您的位置:首页 > 论文页面
基于快照集成卷积神经网络的陨石坑图像分类识别
发表时间:2022-03-31 浏览量:854 下载量:107
全部作者: | 熊月容,康志伟 |
作者单位: | 湖南大学信息科学与工程学院 |
摘 要: | 为提高对陨石坑图像进行分类识别的性能,提出了基于卷积神经网络(convolutional neural network,CNN)集成的陨石坑图像识别方法。利用余弦退火算法将单一的卷积神经模型进行快照集成,构建了快照集成模型。模型将同一训练过程中的不同结点的模型进行集成,既保证了训练模型的多样性,又避免了集成模型训练成本的增加。采用快照集成的CNN模型在陨石坑图像数据集上的实验结果表明,该方法是一种高效准确识别陨石坑图像的深度学习方法。 |
关 键 词: | 航空、航天科学技术基础学科;信号与信息处理;卷积神经网络;集成学习;陨石坑 |
Title: | Classification and recognition of crater images based on snapshot ensemble convolutional neural network |
Author: | XIONG Yuerong, KANG Zhiwei |
Organization: | College of Computer Science and Electronic Engineering, Hunan University |
Abstract: | In order to improve the performance of classification and recognition of crater images, a crater image recognition method based on convolutional neural network (CNN) ensemble is proposed. Single convolutional neural models are integrated to construct a snapshot ensemble model by using cosine annealing algorithm. The model integrates the models of different nodes in the same training process, which not only ensures the diversity of training models, but also avoids the increase of training costs of ensemble models. Using the snapshot ensemble CNN model, the results of experiment on the crater image dataset show that this method is a deep learning method for efficiently and accurately recognizing crater images. |
Key words: | basic subject of aviation; signal and information processing; convolutional neural network; ensemble learning; crater |
发表期数: | 2022年3月第1期 |
引用格式: | 熊月容,康志伟. 基于快照集成卷积神经网络的陨石坑图像分类识别[J]. 中国科技论文在线精品论文,2022,15(1):104-111. |

请您登录
暂无评论