摘要
近几年以来, 随着汽车品牌数量的剧增，汽车电子行业也迅速发展起来，人们在选车的时候开始注重一些电子配置，如：A VM(Around View Monitor)、LDW(Lane Detecting Warning)、FCW(Forward Collision Warning)和PCW(Pedestrian Collision Warning)等等，都是关于机器视觉和图像处理的一些应用。因此，机器视觉和图像处理的技术也伴随着汽车电子行业的发展而兴起。有了这些功能后，汽车在行驶的过程中就方便许多了，特别是在停车的时候，需要将车停放在规定的车位上。由于现有的社会资源已经不能完全满足人们的需求，特别是对没有多少驾驶经验的新手来说，如何将汽车停放在相应的车位里面，成为很多新手为之头疼的事情。针对这个问题，很多国产车和外资车在出厂前就配备了安全辅助驾驶系统，可以对驾驶员停车入库起到辅助作用，而鱼眼摄像头是此技术的关键。所以，鱼眼摄像头和相关的嵌入式系统被广泛的应用在安全辅助驾驶系统中。生态系统理论
如今，市面上许多中高端国产汽车都安装了车载全景影像系统，简称A VM。该系统主要通过在汽车四周分别安装一个鱼眼摄像头，然后采集汽车四周的视频，最后拼接为一幅全景图像并且显示出来，使驾驶员在车内就可以观察到车外的情况。当停车空间相对较窄时，该功能就可以派上用场，这样可以大大降低刮擦等事故发生的风险，并且使驾驶员更容易将车辆停放在预留停车空间内。
但是，随着科技的进步，市面上的A VM系统虽然逐渐得到完善，但还远远不够。如今的A VM系统是建立在四个普通标清鱼眼摄像头上，采集到的画面也是普通画质，甚至很难去识别前方车辆的车牌。如果与前车发生摩擦碰撞，并不能通过视频画面识别谁是主要责任人。所以，A VM系统已经不再仅仅应用于泊车。随着社会的发展和科技的进步，人们对产品的需求和规格不断提高，车载标清全景影像技术如果跟不上科技的进步而必将逐渐被取缔。全新一代车载高清全景影像系统将要诞生，简称HA VM(HD Around View Monitor），在技术和视频画质以及系统功能方面都可以得到全面的提升，也许会引领汽车行业ADAS(Advanced Driver Assistance System)系统迈向全新时代。
文中针对车载高清全景影像系统设计，从硬件板卡电路的设计和试调到软件算法的研究，采用了一种接口类型为LVDS（MAX96705）的高清鱼眼摄像头，和与之匹配的高速串行解串器（MAX9286）对图像视频数据流进行解串，然后对摄像头进行了实验标定和实地标定。通过标定获得的内部参数和外部参数用于
生成全景图像，以实现车载高清全景影像的完美拼接。
任正非一江春水向东流
文中应用了一款高速串行解串器MAX9286芯片，将4通道高速串行信号解串为4组高速的并行MIPI信号，然后直接输入到CPU的MIPI接口，在GPU 中进行图像处理，包括图像的畸变校正、角点检测、透视变换、图像拼接以及图像增强处理，最后直接从CPU的LVDS接口输出全景影像。相比传统的标纳米机器人
清车载全景影像系统，不需要复杂的解码编码电路，也不像CVBS输出容易被干扰，大大降低了设计成本，并且在清晰度上面更胜一筹。
本文主要的研究内容如下：
(1)学习了实现车载高清全景影像系统所必须的基础知识和相关技术。
第一，分析目前常见的车载全景影像系统的实现原理，以及常用的车载全景影像系统方案。第二，研究实现本系统相关的技术方法，包括：畸变校正、角点检测、透视变换、图像拼接、图像增强处理等。
(2)完成车载高清全景影像系统的总体设计方案和详细设计，实现每个模块的具体功能。
分析本系统的设计原则和设计目标，对系统所有的功能模块进行整体化设计，对各个子模块进行详细设计，实现了车载高清全景影像系统的图像采集、内部参数标定、透视变化、图像拼接和图像增强等功能。
(3)实施对系统功能模块的测试。
针对玩具车和真车进行测试，对固定在玩具车和真车周围的四路高清鱼眼摄像头进行内部参数和外部
参数的标定，然后试着在玩具车和真车上面拼接成一幅完整的图像，拼接完成后，对拼接图像是否错位进行测试。
本文研究的主要结果：
(1)系统板卡原理图的设计和对板卡的制作、贴片等工作的跟进，然后经过板卡的硬件调试，让系统板卡可以正常的运行起来，再经过DDR Stress Test优化，让板卡运行速度达到最优。
(2)系统板卡实现了图像采集，对图像进行畸变校正、角点检测、透视变换、图像拼接、图像增强等处理，在显示屏上显示一幅高清全景影像画面。
(3)在玩具车上和在真车上面进行拼接测试，包括摄像头的安装技巧和棋盘格标定布的位置摆放规则，最终在玩具车和真车上都可以实现全景影像功能。
关键词：高清全景影像；鱼眼；图像处理；嵌入式
Research on Vehicle HD Panoramic Image System
Abstract
In the past few years, with the rapid increase in the number of automobile brands, the automotive electronics industry has developed rapidly. People began to pay attention to some electronic configurations when selecting cars, such as: A VM (Around View Monitor), LDW (Lane Detecting Warning), FCW. (Forward Collision Warning) and PCW (Pedestrian Collision Warning), etc., are some applications for machine vision and image processing. Therefore, the technology of machine vision and image processing has also risen along with the development of the automotive electronics industry. With these features, the car is much more convenient during the journey, especially when parking, you need to park the car in the specified parking space. Because the existing social resources can not fully meet people's needs, especially for novices who have little driving experience, how to park the car in the corresponding parking space has become a headache for many newcomers. In response to this problem, many domestic and foreign-owned vehicles are equipped with a safety assisted driving system before leaving the factory, which can help the driver to park the storage, and the fisheye camera is the key to this technology. Therefore, fisheye cameras and related embedded systems are widely used in safety assisted driving systems.
叶面积指数Today, many mid- to high-end domestic cars on the market are equipped with 360 look-around (vehicle panoramic image system), referred to as A VM. The system mainly installs a fisheye camera
around the car, then collects the video around the car, and finally splices it into a panoramic image and displays it so that the driver can observe the situation outside the car. This function can be used when the parking space is relatively narrow, which can greatly reduce the risk of accidents such as scratches and make it easier for the driver to park the vehicle in the reserved parking space.
However, with the advancement of technology, the A VM system on the market has gradually improved, but it is still far from enough. Today's A VM system is built on four common standard fisheye cameras. The captured image is also of ordinary quality, and it is even difficult to identify the license plate of the vehicle in front. If there is a frictional collision with the preceding car, it is not possible to identify who is the main person through the video screen. Therefore, the A VM system is no longer only used for parking. With the development of society and the advancement of
农业项目投资评估technology, people's demand and specifications for products continue to increase. If the standard definition image technology of the vehicle standard definition cannot keep up with the advancement of technology, it will gradually be banned. A new generation of car HD panoramic image system will be born, referred to as HA VM (HD Around View Monitor), which can be comprehensively improved in terms of technology and video quality and system functions, and may lead the automotive industry ADAS (Advanced Driver Assistance System) system. Towards a new era.
This thesis is aimed at the design of vehicle HD panoramic image system. From the design and trial adjustment of hardware board circuit to the research of software algorithm, a high-definition fisheye camera with interface type LVDS (MAX96705) is adopted, and the high-speed serial matching with it is adopted. The deserializer (MAX9286) deserializes the image video data stream and then performs experimental calibration and field calibration of the camera. The internal and external parameters obtained by calibration are used to generate panoramic images for perfect splicing of high-definition panoramic images of the vehicle.
This thesis applies a high-speed serial deserializer MAX9286 chip, which de-strings 4-channel high-speed serial signals into 4 sets of high-speed parallel MIPI signals, and then directly inputs them into the MIPI interface of the CPU to perform image processing in the GPU, including images. Distortion correction, corner detection, perspective transformation, image stitching, and image enhancement processing, and finally output panoramic images directly from the LVDS interface of the CPU. Compared with the traditional standard definition ring system, it does not require complicated decoding and encoding circuits, and it is not easy to be interfered with the CVBS output, which greatly reduces the design cost and is superior in definition.
The main research contents of this thesis are as follows:
(1) The autor studied the basic knowledge and related technologies necessary to realize the vehicle HD panoramic image system.
First, analyze the implementation principles of the current common surround vision system, as well as the commonly used surround vision system solutions. Second, research and implementation of the technical methods related to the system, including: distortion correction, corner detection, perspective transformation, image mosaic, image enhancement processing.
(2) The autor completed the overall design scheme and detailed design of the vehicle HD panoramic image system in order to realize the specific functions of each
module.
The autor then analyzed the design principles and design goals of the system, integrate all the functional modules of the system, and design the sub-modules in detail to realize the image acquisition, internal parameter calibration, perspective change, image mosaic and Features such as image enhancement.
(3) The autor implemented testing of system function modules on the toy car and on the real car, the
internal parameters and external parameters of the four-way high-definition fisheye camera fixed around the toy car are calibrated, and then try to splicing into a complete image on the toy car. After the splicing is completed, whether the spliced image is misaligned carry out testing.
The main results of this thesis are:
(1) The author completed the design of the system board schematic diagram and the follow-up of the board production, patching, etc., and then through the hardware debugging of the board, so that the system board can run normally, and then optimized by DDR Stress Test. Let the board run at optimum speed.
(2)The author realized image acquisition on the system board, and performed distortion correction, corner detection, perspective transformation, image mosaic, image enhancement, etc. on the image, and displays a high-definition panoramic image on the display screen.
(3)The author performed splicing test on the toy car and on the real car, including the installation skill of the camera and the position placement rules of the checkerboard calibration, and finally the panoramic image function can be realized on both the toy car and the real car.
Keywords: HD panoramic image; fisheye; image processing; embedded
nmda受体拮抗剂
豫公网安备41160202000603
站长QQ:729038198
关于我们
投诉建议