Design And Implementation Of Algorithm Of Schistosome Eggs Images Recognition

With the development of computer technology, medical microscopic image processing andrecognition is one of the most important motivations to promote biomedical engineering, which notonly provides a reliable and efficient method for clinical diagnosis, but also develops the medicalscientific research and teaching. For schistosome egg’s microscopic image has impurities andcomplex background, it is hard to process and identify. Based on the existing researches onmicroscopic image identification for parasites and cells, the paper studies the egg image segmentation,feature extraction, selection, classification and recognition method with image processing and patternrecognition technology.First of all, threshold segmentation and edge segmentation methods for eggs are respectivelydiscussed in the paper. In binary images gained by threshold segmentation method, pixel stainingmethod is proposed to sign pixel cluster. The conventional edge dictation factor cannot be applied inmicroscopic images for it has poor anti-noise property. Therefore, level set-based C-V model is usedto detect egg’s edge in the paper.Secondly, with image segmentation, characteristic parameters that are useful to identifyschistosome eggs are extracted. To analyzed egg’s edge information with quantitative analysis method,apart from geometrical characteristic, margin curvature is calculated with multi-scale curvature-basedmethod.Thirdly, the main steps of feature selection approach are estimated, and SFS is employed to selectthe proper ones from10original feature sets.Then, the pattern recognition technology is also researched in the paper. Support vector machine isused as the classifier, and the influences of gaussian kernel function and penalty factor are evaluated.Finally, these parameters are determined by experiments.At last, schistosome recognition system on windows platform is designed and programmed withOpenCV, whose functions contain image processing and pattern recognition. With plenty of test, theoperation and recognition results show that the system is stable and reliable.