Vis enkel innførsel

dc.contributor.authorMaharjan, Samee
dc.date.accessioned2020-12-14T14:55:17Z
dc.date.available2020-12-14T14:55:17Z
dc.date.issued2020-12-18
dc.identifier.isbn978-82-7206-580-4
dc.identifier.issn2535-5252
dc.identifier.urihttps://hdl.handle.net/11250/2719281
dc.description.abstractThe method of high speed photography is used to visualize those phenomena which occur so fast that it is impossible to visualize by the normal human eyes. One of those phenomena is a shock wave propagation during gas explosion experiments. A shockwave is a strong compression wave existing in the supersonic flow field across which gas properties like pressure, temperature, and density increase significantly. This thesis is aimed at developing image processing frameworks, which will process the high speed videos captured during gas explosion experiments and extract some useful information about the shock waves. One way to extract any sort of information about propagating waves from the high speed videos is by tracking the position of the wave front. The common choice of image processing technique to perform this then naturally becomes any kind of edge detection technique. However, when the images are comparatively of low quality in terms of contrast, resolution and include a high amount of noise, basic edge detection techniques might not give the precised result. Hence, some of those image processing techniques, which have the potential to detect edges in the low quality images were studied and implemented in this thesis. The first approach is based on a energy minimizing curve, which moves towards the edges and eventually lies around the edges, widely known as active contour models or Snakes. Based on a classic close contour approach, an open contour model is developed and implemented to contour a wave front from top to bottom of the image (Article I). The second technique studied is region wise image segmentation method called the watershed algorithm (Article II). Both of these methods do track the edges within a required precision however, they are time consuming. The active contour model requires a good initialization of the curve from the user and also includes multiple parameters. The watershed algorithm eliminates any parameter requirements however, requires multiple pre/post-processing steps. The third technique is a statistical object detection method of template matching, which reduces the number of pre/post-processing steps. At first, a binary template matching is implemented in the binary images of the high speed videos (Article IV). This method minimizes the edge detection error but requires an image to be transformed into a binary form. The updated template matching uses a dynamic template that varies its intensity values depending on each considered image (Article III). This approach eliminated the need of image thresholding in order to detect edges. Furthermore, it shows to be more robust and faster compares to previous techniques. Even though it is possible to track the front without any pre-processing by using a dynamic template matching, it shows better results in the filtered image. The images from a high speed camera, when operated in a higher frequency are comparatively of low quality, and also ongoing chemical changes in the flow significantly corrupt the images. A standalone edge detection technique therefore might not be able to track the front as accurately as when it is combined with an image denoising/filtering. Hence, image filtering in both spatial and Fourier domain were also studied and implemented before applying any of above mentioned tracking techniques. The tracking of wave fronts do not only show the structure and position of the waves but also gives a possibility to extract primary information about the shock wave like, shock speed, shock angle, etc. Furthermore, secondary information like Mach number, pressure, and temperature can also be estimated, by combining the primary information and the traditional gas dynamics equations. While calculating shock speed from the tracked shock position, a basic two-point method (distance/time) shows some oscillations in the result. Hence, a relatively new approach non-linear square fit method (NLSFM) was modified and implemented, which reduces the oscillations significantly (Article V). For validation, the estimated pressure for some of the experiments was compared to the reading from pressure transducers, which shows a good match. The results provide insightful information about the reflected shock wave and its boundary layer interactions. The calculated wave properties demonstrate a variation that occurred within a time interval of 300 microseconds (ms) at a distance of 100 millimeters (mm). This information is difficult to extract while using a traditional approach such as pressure transducers. Thus, a combination of the high speed videos and digital image processing has a huge potential to study gas dynamics phenomena is a detailed manner.en_US
dc.language.isoengen_US
dc.publisherUniversity of South-Eastern Norwayen_US
dc.relation.ispartofseriesDoctoral dissertations at the University of South-Eastern Norway;83
dc.relation.haspartArticle I: Maharjan, S., Gaathaug, A.V. & Lysaker, O.M.: Open Active Contour Model For Front Tracking Of Detonation Waves. Proceedings of the 58th International Conference of Scandinavian Simulation Society, SIMS 2017, p. 174-179, 2017. http://dx.doi.org/10.3384/ecp17138174en_US
dc.relation.haspartArticle II: Maharjan, S., Bjerketvedt, D. & Lysaker, O.M.: An Image Processing Framework for Automatic Tracking of Wave Fronts and Estimation of Wave Front Velocity for a Gas Experiment. Communications in Computer and Information Science 842, (2019), 45-55. https://doi.org/10.1007/978-3-030-19816-9_4. Reprinted with permissionen_US
dc.relation.haspartArticle III: Maharjan, S.: Wave Front Tracking in High Speed Videos Using a Dynamic Template Matching. Lecture Notes in Computer Science 11867, (2019), 531-542. https://doi.org/10.1007/978-3-030-31332-6_46. Reprinted with permissionen_US
dc.relation.haspartArticle IV: Maharjan, S., Bjerketvedt, D. & Lysaker, O.M.: Processing of high-speed videos of shock wave boundary layer interactions. Signal, Image and Video Processing, (2020). https://doi.org/10.1007/s11760-020-01782-5en_US
dc.relation.haspartArticle IV: Maharjan, S., Bjerketvedt, D. & Lysaker, O.M.: Information Extraction from High Speed Videos of Reflected Shock Wave Interaction With Boundary Layerteractions. Manuscript. Not available onlineen_US
dc.relation.haspartUnpublished work: Machine learningen_US
dc.rightsNavngivelse-Ikkekommersiell-DelPåSammeVilkår 4.0 Internasjonal*
dc.rights.urihttp://creativecommons.org/licenses/by-nc-sa/4.0/deed.no*
dc.titleAn Image Processing Framework for High Speed Videos from Combustion and Gas Explosion Experimentsen_US
dc.typeDoctoral thesisen_US
dc.description.versionpublishedVersionen_US
dc.rights.holder© The Author, except otherwise stateden_US
dc.subject.nsiVDP::Teknologi: 500::Informasjons- og kommunikasjonsteknologi: 550::Teknisk kybernetikk: 553en_US


Tilhørende fil(er)

Thumbnail

Denne innførselen finnes i følgende samling(er)

Vis enkel innførsel

Navngivelse-Ikkekommersiell-DelPåSammeVilkår 4.0 Internasjonal
Med mindre annet er angitt, så er denne innførselen lisensiert som Navngivelse-Ikkekommersiell-DelPåSammeVilkår 4.0 Internasjonal