We propose a measurement method based on binocular vision aiming at the rapid and accurate measurement of truck dimension-al parameters. The composition of the measurement system and the principle of binocular measurement are discussed. The vehicle target is detected by image difference method and the image segmentation method based on HSV color space is used to extract the vehicle body. Then we design a support vector machine classifier based on Hu moments to classify vehicle types. The rectangular carriage of the vehicle is extracted by Hough transform if classified as a truck. Finally,the edge measurement points are determined,and the stereo matching is realized by feature matching technology,and the dimension automatic measurement is completed. The measurement method has the ad-vantages of simple installation structure,low occupancy space,low cost and fast measurement speed compared with the traditional meas-urement of laser radar and infrared light curtain. Experiment shows that the measurement system error is less than 3％ and its time con-sumption is less than 2 second. The truck body size measurement can be achieved with accuracy and little time consumption.%针对货车尺寸的快速精确测量,提出了一种基于双目视觉的测量方法.论述了测量系统的组成、双目测量的原理,并通过图像差值法检测车辆目标,在此基础上应用基于HSV颜色空间的图像分割方法提取车身.设计了一个基于Hu矩的支持向量机分类器对车型进行分类,对分类为载货汽车的对象通过霍夫变换检测提取矩形车厢.最后确定边缘测量点,结合特征匹配技术实现立体匹配,并完成尺寸自动测量.该方法与传统的激光雷达法和红外光幕法相比,具有安装结构简单,占用场地少,成本低,测量速度快的优点.实验结果表明,该算法测量精度低于3％,耗时低于1.5 s,实现了载货汽车类车辆长度尺寸及车厢尺寸测量的精确性和快速性.
展开▼
