摘要
针对车载视频中的障碍目标,通过单目视觉测距法对目标位置进行测量;利用物像位置公式推倒障碍目标的外形求解方法;基于视频跟踪系统,采用帧差法得到运动目标的速度参数、方向参数、加速度参数等计算障碍目标的大致运行轨迹。结果表明,对参数的实时测量分析能有效避免车道的防偏和防撞问题,为车载视频系统捕捉多个目标、分析处理多个目标、及时合理规避目标障碍等提供了一定的理论依据。
Barriers to target of vehicle-mounted video, such as the target location, is measured by the monocu- lar vision algorithm. The shape solving method is obtained by the image formula. The target trajectory of speed, di- rection and acceleration are obtained by the frame difference method in the video tracking system. Results show that parameter measurement avoids the car collision effectively. It provides a theoretical basis for the vehicle - mounted video system to capture multiple targets, analyze and deal with multiple targets, and timely avoid obstacles.
出处
《电子科技》
2017年第2期90-93,共4页
Electronic Science and Technology
关键词
车载视频
障碍目标
单目视觉
自动驾驶
vehicle - mounted video
barrier to target
monocular vision
auto - pilot
