运动车辆的多传感融合跟踪

针对单一传感器可靠性低、有效探测范围小的缺点,提出了采用雷达与机器视觉融合来实现路面运动车辆跟踪的新方法.该方法采用动力学模型对车辆运动进行描述,考虑了车体运动与车轮速率、转向角之间的关系,比用线性模型更符合车辆实际行驶时的复杂运动状况.通过基于雷达量测的扩展卡尔曼滤波估计建立视觉窗口,再根据图像灰度信息自适应调整窗口中心位置及尺寸,有效地限制了后继图像处理的工作区域,提高了系统的实时性.新方法采用数据融合技术,充分利用雷达与图像传感的量测信息,改善了对机动目标的状态估计.实验证明,该方法能明显提高路面运动车辆位置和方向角的跟踪精度.

Vehicle Tracking Method Based on Multi-Signal Fusion

Maneuver vehicle tracking system based on a single sensor has defects of low reliability and narrow tracking range. To enhance system performance, a new method is presented for the maneuver target tracking by integrating the radar and machine vision information. For the mobile case, a dynamic model of the vehicle motion, which takes account of the relationship between the rigid body motion of vehicle and the steering and drive rates of wheels, is adopted. A vision tracking window, whose center and size can be adaptively adjusted according to the image intensity data, is produced based on radar-based extended kalman filter. This method limits the working area for subsequent image processing and therefore improves real time performance. Data fusion technology that can integrate data from radar and image sensor to make full use of both measurements is then adopted, thus further improves the state estimation performance for maneuver targets. The experiments show that the proposed method can significantly improve position and orientation tracking accuracy of the moving vehicle.