硅藻细胞显微图像骨架树匹配方法研究

Research for Skeleton Tree Matching of Microscopic Image of Diatom Cells

基于形态学差异的图像分析技术近年来成为藻种鉴定的一种重要手段。为了识别硅藻门内角毛藻属显微图像,提出了一种骨架树相似度匹配方法。构建目标曲面矢量模型,在保留微弱角毛信息的前提下精确分割角毛藻细胞;利用骨架表征目标,采用竞争式跟踪策略将目标骨架层次性分解,根据拓扑关系将各基元映射至骨架树结构;通过构建细胞拓扑和局部特征将角毛藻辨识任务转化为骨架树相似度匹配问题,实验证明该方法对常见的若干种角毛藻有较高的识别准确率。

The image analysis technology based on morphological differences has become an important method of algae recognition in recent years. A skeleton similarity matching method was proposed to recognize microscopic images of diatom cells. The gray surface vector model of image was established to make segmentation by keeping obscure seta; The skeleton was used to represent the Chaetoceros, and it was decomposed hierarchically by competition strategy. The Chaetoceros object was represented by skeleton tree by forming the rachis elements and branches into it. The similarity mode for microscopic images of Chaetoceros was established by defining the topological and geometric difference. Experimental results show that this algorithm can achieve the better recognition of several kinds of Chaetoceros.