Study Of Methods Of Particle Counting Based On Microscope Image

Particle counting can be applied in many fields, such as medical diagnosis, biological research, health monitoring and industrial production. The traditionally manual counting is not only inefficient and time-consuming, but also prolonged labor eyesight makes eye fatigue and has a serious impact on the counting results. With the development of computer graphics, more and more researchers begin to use computer software to achieve automatic particle counting, it has become a hot area of research.A counting system software consists of three main process steps, the first part is the image pre-processing, the second part is the segmentation of the object and background, the third part is splitting the overlapping objects. The second and third parts are the difficulties of the particle counting system. The segmentation results of background and target directly affect the effectiveness of the system, the splitting quality of overlapping objects affects the accuracy of the result counting. While it already has some counting algorithm and software, most of them are developed for certain particle and have strong limitations. And these software usually has low speed and accuracy.Particle counting method proposed in this paper includes two counting methods, one is for colony counting, the other is for cell.For colony counting, this paper presents a new image segmentation algorithm, which is a new iterative threshold with pre-particle filter. The segmentation method can shield effects of background, contaminants and impurities. And processing speed is high. Then, we discriminate the adhesion colonies and use watershed segmentation combining particle filter to split them accurately. Finally we get the number of all colonies. Experiments show that this method for colony counting takes less than1s to process one image, for the same type of colony, but multiple images containing different number, the average error of counting is1.89%, and for different kinds of colony, the average error is1.96%, analysis time of each image is less than1second.Since the cell and colony has many different points, such as form, size, environment, etc., cell image needs different processing and analysis method. In this paper we use HIS dual-threshold segmentation to divide cells and background and correct the result with holes-filled method. For overlapping cells, we use contour tracking method to find the separation point of the overlapping cells and get the number of cells. Experimental results show that the average error of method of cell counting is less than2.5%, and the speed processing one image is2.4s.The particle counting system of the paper realizes the automatic counting of colony and cell, the result is very accurate and the speed is quick. It has been used in microscopic processing software.