QQ客服 官方微博 反馈留言

题名	基于BRISK算法的图像拼接技术研究
作者	陈思聪
学位类别	硕士
答辩日期	2015-05
授予单位	中国科学院大学
导师	刘晶红
关键词	图像拼接 BRISK特征 RANSANC算法 模板匹配
其他题名	Research on Image Mosaic Based on BRISK Algorithm
学位专业	光学工程
中文摘要	为了对较大的目标场景获取较宽视角的图像，在机器视觉领域通常有两种方法，一种方法是通过缩短相机焦距来摄取完整的场景，这种方法会降低图像的分辨率；另一种方法是采用鱼眼镜头、广角镜头、反/折射系统等，该方法需要复杂的硬件设备，价格昂贵，在图像边缘容易产生较大的扭曲变形。 随着计算机技术的发展，图像拼接技术成为了一种新的解决方法。图像拼接是通过计算机技术对两幅或者两幅以上具有重叠区域的图像进行自动配准对齐，再通过图像融合技术生成一幅等效的具有宽视角的高分辨率图像，该方法降低了对摄像设备的硬件要求，减小了设备体积，具有自动化、智能化、低成本等优点，因此本文对图像拼接技术进行了深入研究。 论文对图像拼接技术的发展现状进行了调查研究，对图像拼接技术进行了概括和总结，分别分析和阐述了组成图像拼接的三个重要关键步骤:图像预处理、图像配准以及图像融合。深入分析了Harris，SIFT，SURF，BRISK，FAST，AGAST等具有经典意义的图像特征匹配算法，为后续研究奠定了理论基础。 针对传统的基于特征的图像拼接方法的缺点，本文提一种基于改进BRISK算法的图像拼接方法。传统的BRISK算法在拼接平移方向上存在较大的误差。针对该问题，首先通过FAST角点提取出图像的特征点，采用BRISK描述采样器进行特征匹配，引入RANSANC算法剔除误匹配对，实现图像尺度和旋转的校正，再利用互相关模板匹配进行图像的精确匹配，完成平移校正，最后经图像融合实现图像的精准拼接。 本文在两幅图像重叠区域的选择若干同名点对作为十字监测点，计算拼接图像中对应监测点像素坐标距离的均方根误差，对图像拼接精度进行了量化分 析。对标准数据库图像的实验结果表明，本文算法是一种运算时间短、精确度高、拼接效果良好的图像拼接方法。
英文摘要	There are two approaches generally used to obtain wide angle of view images for the large object scenes in the computer vision community.One method is shortening the camera's focal length to shoot the complete scene, but it will reduce spatial resolution of images.The other method is useing fisheye lens, wide-angle lens and reflection/refraction optical system, this method need complex and expensive hardware devices, and easily produce larger distortion near the edge of an image. With the development of computer technique, the technology of image mosaic offers a new idea to solve these problems. Image mosaic is to use the image automatic registration techniques to align two or more images with overlapping area, then use the image fusion technology to generate an equivalent image with wide viewing angle and high image resolution. This method decreases the hardware requirement of cameras, narrows the equipment, and has the advantages of intelligence, automation, and low cost. Therefore, the algorithms of image mosaic are studied in this article. This thesis investigated the developing status and trends of image mosaic technology, and the theory of image mosaic is generalized and summarized. The key techniques, such as the image preprocessing methods, the image registration algorithms, and the image fusion technology are analyzed and expounded. The classic image feature matching algorithms, such as Harris, SIFT, SURF, BRISK, FAST, and AGAST, are analyzed in-depth, and this laid the foundation for further work. To get accurate space stitching images, this paper proposes an algorithm combined with BRISK and cross-correlation template matching algorithm. There are some considerable errors in the splicing translational direction, when using traditional BRISK algorithm. To counter this problem, BRISK algorithm can be used for the image of the scale and rotation correction .Then the template matching method can be used to solve the translation correction.At the same time, RANSANC algorithm is added to the BRISK algorithm to achieve precise stitching. Some of the nomonymy points are picked out as cross monitoring points in the overlapping region.Then calculating the root-mean-square error of the distance between pixel coordinates of corresponding monitoring sites, and the image matching precision can be quantified.The experimental results with database images show that it is a kind of short operation time, high precision image matching method with good results.
公开日期	2015-12-24
内容类型	学位论文
源URL	[http://ir.ciomp.ac.cn/handle/181722/48827]
专题	长春光学精密机械与物理研究所_中科院长春光机所知识产出
推荐引用方式 GB/T 7714	陈思聪. 基于BRISK算法的图像拼接技术研究[D]. 中国科学院大学. 2015.

个性服务

查看访问统计

相关权益政策

暂无数据

收藏/分享

除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。