改进GM(1,1)模型的前车检测与跟踪研究

Research on detection and tracking of front vehicle based on modified GM(1,1) model

提出了一种改进灰色预测模型GM(1,1)的前方车辆检测与跟踪方法,利用Hough变换识别两侧车道标识线,缩小前方车辆检测与跟踪区域,完成对前方车辆的检测之后,通过改进GM(1,1)模型的持续更新,搜索其运动规律,并对前方车辆的运动轨迹进行预测,根据预测结果实现对前方车辆的跟踪。实验结果表明,该方法不需要对随机噪声序列和目标运动规律进行假设,克服了随机噪声和分离合并的影响,具有较好的实时性和鲁棒性,适合于范围较小的前方车辆检测与跟踪。

A new detection and tracking method of front vehicle based on modified grey forecasting model GM （ 1, 1） is proposed. Using Hough transform recognition on both sides of driveway logo lane, setting front vehicle detection and tracking area, and the detection of front vehicle is completed, through the continuous updating modified GM （1, 1） model, movement law of front vehicle is searched out, the trajectories of front vehicle is predicted, finally according to the prediction result finishs the tracking of front vehicle. The experimental results show that this method do not need hypothesis of random noise sequence and target movement law, overcome the influence of random noise and separation merger, and has a good real-time and robustness, and is suitable for small areas detection and tracking of front vehicle.