computer vision system to wave flume measurements and they detect the free surface motions along flume section, and the floating body motion was very efficiently measured without conventional sensors in flume. Wang et al. (2012) developed the modified plane-based camera calibration method to apply optical-based measurement system for flume tank experiments, and they found that the RMS error was lower than one millimeter in their flume tank tests.