Long Distance Sea Surface Images Fast Sparse Matching by Integrated Feature Vector
Long Distance Sea Surface Images Fast Sparse Matching by Integrated Feature Vector
复制标题
DOI：
10.1007/978-981-32-9441-7_86
复制
发表时间：
2019
期刊：
影响因子：
--
通讯作者：
Ying Yang;Cunwei Lu;Chenhao Chen
中科院分区：
文献类型：
--
作者：
Ying Yang;Cunwei Lu;Chenhao Chen
对应匹配是双目立体视觉系统的一项基本任务。针对长距离海面图像无明显结构和颜色分布特征，同时对处理精度和速度要求较高的情况，提出一种快速备件对应匹配方法。针对缺乏明显特征的问题，本文从海面图像中提取海浪作为特征点，采集海浪四个方向的大小、形状、亮度、位置和长度，实现描述子融合，形成特征向量。为了快速准确地完成立体匹配，本文利用余弦相似度来匹配第一步生成的特征向量。通过最终的3D测量结果进行对应性验证。在各种真实海面图像对上进行实验来验证所提方法的有效性。我们的方法与一些最先进的方法进行了比较，以显示我们的方法在精度方面的优势。
Correspondence matching is a fundamental task of stereo binocular vision system. This paper proposes a fast spares correspondence matching method used for long distance sea surface images which possesses no obvious structure and color distribution features, simultaneously, requires high processing accuracy and speed. To solve the problem of lacking obvious features, the paper extracts sea waves out from sea surface image as feature points and gathers sea waves’ size, shape, brightness, location and lengths in four orientations to realize descriptors fusing and form feature vector. To complete stereo matching fast and accurately, the paper utilize cosine similarity to match feature vectors generated in first step. The correspondence verification is carried out by the final 3D measurement result. Experiments on various real sea surface image pairs are operated to validate the effectiveness of proposed method. The comparisons between our method and some state-of-the-art methods are carried out to display the advantage of our method in precision.