车载全景视觉系统的研究与实现

发布时间：2019-03-05 21:07

【摘要】：随着国内汽车使用数量的飞速增加,越来越多的交通问题开始凸显,其中发生频率最多的事故是车道拥挤和低速泊车导致的车辆剐蹭事故,而导致这一事故的原因很大程度上是因为驾驶员驾驶车辆时存在大量的视野盲区。车载全景视觉系统旨在给驾驶员提供一幅车辆四周的自上而下的全景鸟瞰图,帮助驾驶员消除视野盲区,从而降低此类交通事故的发生。本文主要针对车载全景视觉系统进行研究,介绍了鱼眼摄像机成像原理和畸变矫正方法,设计了全景视觉图像处理算法,并提出了算法的加速方法,最后进行实车实验验证,并在嵌入式平台上加以实现。首先,分别介绍了理想摄像机成像原理和鱼眼摄像机的成像原理,并分析了理想摄像机和鱼眼摄像机之间的区别和联系,介绍了鱼眼摄像机的畸变矫正算法和图像插值原理,并利用张正友平面标定法求取了鱼眼摄像机的内参矩阵和畸变系数,成功实现了鱼眼图片的畸变校正。其次,介绍了图像透视变换的基本原理,并设计了基于最小二乘法的最优透视变换矩阵求取方法,然后介绍了传统的基于不变特征的图像拼接方法,并分析了此方法的局限性,提出了改进的基于棋盘格配准的图像拼接方法,最后利用图像融合算法消除了因图像亮度不同而产生的拼接缝。最后,在I.MX6平台上搭建了嵌入式Linux操作系统环境,并将Opencv交叉编译到该嵌入式系统使用。然后在真实车辆上进行全景实验,设计了基于映射思想的算法加速方法,同时利用多线程技术进行算法加速,并在此嵌入式平台上进行实验验证。实验结果表明,本文所提的方法成功实现了全景视觉系统,达到了预期的目的,但是在全景图的质量上有待于进一步提高。
[Abstract]:With the rapid increase in the number of cars used in China, more and more traffic problems begin to be highlighted. Among them, the most frequent accidents are vehicle scratches caused by lane congestion and low-speed parking. The cause of the accident is largely due to the presence of a large number of blind areas of vision while the driver is driving the vehicle. The purpose of the vehicle panoramic vision system is to provide drivers with a top-down panoramic aerial view around the vehicle to help the driver eliminate the blind area of vision and thus reduce the occurrence of such traffic accidents. This paper mainly focuses on the research of vehicle panoramic vision system, introduces the imaging principle and distortion correction method of fish eye camera, designs the panoramic vision image processing algorithm, puts forward the acceleration method of the algorithm, and finally carries on the real vehicle experiment verification. And it is implemented on the embedded platform. Firstly, the imaging principle of the ideal camera and the fish-eye camera are introduced, and the difference and connection between the ideal camera and the fish-eye camera are analyzed. The distortion correction algorithm and the image interpolation principle of the fish-eye camera are introduced. The internal parameter matrix and distortion coefficient of fish eye camera are obtained by using Zhang Zhengyou plane calibration method, and the distortion correction of fish eye image is successfully realized. Secondly, the basic principle of image perspective transformation is introduced, and the method of finding the optimal perspective transformation matrix based on least square method is designed. Then the traditional image splicing method based on invariant feature is introduced, and the limitation of this method is analyzed. An improved image mosaic method based on checkerboard registration is proposed. Finally, the image fusion algorithm is used to eliminate the splicing seam due to the different brightness of the image. Finally, the embedded Linux operating system environment is built on the I.MX6 platform, and the Opencv is cross-compiled to the embedded system. Then a panoramic experiment is carried out on a real vehicle, and an algorithm acceleration method based on mapping idea is designed. At the same time, multi-thread technology is used to accelerate the algorithm, and the experiment is carried out on this embedded platform. The experimental results show that the method proposed in this paper has successfully realized the panoramic vision system and achieved the desired purpose, but the quality of the panoramic image needs to be further improved.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：U463.6;TP391.41

【参考文献】