Image Measurement Of Characteristic Parameters Of Aerated Water Flows

Recently, with the development of computer techniques and image processing, image measurement, as an un-intact lossless detection method, has been paid more attention by most researchers during the study of characters of aerated water flows.This project comes from the National Fund of Natural Science (No. 50079020), whose aim is to study the measurement of the characteristic of aerated water flows by image processing. Firstly, the experimental system of the project is introduced. Then, the design of the system, which satisfies the image measurement, is obtained from the construction of the experimental system.Because the bubbles and water are both colorless and transparent in aerated water flows, the contrast of the captured images is low, and uneven illumination exists. Two bubbles extraction methods are proposed to eliminate the un-balance of illumination and higher accuracy bubble extraction is obtained. The first method is adaptive threshold segmentation based on local blocks of image. Firstly, the original image is parted into disjoint sub-blocks with same size. Then, different threshold is used in different block to extract bubbles. Finally, an optimal extract result is obtained by synthesizing of different parted blocks. The second method is bubble extraction based on adaptive local region. It is a two-level dynamic threshold method. In the first level, local regions of each bubble are acquired by calculating the local-variance of the image. In the second level, different threshold is used to each of the regions obtained in the first level. The accuracy of segmentation is improved because there are onlyone or a few bubbles in each region, and the illumination disturbance of different objects can be reduced efficiently.Generally speaking, the segmentation method itself couldn't judge whether there are overlapping bubbles or not in the images; also couldn't split them. A concavity detection method of the overlapping bubbles, based on the projection of the object contour is proposed here; then departing point pair is obtained by matching the concavities; finally, the splitting curve is obtained by the departing point pair constrained based minimum mean square error(MMSE) ellipse fitting. Thus the areas of the overlapping regions are compensated efficiently and the accuracy of measurement is highly improved.The volume of the bubbles can be estimated on the bubbles' area information, and a genetic algorithm (GA) based method is used to match and track bubbles in the image sequences, which solve tracking problem under complex conditions efficiently, such as some bubbles may have kinetic occlusion and crossover, some newly generated bubbles may entered into flow field and run away. A smooth kinetic locus is obtained by fitting the discrete centroids with a cubic spline function and at the same time the visual measurement is realized.