摘要
为了解决藻类分类识别中人工选取特征困难的问题,提出了一种基于深度学习的藻类分类识别方法。首先,对训练和测试样本集数据进行处理,得到所需数据的格式;其次,研究各种深度学习模型,理解卷积层、全连接层等的作用,基于Caffe设计深度学习网络模型;最后,根据设计的深度学习网络模型,比较各个模型的性能,得到最好的模型。实验结果表明,使用该方法做藻类分类,优于张松等基于视觉词包模型训练SVM分类器的方法,得到比较理想的效果。
In order to solve the problem of difficult manual feature selection in algae classification and recognition,a method of algae classification and recognition based on deep learning is proposed. Firstly, the data of training and testing samples are processed to get the format of the data needed. Secondly, various deep learning models are studied to understand the roles of convolution layer and full connection layer, and a deep learning network model is designed based on Cafe. Finally, according to the designed deep learning network model, the performance of the models are compared, and the best model is obtained.The experimental results show that this method is superior to the other methods based on visual word package model such as Zhang Song to train SVM classifiers, and achieves better results.
引文
[1]郭春锋.中国海常见有害赤潮藻显微图像识别研究[D].青岛:中国海洋大学, 2011.
[2]张松,周亚丽,张奇志.词包模型在藻类分类识别中的应用[J].北京信息科技大学学报, 2016, 31(1):28-32.
[3]张帆.人工镜检难点与AlgacountTM藻类监测技术优势分析[A].见:第二届全国藻类多样性和藻类分类学术研讨会论文集[C]. 2010.
[4]赵文仓,姬光荣,周立俭,年睿.浮游植物细胞图像识别方法的研究[J].计算机工程, 2015, 31(24):143-144.
[5]王迪.基于线性判别分析的赤潮藻类流式图像自动识别研究[D].厦门:厦门大学, 2014.
[6]史加荣,马媛媛.深度学习的研究进展与发展[J].计算机工程与应用, 2018, 54(10):1-10.