【作者】 单红梅；

【导师】 苏建；

【作者基本信息】 吉林大学 ， 载运工具运用工程， 2011， 博士

【摘要】 在我国机动车检测技术的发展过程中,硬件技术普遍受到重视,而一些难度大、投入多、社会效益不明显的检测项目常被忽略,使这些检测项目缺少相应的检测设备,严重影响检测线上的检车速度,阻碍了检测技术向智能化、自动化方向发展。轴距左右差的检测就是被忽略的一项。轴距左右差对汽车使用性能有着显著的影响。当差值超过一定限度时,车辆行驶稳定性变差;当制动或高速行驶时,会使汽车发生激转而侧滑或翻车;当直线行驶时,会增加车身占道宽度,影响交通安全,增加轮胎磨损,带来经济损失和潜在危险。自2001年起,颁布了有关汽车轴距左右差的国家标准。国标《营运车辆综合性能要求和检验方法》GB18565-2001规定:整车装备应齐全、完好、有效,各连接部件紧固完好。车体应周正,车体外缘左右对称部位(在离地高1.5m内测量)左右轴距差不得大于轴距的1.5/1000。2004年,国家对轴距左右差的要求日趋严格,颁布了交通行业标准《营运车辆技术等级划分和评定要求》JT/T198-2004规定:车体外缘左右对称部位(在离地高1.5m内测量)左右轴距差不得大于轴距的1.2/1000。接着,又颁布了国标《机动车运行安全技术条件》GB7258-2004规定:汽车在直线行驶过程中,汽车的前后轴中心的连线应与路面方向保持一致。2005年,国标《汽车综合性能检测站能力的通用要求》GB/T 17993-2005规定:汽车轴距左右差测量范围:0～20000mm(轴距差);分辨力1mm。由此可见,对轴距左右差的检测精度要求越来越高,这就要求有更先进、更精确的检测方法。基于上述背景,本课题研究了基于立体视觉轴距左右差的检测系统,替代传统的人工测量的检测方法,利用双目立体视觉的检测系统,实现对轴距左右差不停车的动态测量,并满足检测精度要求。论文主要围绕图像处理技术,以及角点提取、立体匹配、三维重建、摄像机标定等关键技术,进行了全面、深入的研究和探讨,提出了本课题的检测算法,并通过实车试验证明本检测系统具有较好的检测精度和重复性。本文的研究内容如下:1、论述本课题研究的背景和意义,分析了汽车轴距左右差检测技术的国内外研究现状,提出了基于立体视觉汽车轴距左右差检测方法。2、介绍了立体视觉测距原理,以及立体视觉的构成和研究内容。阐明了汽车轴距左右差的检测原理。提出了基于立体视觉汽车轴距左右差检测系统方案设计。3、运用图像识别关键技术,获得轮毂图像。分析采用图像处理关键技术中适用于本论文研究的检测方法:中值滤波、Otsu算法的图像分割方法,区域填充选取扫描线种子填充算法,数学形态学用来移除边界,再利用形态学开运算处理得到轮毂图像,通过最小二乘法进行圆特征的提取,得到轮毂中心的像素坐标。研究了立体视觉算法:采用Harris算法提取靶标角点,以及立体视觉中的匹配技术,满足检测系统精度要求。4、研究了基于立体视觉汽车轴距左右差检测的标定系统。首先分析摄像机需要标定的参数,并说明了世界坐标系、摄像机坐标系、图像坐标系的定义以及它们之间的变换关系,然后分析了摄像机成像模型,研究了摄像机标定算法,采用了基于一阶径向畸变的三维立体靶标标定方法,并对算法进行了详细的描述。并用靶标进行试验,验证了本文的标定算法。通过不同基线距离、不同标定距离和不同倾斜角的标定试验,优化了传感器结构参数。5、研究了汽车斜直线行驶时的轮毂中心的三维重建模型。首先介绍了点和直线的三维重建算法,然后用轮胎试验完成了不同轮距的汽车,在不同角度斜直线行驶时的轮毂中心的三维重建,并分析了绝对误差曲线。验证了本文的三维重建算法。6、基于立体视觉汽车轴距左右差检测系统的试验研究。合理设计了基于立体视觉汽车轴距左右差检测系统软、硬件系统方案。利用实车对本文设计的系统进行了试验研究,并对汽车曲线行驶引起的误差进行校正并加以补偿。试验结果表明,该检测系统能够动态检测轴距左右差,满足检测线的检测要求,检测结果在误差范围内。本研究工作的主要创新点为:1、首次提出了一种应用立体视觉技术对汽车轴距左右差进行非接触三维测量的新方法。研发设计了汽车轴距左右差检测系统,该系统可以有效修正被测车辆相对于检测系统的纵向中心线偏差,实现汽车轴距左右差的准确测量。2、建立了基于车轮轮毂图像识别匹配算法,使数字图像处理算法与实际的检测项目相结合,完成了检测系统软硬件的开发,实现了汽车轴距左右差的精确测量。3、建立了车轮轮毂图像的三维重建算法,可以对汽车曲线行驶时的轮毂中心坐标进行提取,不仅获得了轮距左右差的信息,而且获得了车辆纵向中心线的偏差信息,保证了测量精度,实现了车辆的动态检测。论文所作的研究将立体视觉检测技术与实际检测相结合,在算法及精度保证等方面取得了进展。同时,将立体视觉技术引入到汽车轴距左右差的检测中,实现了检测过程的自动化。更多还原

【Abstract】 In the course of automatic detection technology, universal attention to hardware technology, ignore some testing item of difficulty and invested more than and social not obvious, that these tests are lack of appropriate detection equipment, which has affected the speed of automobile detection, hinder the Intelligent development and automation of Detection technology. Ignore the detection of the wheelbase difference. The wheelbase difference has a significant influence on automotive performance. When the difference exceeds a certain limit, vehicles stability changed for the worse; when braking or running at high speed, it will enable the auto will occur to turn to radical sliding or overturn; when straight line travel, it will Increases the body’s road width, impact on road safety, increase tire wear, brought economic losses and potentially dangerous.Since 2001, it is enacted national standards about the automobile wheelbase difference. According to“National Standards multiple performance requirement and detecting methods for commercial vehicle”GB18565-2001, the equipment requirements for vehicle goes: vehicle equipment should be complete, intact and effective and fastening parts should be well-connected. The body should be modified, (from 1.5m high with measurement), wheelbase defect between the right and the left should be less than 1.5/1000 of the wheelbase. In 2004, about national wheelbase difference are more stringent, promulgated the transportation industry standards, According to“Operating vehicle technology classification and assessment requirements”JT/T198-2004: The body should be modified, (from 1.5m high with measurement), wheelbase defect between the right and the left should be less than 1.2/1000 of the wheelbase. Then, also promulgated“National the motor vehicle running safety technical conditions”GB7258-2004, Automobile in straight line travel process, the center of front axle and rear axle of Automobile Should be consistent with the direction of the road. In 2005, "Specifications for multiple-function detecting test station of vehicle transport" GB/T 17993-2005: 0 ~ 20000mm (wheelbase defect);error allowed 1mm. This shows that the request of wheelbase difference detection accuracy is more and more high. This requires more advanced and more accurate detection methods.Based on above background, this subject study on testing system of the wheelbase difference based on stereo vision, by using the binocular stereo vision, realization do not stop the dynamic measurement, and meet requirements for measurement accuracy.The Key technologies, such as the Image processing technology, camera calibration, as well as Corner extraction and stereo matching and 3D reconstruction and camera calibration, are studied and discussed thoroughly. This paper shows detection algorithm of this subject, and detection of the automobile body. Experiments show that, this testing system has better detection accuracy and repeatability.The main research works are as following:1. Study the background and meanings of the paper, analyzed detection technology of the wheelbase difference internal and overseas, proposed testing method of the wheelbase difference based on stereo vision, and describes the key features of the paper.2. Produced ranging principle of stereo vision, and composition and contents of stereo vision, illustrates the detection principle of the wheelbase difference. Uses camera installation program of axial mode, proposed the project design of the wheelbase difference based on stereo vision.3. Using pattern recognition key technologies, obtains the wheel hub image. Analysis by image processing applies to this thesis research in the key technologies of detection methods: Median filter and image segmentation method of OTSU algorithm, area fill select Scan line seed fill algorithm, used mathematical morphology to remove the boundary, and opens operation processing of mathematical morphology to obtain the wheel image, by least square method to circular feature extraction, obtains pixel coordinates of wheel center. Study on stereo vision algorithm, uses Harris algorithm to extraction of target corner, and stereo vision matching technology, uses outside polar line geometric constraint, it is better to eliminate the undesirable match under various conditions, obtains the very high correct match rate, meets the precision of a measurement system requirements.4. Study calibration system of the wheelbase difference based on stereo vision. First of all, analysis camera parameters are required, and described the world coordinate system, camera coordinate system, the definition of the image coordinate system and the transforming relationship between them, as well as the camera imaging model, research on camera calibration algorithm, uses calibration method based on a step radial direction distortion, and description of the algorithm in detail. Finally realizes the rapid and accurate calibration of the parameters inside and outside and structural parameters of the sensor. And experimented with the target, verify the calibration of this algorithm. Through calibration experiments of different baseline distance, Description Effect of baseline distance calibration accuracy. 5. Study on 3D reconstruction models of several typical track width vehicles when Oblique straight line driving. First introduced 3D reconstruction algorithm of point and the straight line, then completion of the experiment with tires of different tread vehicle, 3D reconstruction of wheel center At different angles when Oblique straight line driving, and drawn the absolute error curve. It is verified Accuracy of 3D reconstruction algorithm of this paper.6. The experimental study of the wheelbase difference based on stereo vision. Reasonable design system programs of hardware and software. Uses vehicle body to a experimental study on the design of this system, and correction and compensate the Error when Automobile curve travel. The experiment results show that, this detection system can meet the test line of real-time detection requirement; test results were consistent with error range.The main contents of the study and the results are as follows:1. Based on stereo vision in automobile wheelbase difference non-contact 3D measurement method.Research and design for a variety of typical track automobile wheelbase difference detection system,the system can realize the accrate measurement of left and right axle distance difference.2. Based on the wheel hub image matching algorithm, the digital image processing algorithm and the actual testing project combining,complete detection system hardware and software development,relize automobile wheelbase difference measurement.3. Based on wheel hub of 3D image reconstruction algorithm,which allows the extraction of the hub center coordinate more accurately,thereby improving the wheelbase difference calculation precision.The research and the development work made by this paper is a typical example of application which makes a number of research results in the technology and algorithms. At the same time, this paper leads the technology of stereoscopic vision into the detection of wheelbase difference, realizes the automation process of vehicle detection.更多还原