Scanning Microscopic Particle Image Recognition System Research And Development

As image processing technology in the computer field has continuous researched in recent years, development of microscopic image became faster, particle image recognition get more attention to people gradually. Particle identification refers to extract the characteristics of particles and analysis in the particle image, In the traditional method for identifying particles slide, microscopy is currently one of the popular particle identification method, but microscopy exist limitations in the process of identifying, such as the industrial cameras is commonly used as microscopy camera, which CCD size is small, shooting the picture information is not rich, there are less particles in each image. Due to the three-dimensional particle being filmed for the two-dimensional images, there is overlap between particles, need split on adherent particles before recognition. But each kind of segmentation algorithm is suitable for limited kinds of particles, is not entire, if segmentation algorithm is used unreasonably, would influence on particle identification directly.Aiming at the shortcomings of the current particle identification methods, with a frame based on the combination of CMOS camera and lens, while image acquisition, image matching, image data processing and particle image recognition as four modules for architecture, scanning microscopic image particle recognition system is proposed.The key of Scanning microscopic image particle recognition system is image matching and image data processing. Image matching algorithm is mainly to the image feature extraction and matching, aimed at the deficiencies of the present algorithm in the situation of feature extraction is slow, characteristics describe is complex, improvement of SURF algorithm in the process of feature points extraction reduce the number of feature points, reduce the dimension of feature vector. With based on vector angle of matching algorithm filter characteristic vector set in the feature vector matching process, reduce the number of matching object and improve the efficiency of image matching algorithm. MIRXS file format as the format of the image display in image processing process, and through the block processing of large images, reduces the memory consumption. Use the Direct2 D technology which through hardware to speed up the display, instead of the traditional GDI software to accelerate, achieving translation, scaling, rotation, and so on. To tag the interest region in the image, and through the XML file to store tag informations. In the process of particle identification according to the image of the block, in a certain order recursive identification in large image, solving the problem of insufficient memory while eliminate the special particles. With the level set method based on the C-V edge detection, obtaining particles geometry features. Experiments proved that the scanning microscopic image particle recognition system can quickly collect particles slide, image matching effect is good, the identification efficiency is high.