Research And Implementation Of Blood Cell Recognition And Classification

With the development of science and technology, digital image processing has been applied to various fields; especially plays an important role in medicine. This paper mainly study recognition and statistics of blood cells, and use GUI function in MATLAB to design a Software User Interface.First, we do some pretreatment for blood cell image. Using 3×3 median filtering to do image de-noising, analyzing the color space of RGB and HSV, comparing the method of common edge detection. And then we present a method called mathematical morphology. We use mathematical morphology to deal with some blood cell images, so that it can improve accuracy rate of cell recognition and statistics.Second, we analyze the biological features and geometric features of blood cells. Biological features mainly refer to shape and quantity. Geometric features mainly refer to perimeter, area, long and short axis, equivalent diameter, circularity and rectangularity. After contrastive analyzing the data of biological features and geometric features, we can decide whether these cells are white blood cells or red blood cells. It’s very useful for recognizing, separating and counting.Third, we use two methods to recognize blood cells. The first method is to use threshold segmentation. We analyze 6 color components combine with spectrum diagram to determine the scope of the threshold value. The second method is to use support vector machine. This method is to choose some characteristic vectors, and take what we need. After comparing and analyzing these two methods, we find that the first method is more complex, the second method is relatively simple. However, the second method can extract independent parts when we recognize cells.Fourth, we use connected domain algorithm and watershed algorithm to count white blood cells and red blood cells. After that, we contrastive analyze the advantages and shortcomings of these two algorithms, and improve them. After improving two algorithms, the accuracy of counting white blood cells can reach 100%, the accuracy of counting red blood cells can reach over 90%. After comparing two methods, we find that watershed algorithm is better.Finally, we design Graphical User Interface of MATLAB software. This software interface contains all the treating processes of picture processing which process blood cell images. It includes pretreatment, feature analysis, segmentation, recognition, counting and etc. This GUI is simple and intuitive, easy to operate and running smoothly. Users needn’t understand or be familiar with how does the program run inside. They can easily realize the man-machine interactive function, complete the mission what they want to do.