基于主成分特征的相似微藻分类算法的研究

Study on classification algorithm of similar microalgae based on principal component features

海洋微藻是一种光合自养型生物,随着水体富营养化日益严重,藻类大量繁殖逐渐形成水华和赤潮,如何快速准确识别有益藻和有害藻是当务之急。但是许多藻类形态彼此相近,难以分辨。为了解决这个问题,文中首先将输入的图像进行标准化处理,计算图像第一主成分载荷特征;其次设计一个基于逻辑回归的二分类模型,分两类标签进行训练;最后构造代价函数以及sigmoid函数,根据梯度下降法得出的预测结果与实际情况对比,计算最终准确率为92.86%。与广泛使用的藻类二分类算法SVM分类器比较结果显示,在相似藻的分类精度上提高了1.86%。

Marine microalgae is a kind of photosynthetic autotrophic organism.With the increasing severity of algal bloom,algal blooms and red tides are formed gradually.Therefore,how to quickly and accurately identify beneficial and harmful algae is an urgent task.In order to solve this problem that many algae are indistinguishable from each other,the paper proposes several processes.Firstly,the input image is standardized and the first principal component load is calculated.Secondly,a binary classification model based on logistic regression is designed,and two kinds of labels are trained.Finally,the cost function and sigmoid function are constructed,and the accuracy of the gradient descent is 92.86%.Compared with the widely used binary algal classification algorithm SVM classifier,the result shows that the classification accuracy of similar algal is improved by 1.86%.