显微图像分析技术在赤潮生物识别中的应用

为克服人工识别赤潮生物种类和数量时的困难，基于图像分析法对赤潮生物图像进行自动分析，针对赤潮生物显微图像的特点。研究图像的预处理、分割、特征提取和分类的处理过程，结合理论分析和实验结果。与其他自动分割方法进行比较，证明采用类别方差法可以有效进行赤潮生物图像的自动阈值分割，解决了赤湖生物显微图像自动分割的难题，采用形态学的方法，解决相连赤潮生物的分割问题，对获得的轮廓和灰度特征进行选择，作为分类器的测试和学习数据使用，建立分类树结构，采用树状判别机制实现几类赤潮生物的识别分类。实验结果表明，与人工识别相比，其识别正确率大于90％。

Application of Micro-Image Analyzing Technology to the Recognition of Harmful Algae

The micro-image of the harmful algae was analyzed automatically on the basis of the image analysis method. Aiming at the characteristic of the harmful algae's micro-image, this paper studied the pre-processing, segmentation, feature extraction and classification of the harmful algae's image. Compared with the other methods, OTSU was found to be effective in segmenting the harmful algae's image and solving the problem of the segmentation. By using the mathematical morphology methods the segmentation problem of the jointed algae was further solved. The features of the shape and gray were selected and used as testing and training data of the classifier. The classifier was applied to the identification of several kinds of algae by constructing the cluster tree. The result indicated that the classification accuracy reached up to 90%, compared to the artificial identification.