为了在海量舰船网络数据库中准确获取目标数据，提高数据库目标数据检索的成功率与检索精度，基于哈希学习算法，对舰船网络数据库目标数据检索方法展开研究。通过深度卷积神经网络与哈希学习的有效结合，构建舰船网络数据库目标数据检索模型，利用深度卷积神经网络提取舰船网络数据库中的数据特征，通过哈希学习层获取舰船网络数据库中数据特征的检索匹配哈希码，并在检索模型中引入加权余弦三元组损失函数完成模型训练，将目标数据作为训练后的模型输入，通过匹配输出的目标数据特征哈希码和舰船网络数据库中数据特征检索哈希码，实现舰船网络数据库目标数据检索。实验表明：该方法可实现舰船网络数据库目标数据检索，检索成功率为100%，可获取准确的经纬度数据；对图像以及文本数据库目标数据均值平均精度最高为82%、78%，检索精度较高，检索的实际应用性能较高。
In order to accurately obtain target data in massive ship network databases and improve the success rate and retrieval accuracy of database target data retrieval, a research was conducted on target data retrieval methods for ship network databases based on hash learning algorithms. Through the effective combination of deep convolutional neural network and hash learning, the target data retrieval model of the ship network database is constructed, and the data feature extraction in the ship network database is realized using the deep convolutional neural network. Based on the obtained features, the retrieval matching hash code of the data feature in the ship network database is obtained through the hash learning layer, and the weighted cosine triplet loss function is introduced in the model to complete the model training. Finally, the target data is used as the input of the trained model, and the target data retrieval in the ship network database is achieved by matching the output target data feature hash code with the data feature retrieval hash code in the ship network database. The experiment shows that this method can achieve target data retrieval in ship network databases, with a success rate of 100%. It can obtain accurate longitude and latitude data, and the average accuracy of image and text database target data is the highest at 82% and 78%, with high retrieval accuracy and practical application performance.
2023,45(17): 182-185 收稿日期：2023-06-07
DOI：10.3404/j.issn.1672-7649.2023.17.037
分类号：TM391.41
作者简介：许自龙(1981-),男,硕士,讲师,研究方向为计算机技术及数据库技术
参考文献：
[1] 蔡梦玲. 基于贝叶经信息资源的特色数据库建设[J]. 图书馆杂志, 2021, 40(2): 111-116.
[2] 张天, 靳聪, 帖云, 等. 面向跨模态检索的音频数据库内容匹配方法研究[J]. 信号处理, 2020, 36(6): 966-976.
[3] 罗鹏程, 王继民, 王世奇, 等. 基于深度学习的科学数据集检索方法研究[J]. 情报理论与实践, 2022, 45(7): 49-56.
[4] 宁海涛. 基于分块主色法的图像无序激增数据检索仿真[J]. 计算机仿真, 2021, 38(1): 375-379.
[5] 赵英豪, 吕亮, 徐青, 等. 一种面向海量时空数据的多维检索策略[J]. 测绘科学, 2020, 45(6): 199-204.
[6] 薛朝辉, 张瑜娟. 基于卷积核哈希学习的高光谱图像分类[J]. 遥感学报, 2022, 26(4): 722-738.
[7] 聂秀山, 刘兴波, 袭肖明, 等. 基于相似度驱动的线性哈希模型参数再优化方法[J]. 软件学报, 2020, 31(4): 1039-1050.
[8] 黄雅婕, 翟俊海, 周翔, 等. 基于哈希学习的投票样例选择算法[J]. 计算机应用, 2022, 42(2): 389-394.
