有害赤潮显微图像识别的三级分类器设计

The Design of Three Levels Classfiers on Microscopic Images Recognition of Harmful Algae Blooms

传统的赤潮藻鉴定主要依赖经验丰富的藻类专家依据种的生物形态学特征在显微镜下通过人工目视判读、分类,该方法存在专业水平高、分类人员断层、耗时费力等问题。基于赤潮藻类细胞生物形态学分类特征,通过对藻类细胞3种细节特征(有无角毛、横纵沟、尖顶刺)进行有效的自动提取和描述,提出了显微图像自动分类识别的思想。设计三级两类分类器,建立树状判别体系,将大样本集有效划分为小样本集,构建赤潮藻显微图像自动诊断识别系统,结果表明:多级分类器的设计思想减少了训练时间,提高了识别准确率。

Traditionally identification of HAgs mainly relies on well-experienced algae operators using microscope based on characteristics of biological morphology, which needs high level professional experts that are becoming less and less. Based on the bio-morphological classification features of the HAg cells, the automatic classification system of the microscopic images was presented with the effective automatic extraction and description of three detailed features （seta, cingulum or sulcus, spine） of algae cells. Then we designed the three levels of two types of classifiers, and established the treelike identification system to divide the large sample set into small sets in order to build the microscopic images automatic diagnosis and recognition system. The experimental results show that the design of classifier can reduce the training time as well as improve the recognition accuracy rate.