The Micro Image Segmentation And Stomata Index Measurement Of Leafy Herbal Medicine

The identification of the Chinese herbal medicine will contribute to the discovery of the new medicine and the crack-down on the fake and shoddy medicine.In the light of the low efficiency of the traditional method of identification of the Chinese herb,this paper discovered the new ways of identification and presented two feasible identification approaches on the basis of image segmentation.The first approach is the micro image segmentation of leafy herbal medicine based on feature space clustering.And the second approach is the micro image segmentation of leafy herbal medicine based on threshold segmentation and the target counting method.These two approaches found solutions to the various difficulties in the micro image segmentation of leafy herbal medicine,such as the huge amount of noise disturbing the segmentation,and uneven of light,the existence of the incision and abrasion,great number of categories and uneven tincture.In the first study of the approach,in order to solve the problem of hard to identify the initial clustering center in feature space clustering technique,this essay brought forward the K-means clustering segmentation counting method based on kernel density and the maximum distance in the intra-class.The counting method used a special storage structure to store the data of color space,sort the colors in the image according to the density features of the colors,and cluster accordingly,and then decided whether it belonged to the existing cluster or formed a new cluster according to whether the distance between the color waiting to be sorted and the center of existing cluster is smaller than the maximum distance.The second approach is original and creative to some extent in that it made full use of the hypothesis that the image edges and the target pixels are normally distributed to solve the single-peak problem during the threshold segmentation.After the gray level transformation of the image,removal all sources of noise and edge detection,the threshold value will be set.And use the image edge information to get the threshold value of the image segmentation,and divide the target area without the edge area.Most non cellular target and cell hole will be eliminated through removing of the small area and hole-filling technique;the division and counting of the cells will be achieved through marking the area and calculating the perimeter of every area.The segmentation of stoma will be based on its distinct features from other targets,and according to the hypothesis that the target pixels and backing pixels are normally distributed,the single-peak image is segmented,thus the problem of the target being too small and no apparent trough in the image histogram.The technique brought forward in this paper solved the problem of the segmentation of micro image of leafy herbal medicine and the measurement of micro constant,and was significant in finding some solutions to problems existing in the current image segmentation.