| Particles are one of the most important forms of chemical substances.Particle size has an important influence on the properties of chemical substances.Therefore,particle size analysis technology has gradually developed into an important branch of measurement.With the rapid development of image acquisition and processing technology,the method of particle size analysis based on image processing has attracted wide attention.This method is fast and accurate and it can be applied to the measurement of different particle sizes and shapes.Euler number,as an important topological property of image,is often used to describe the structural characteristics of image.It is widely used in image understanding and pattern recognition,computer vision and other fields because it remains unchanged when the image is stretched,contracted,rotated and irregular elastic deformation.It has a strong robustness property.This dissertation mainly focuses on the optimization of Euler number algorithms of a binary image and the its application in paper fillers particle size analysis.Firstly,the factors affecting the computational efficiency in conventional Euler number computing algorithms are analyzed in detail,and effective strategies such as state transition,multi-row scanning and interlaced scanning are used to reduce the repeated accesses of pixels,so as to improve the computational efficiency of algorithms.Then,Euler formula in graph theory is applied to reduce the number of bit-quad types used in conventional Euler number algorithm and an optimized Euler number computing algorithm is proposed.Finally,the optimized Euler number computing algorithm combines with connected component labeling algorithm is applied to the particle size analysis of paper fillers.Thus,a convenient and accurate method for particle size analysis of paper fillers is explored which will provide technical support for particle measurement and analysis based on image processing.The main research results are as follows:(1)In conventional run-based Euler number computing algorithm of a binary image,the number of runs in each row and the number of neighboring runs in adjacent rows are obtained by scanning the image row by row,and the difference between them is calculated to obtain the Euler number of the given image.In order to overcome the problem of redundant storage of runs' position in conventional run-based Euler number computing algorithm,an image scanning method with two interlaced scans is proposed.Firstly,it scans the odd rows of pixels in the image,searches and counts the runs,records their end position in order.Then,it scans the even rows of pixels in the image,searches and counts the runs.At the same time,neighboring runs in the adjacent rows of the processing run needs to be found.The experimental results demonstrated that the improved algorithm is more efficient than the conventional run-based computing algorithm.(2)In the conventional bit-quad-based Euler number computing algorithm,the number of specific types of bit-quads need to be counted in the image.The order of scanning image and processing pixels is from left to right and from top to bottom.For the processing pixel,it is necessary to access the other three pixels in the corresponding bit-quad to determine whether the bit-quad is needed to be counted.Inevitably,there are redundant accessed pixels.In the dissertation,effective strategies of state transition and multi-row scanning is proposed to reduce the redundant accesses of pixels from horizontal and vertical directions in the given image in order to improve algorithm efficiency.Theoretically,the average number of pixels to be accessed for processing a bit-quad would decrease with the increase of the number of bit-quads being processed simultaneously.At the same time,the more bit-quads being processed simultaneously,the more states need to be considered for state transition and the algorithm will be complicated.Thus,the algorithm will be less efficient in implementation.The balance between the maximum number of rows in each scan and the complexity of the algorithm is explored in the dissertation and the optimal number of rows in each scan is verified by experiments.Lastly,an efficient bit-quad-based Euler number computing algorithm is obtained.(3)In this dissertation,4-neighborhood graph-based Euler number computing algorithm is extended to 8-neighborhood for computing the Euler number of a given image.The given image is transformed into a graph according to certain rules and according to Euler's formula in graph theory,the number of nodes,edges and fundamental squares in the graph corresponding to the given image is used to calculate the Euler number.In view of the fact that the poor efficiency of directly counting nodes,edges and fundamental squares in the graph,they are replaced by processing the bit-quad in the given image in the proposed algorithm.The increment of each bit-quad to the Euler number is analyzed,and the bit-quads influencing the Euler number of images in practice are explored.At the same time,the state transition and multi-row scanning strategies are used for reducing the redundant accesses of pixels,thus an efficient graph-based Euler number computing algorithm is proposed.(4)In this dissertation,image processing method is used to analyze the particle size distribution of paper fillers.The filler particles image obtained by scanning electron microscopy is preprocessed,and the connected component labeling algorithm and Euler number computing algorithm are used to identify and analysis the filler particles in the image.Then,the particle size distribution of fillers is obtained by calculating.Finally,we compare the results obtained by image processing and the results measured by laser diffraction,analyze the causes of the differences.Thus,we can provide technical support for the correlation between filler size characteristics and paper performance in the paper fiber network structure,and provide ideas for further improvement of the particle size analyzing method.The experimental results on different types of images demonstrated that the improved Euler number algorithm is more efficient than the conventional Euler number algorithms.When computing the Euler numbers of different types of images,the most efficient algorithm can be selected to improve the overall performance of real-time image processing system.Furthermore,the particle size analysis method based on image processing proposed in this dissertation is consistent with the results obtained by laser diffraction measuring method when it is used to calculate the particle size of PCC filler with an enlargement of 1500 times.The median diameter error between two methods is less 1%.The automatic particle analysis method based on image processing can not only greatly reduce the cost,but also improve the speed and accuracy of counting and analysis greatly.The research results of this subject can be applied to particle analysis in different fields.By improving existing algorithms to enhance the robustness to particle types,it can provide technical support for real-time image analysis system in corresponding fields. |
PDF Full Text Request