智能汽车赛道图像识别技术的研究

Research of Recognition Technology of Smart Car's Track Image

改进现有边沿提取算法,提出二阶绝对差分,对智能汽车赛道中心线连续遍历,计算各相邻三步的绝对信号跳跃斜率的绝对值,最后加以阈值判断。旨在识别图像中有用的赛道信息,剔除智能汽车在弯道时采集的部分相邻赛道图像,再对有用赛道中心线加以一次线性拟合,求出赛道的弯曲程度及规划出弯道行驶路径。同时,提出一次线性插值处理十字路面的边沿信息缺损情况,再精确解算十字路面的中心线,最后对其加以线性拟合,从而实现高速无扰行驶。

Existing algorithm of edge detection was improved. And second-order absolute difference was presented, to traverse the track center line of smart car continuously. Then, for each signal, calculating the absolute value of the signal＇s forward difference and the backward difference of signals. Next, calculating the absolute value of the two absolute values＇ difference, finally using the threshold to constraint. The purpose is to recognize the useful track information in image, and eliminate the partial image of neighbor track. Then processing the useful track center line by using linear fitting, to calculate the degree of crook of track and plan the marched route in curve. Meanwhile, introducing a method to process the defective information of edge in cross road by using a linear interpolation, for calculating the track center line of cross road accurately. Finally, to process it by using linear fitting to achieve fast robust driving.