基于特征级融合的多波段舰船目标识别方法

发布时间：2018-01-26 12:11

本文关键词： 图像融合区域协方差特征级融合目标识别　出处：《光谱学与光谱分析》2017年06期 　论文类型：期刊论文

【摘要】：针对复杂情况下海上舰船目标单波段特征识别能力不足的问题,研究可见光、中波红外和长波红外三波段特征图像融合技术,重点解决图像融合方法中存在的算法耗时和融合策略选择的问题,提出了一种新的基于区域协方差矩阵的多波段特征级融合方法,针对可见光图像和红外图像分别设计11维和5维特征向量,协方差矩阵可以将多个特征进行融合,既保证了不同目标之间的区别性,同时又减小计算量。该方法首先利用显著性检测,快速定位图像中的目标区域,然后,针对不同波段图像设计的特征向量定义协方差阵的距离计算公式并进行匹配,通过对图像的一次遍历操作获得积分图像,在协方差计算时达到快速计算的目的,最后利用k-阶最近邻算法对多种舰船目标进行分类识别。利用实拍的3 400余张三波段舰船目标图像作为测试数据。实验主要分为两部分,首先对比单波段和三波段融合识别的识别率,验证所提出的融合方法具有更广的应用范围;然后,在计算效率上对比多种传统的像素级方法,验证采用的特征级融合在计算时间上的优势。实验结果表明,该方法可达到95.1%的识别率,单帧计算耗时约为0.5s,在实时性和检测率方面都有明显提高。
[Abstract]:Aiming at the problem that the capability of single band feature recognition of marine ship targets is insufficient in complex situations, the technology of feature image fusion in visible, medium and long wave infrared bands is studied. In order to solve the problem of time-consuming and strategy selection in image fusion, a new multi-band feature level fusion method based on regional covariance matrix is proposed. For visible image and infrared image, 11 and 5 dimensional feature vectors are designed respectively. Covariance matrix can fuse multiple features, which ensures the distinction between different targets. At the same time, the computational complexity is reduced. Firstly, the salience detection is used to locate the target region of the image quickly, and then the target area in the image is quickly located. The distance calculation formula of covariance matrix is defined and matched according to the eigenvector of different band image design. The integral image is obtained by one traversal operation of the image, and the purpose of fast calculation is achieved when covariance is calculated. Finally, we use k- order nearest neighbor algorithm to classify and recognize various ship targets, and use more than 3 400 real-shot images of three-band ship targets as test data. The experiment is mainly divided into two parts. First, comparing the recognition rate of single-band and three-band fusion recognition, it is verified that the proposed fusion method has a wider range of applications. Then, compared with many traditional pixel level methods in computational efficiency, the advantages of feature level fusion in computing time are verified. The experimental results show that the recognition rate of this method can reach 95.1%. The computation time of single frame is about 0.5 s, and the real time and detection rate are improved obviously.
【作者单位】：海军航空工程学院控制工程系;中国国防科技信息中心;91206部队;
【基金】：国家自然科学基金项目(61303192)资助
【分类号】：TP391.41
【正文快照】： 2.中国国防科技信息中心,北京1001423.91206部队,山东青岛264001引言图像融合技术是光信息处理和计算机视觉领域的重要研究内容,利用多源传感器或同一传感器在不同时刻拍摄的图像,将其中有用的信息进行结合可以克服单一传感器在几何、光谱和空间分辨率等方面的局限性和差异性,

【参考文献】