Development of flow velocity measurement techniques in visible images; Improvement of particle image velocimetry techniques on image process

Kimura, Nobuyuki; Miyake, Yasuhiro*; not registered; Kamide, Hideki ; Hishida, Koichi

Noise reduction system was developed to improve applicability of Particle Image Velocimetly (PIV) to complicated configure bounded flows. For fast reactor safety and thermal hydraulic studies, experiments are performed in scale models which usually have rather complicated geometry and structures such as fuel subassemblies, heat exchangers, etc. The structures and stuck dusts on the view window of the models obscure the particle image. Thus the image cxcept the moving particles can be regarded as a noise. In the present study, two noise reduction techniques are proposed. The one is the Time-averaged Light Intensity Subtraction method (TIS) which subtracts the the-averaged light intensity of each pixel in the sequential images from the each corresponding pixel. The other one is the Minimum Light Intensity Subtraction method (MIS) which subtracts the minimum light intensity of each pixel in the sequential images from the each corresponding pixel. Both methods are examined on their capabilities of noise reduction. As for the original "bench mark" image, the image made from Large Eddy Simulation was used. To the bench mark image, noises are added which are referred as sample images, Both methods reduce the rate of vector with the error of more than one pixel from 90% to less than 5%. Also, more than 50% of the vectors have the error of less than 0.2 pixel. The analysis of uncertainty shows that these methods enhances the accuracy of vector measurement 312 times if the image with noise were processed, and the TIS method has 1.1 2.1 times accuracy compared to the TIS. Thus the present noise reduction methods are quite efficient to enhance the accuracy of flow velocity fields measured with particle images including structures and deposits on the view window.