Mineral Particles Based On Particle Size Analysis Of Image Processing And Parameter Analysis Of The Study And Realization

Granularity is the important quality index for granule material. So the precise measurement to the granularity plays a key role to the technique and practical application of the sequential processsing. Untill now, the common standard of granularity measurement is sedimentation test and other artificial operations charaterized as time-consuming process and too many examining steps in addition to some negative effects from subjectivity. The stuty trys to find a way to get the granularity information automatically based on visual computer granularity measurement technique, which avoids the subjectivity and inaccuracy from artificial measurements. As a matter of fact, image processing-oriented technique proves its extensive applicability in many fields, for example, image processing of leucocyte count in medical field. So, the technique has been paid much attention both in China and overseas countries.The main work and innovation points in this dissertation are as follows:(1) An improved optimized algorithm OTSU is introduced. With this algorithm, image is divided in order to obtain the target i.e particle by seperating the particles and backgrounds. This method is less affected by changes in image contrast and brightness changes and the size of the target ratio, since taking into account the within-class distance of the impact of image segmentation, so that the method has a certain practicality.(2) Having proposed the seed-point optimized algorithms based on a distance transform.Using the watershed segmentation algorithm dividing the image,we must obtain the seed-point which directly are related to the algorithm effects.In practice we find that when definiting the limit corrosion with the ordinary distance function, it has one main seed-point and more interfering seed-point i.e.redundant seed-point situation which makes the watershed segmentation over-segmentation problems.While with this optimized algorithm to obtain seedpoints, there is no redundant seed-point at large thus eliminating over-segmentation phenomenon.(3) Having designed a vertical control method,the partical long diameter and short diameter must be vertical while calculating the particle parameters, due to the image presentation factors, two lines are not absolutely vertical in most cases, thus,there must be vertical control mechanism to make the obtained short-track closest to the vertical when calculating the short-track.The short-track datas are the most representatives after adopting this mechanism.(4) The paper have designed the conversion method of the physical size and image: the ruler function method and direct calculation zoom ratio method.Because the image is only subjectively exsitence in the computer,its corresponding physical size can not be specificly measured.In the dissertation the two designed conversion methods can both accurately calculate their sizes,while the gauge ruler function method can achieve the same result as the manual measurement.