嵌入式汽车纵向碰撞预警系统研究

Research on Embedded Longitudinal Collision Alarm System for Vehicle

在嵌入式硬件平台上采用Linux操作系统研究了防止汽车碰撞的预警算法。在车辆的图像识别基础上,利用透视投影变换原理实现本车与前车的间距测量,依据图像序列帧与本车速度信息确定前车的行驶状态,基于前车状态确定安全系数权重,建立汽车纵向碰撞预警模型。车辆行驶实验表明,纵向碰撞预警模型能够对处于最小安全车距的汽车给以准确预警,并使其计算的最小安全车距符合驾驶员的习惯车距。

An alarm algorithm to prevent the vehicle collision is researched with Linux operating system on the platform of embedded hardware.On the basis of image recognition of vehicles,the measurement of the space between a vehicle and another one ahead was performed by using the transform principle of the perspective projection to make sure the driving states of the one ahead according to serial images and its speed.Based on the state of the vehicle ahead,the weight of safety coefficients was ensured and the vehicle longitudinal collision alarm model was established.Driving experiment results prove that the longitudinal collision alarm model can correctly warn the vehicle in shortest safe distance to another vehicle straight ahead and the shortest safe distance calculated is in conformity with the accustomed distance of drivers.