Preprocessing And Segmentation For Microstructure Images Of Mongolian Flower Herb

Traditional Mongolian medicine is one of most important parts in Mongolian cultural heritages, and also is one of bright spots in eastern traditional medicinal treasury. Mongolian Herbs are utilized more and more in the modern medicine today. However, the precise and impersonal identification methods was imperative for powdered Mongolian Herbs, because of the variations caused by different production spots, different harvest seasons, different storing and processing methods of Herbs, and coming forth of some fake medicines. Almost all of preparations of Mongolian traditional medicines were made of several to dozens of powdered Herbs, therefore, it is difficult to identify the exact components of the preparations according to their appearances. For developing and improving the identification and certification methods of traditional Mongolian medicine, commonly used Mongolian Herbal Flowers, were emulated and identified by using Pseudo-Jacobi (p=4, q=3)-Fourier Moments (PJFM’s), image processing and pattern recognition in this paper on the basis of skillfully combination of advanced computer techniques and traditional microscopic identification methods-according to the distinct microscopic characteristics of the Herbal flowers and their specificity and stability of the tissue structures. Original image database of more than ten species of Mongolian Herbal Flowers were established using their main microscopic characteristics, and these microscopic characteristic images were digitalized and reconstructed in our study.Preprocessing of herbal flower image is the important precondition for micro-images recognition and identification. It is important for computer-aided plant identification to get a high-quality micro-image of herbal flowers which is preprocessed. Preprocessing method of herbal flower image based on mathematical morphology with fusion of other methods is advanced in this paper. Dilating and eroding algorithm are used for eliminating isolated points in the image and filling holes in pollen grain. At the same time, microstructure images of Mongolian herbal are complicated background multi-target images. One simple segmentation algorithm can’t obtain ideal auto segmentation results. So we developed semi-automatic segmentation method based on Matlab function. We use roipoly in MATLAB to specify a polygonal region of interest (ROI) within ideal target image. Then roipoly returns a binary image that can be used as a mask to multiply with the original image to select the object. By using this method, a multi-target herbal flower image and a single target herbal flower image are processed separately. And the basic shape features of original image are preserved and clear edge of pollen grain is detected. Thus favorable precondition is created for the object selection of pollen grain, and may pave the way for accomplishing automatic check and microscopic identification. This study results can supply a gap for the digitalization and visualization of Mongolian Herbs, and establish an important foundation for globalization of Mongolian Herbs.