浮游植物细胞图像识别方法的研究

Study of Recognition Method of Phytoplankton Cell Image

讨论了如何利用数学形态学方法提取浮游植物细胞图像的面积(A)、周长(P)以及通过图像细化提取细胞的长(L)等特征信息;提出了利用这3个特征参数的比值(A/P,A/L和P/L)形成的特征向量作为进行细胞分类的依据,然后提出了利用最近邻准则对浮游植物进行自动识别的试探聚类算法,该方法可以通过学习提高识别率。通过对17种浮游植物细胞图像的识别实验,证明了该方法的识别准确率在95%以上,可以有效预防大面积赤潮的发生,对于赤潮爆发后的鉴定也有一定意义。

This paper talks about how to compute the phytoplankton cell image＇s area （A） and perimeter （P） using the mathematical morphology method, and how to calculate the length （L） and other characteristic information using the image thinning method; puts forward a way to classify the phytoplankton cell by the eigenvector which consists of the specific value of three choJ＇acteristic parameters, such as A/P,A/L and P/L; then presents a tentative auto-recognizing clustering method of the phytoplankton using the nearest neighbor rule, and the method may enhance the recognition ratio through learning. According to the recognition experiments of 17 sorts of the phytoplankton cell image, the recognition rate of this method is over 95%. This method could effectively prevent the extensive area red tide and should have significance to identify the red tide that has broken out.