计算机应用与软件2017，Vol.34Issue(9)：57-63,7.DOI:10.3969/j.issn.1000-386x.2017.09.012
构建适用于深度学习的海浪样本数据集的并行算法实现及性能优化
PARALLEL IMPLEMENTATION OF WAVE SAMPLE DATA SET FOR DEEP LEARNING ALGORITHM AND ITS PERFORMANCE OPTIMIZATION
摘要
Abstract
The deep learning algorithm is used to achieve the classification of ocean waves,data from Yang Shan Port video monitoring and synchronous wave measurement.Aiming at the problem that the construction of the wave sample data set in the image processing part has large computation in the offshore wave recognition system,a parallel computing scheme for image processing of ocean wave sample dataset is designed.When building a wave sample dataset,the key frame was extracted from the video,and the weighted mean filter denoising was used to generate a corresponding label of the wave level.Hence,the construction of the ocean wave sample data set was realized.We adopted OpenMP to perform parallel algorithm simulation for wave sample data set image preprocessing,and improve the performance optimization of related code.Experimental results show that the proposed parallel algorithm greatly improves preprocessing speed and utilization ratio of multi-core over a serial algorithm.Its speedup ratio can reach 24.29 as the thread is increased to K =8.We conclude that the algorithm has good scalability,high performance,convenient use,inexpensiveness and good practical value.关键词
深度学习/海浪样本数据集/关键帧/加权均值滤波/并行优化/OpenMPKey words
Deep learning/Wave sample data set/Key frames/Weighted mean filtering/Parallel optimization/OpenMP分类
信息技术与安全科学引用本文复制引用
邹国良,陈长吉,郝剑波..构建适用于深度学习的海浪样本数据集的并行算法实现及性能优化[J].计算机应用与软件,2017,34(9):57-63,7.基金项目
国家自然科学基金项目(11402142). （11402142）
