机器人双目图像实时立体匹配算法

发布时间：2018-04-12 14:08

本文选题：双目图像 + 立体匹配　；参考：《北京理工大学学报》2017年09期

【摘要】：针对移动机器人平台上的双目视觉深度信息获取问题,研究了一种适用于动态图像序列的双目图像立体匹配算法.采用成熟的半全局立体匹配算法对图像快速立体匹配;分析了在复杂环境下的运动平台上完成立体匹配所面临的主要噪声干扰,包括白斑噪声和断层闪烁噪声;为此,分别提出了白斑滤波算法和视差图时间域滤波方法以抑制这两种噪声.试验结果表明,所提出的算法能够有效抑制动态双目图像在立体匹配时的噪声干扰,最终能够实时获得连续清晰的视差图序列.
[Abstract]:Aiming at the problem of obtaining binocular visual depth information on mobile robot platform, a binocular image stereo matching algorithm suitable for dynamic image sequence is studied.A mature semi-global stereo matching algorithm is used for fast stereo matching of images. The main noise interference, including white spot noise and fault flicker noise, is analyzed when stereo matching is completed on a moving platform in complex environment.White spot filtering algorithm and parallax time domain filtering method are proposed to suppress these two kinds of noise.The experimental results show that the proposed algorithm can effectively suppress the noise interference of dynamic binocular images in stereo matching, and finally obtain continuous and clear parallax image sequences in real time.
【作者单位】：哈尔滨工程大学自动化学院;
【基金】：国家自然科学基金资助项目(51409053、61175089) 黑龙江省自然基金资助项目(E201414)
【分类号】：TP242;TP391.41

【相似文献】