面对现有船舶图像检索存在的“语义鸿沟”问题，导致无法满足用户实际检索需求，研究集成局部和全局特征的船舶图像检索算法，实现不同条件下的船舶图像检索。通过颜色矩方法提取船舶图像全局颜色特征；通过块截断量化编码理念，采用模糊C－均值聚类算法将船舶图像分割成子块，构建子块的二值位图，表示图像局部颜色特征；结合小波变换提取船舶图像纹理特征；求解待检索船舶图像与数据库中船舶图像的各特征相似度，获取总相似度，选取总相似度最大图像作为图像检索输出结果。实验结果表明：该算法可有效提取船舶图像纹理特征；尺度变化、光照变化、旋转条件下的船舶图像检索性能较好，平均匹配率97.94%，平均匹配时间为11.9 ms，检索速度快，操作简单，能够满足用户实时性检索需要。
In the face of the semantic gap problem existing in the existing ship image retrieval, which can not meet the actual retrieval needs of users, the ship image retrieval algorithm integrating local and global features is studied to achieve ship image retrieval under different conditions. The global color feature of ship image is extracted by color moment method. Through the concept of block truncation quantization coding, the ship image is divided into sub-blocks using fuzzy C-means clustering algorithm, and the binary bitmap of the sub-blocks is constructed to represent the local color characteristics of the image. The texture feature of ship image is extracted by wavelet transform. Solve the feature similarity between the ship image to be retrieved and the ship image in the database, obtain the total similarity, and select the ship image with the largest total similarity as the ship image retrieval output. The experimental results show that the algorithm can effectively extract the texture features of ship images. The ship image retrieval performance under the conditions of scale change, illumination change and rotation is good. The average matching rate is 97.94%, the average matching time is 11.9 ms, the retrieval speed is fast, and the operation is simple, which can meet the real-time retrieval needs of users.
2023,45(5): 170-173 收稿日期：2022-10-28
DOI：10.3404/j.issn.1672-7649.2023.05.033
分类号：TP391
基金项目：四川省科技计划重点研发项目（2022YFG0206)；四川省级知识产权专项资金项目（2022-ZS-00156）
作者简介：邢伟寅(1975-),男,博士研究生,高级工程师,主要研究方向为软件工程及计算机应用
