Development of a System Measuring Vehicle Motions on Video Images

为了跟踪车辆的运动轨迹，本研究开发了基于视频图像的车辆运动测量系统。视频图像是由放置在观测气球上的摄像机拍摄的，覆盖了大约200米的高速公路路段。在建立了人工测量系统后，我们开发了自动测量系统，并对系统的性能进行了检查。该系统除了交通检测器通常能观察到的交通量、密度和速度等观察指标外，还可以观察微小的车辆运动。该系统的过程由两部分组成：1.图像转换视频图像通过调整预定的标准点，将视频图像转换为与标准图像相同的坐标。2.车辆识别利用变换后的图像，通过比较背景和亮度来识别每辆车。通过比较手动系统和自动系统的性能来检查系统的性能。结果表明，自动系统的车辆识别率与人工系统的车辆识别率基本相当，但自动系统的识别率存在一定的差异，这主要是由于识别弱反射车辆和与人工系统的轨迹不同造成的。

This study develops the system of measuring vehicle motion on video images for the purpose of tracking vehicle trajectory. The video images are taken by the video camera being placed on an observation balloon, and covers about 200 meters' expressway section. After we established manual system of measuring, we developed automatic system of measuring and checked the system performance.In addition to usual observation index observed by traffic detectors such as traffic volume, density and velocity, this system enables to observe micro vehicle motions.The process of this system are consist of two parts as follows ;1.Image TransformationVideo Images are transformed to same coordinates as the standard image, by adjusting predetermined standard points.2.Vehicle RecognitionUsing transformed image each vehicles are recognized by comparing brightness with one of background.The system performance are checked by way of comparing performance of manual system with one of automatic system. In result, rate of vehicle recognition of automatic system is almost equal to one of manual system, however, accuracy of automatic system has some difference mainly caused by recognizing feeble reflective vehicle and tracks to one of manual system.