基于DBSCAN的复杂环境下车道线鲁棒检测及跟踪

Robust lane detection and tracking in complex environment based on DBSCAN

为了提高车道线检测的准确性、实时性和鲁棒性,首先,利用逆透视点变换减少图像形变;根据颜色和几何特征,基于DBSCAN算法实现聚类簇划分。然后,利用基于抛物线模型的随机采样一致性拟合方法初步完成车道线提取,并针对不同的环境干扰,制定了相应的优化策略,实现了自车道线的鲁棒检测。最后,利用卡尔曼滤波对车道线模型进行跟踪处理,保证系统的稳定性。实验证明,本文算法在多种复杂环境下都能准确识别自车道线,能够满足辅助驾驶系统的实际需求。

In order to improve the accuracy,real-timing and robustness of the vehicle lane detection,firstly,the inverse perspective point transformation is applied to solve the problem of perspective deformation.According to the color and geometric features,the lane lines are divided into different clusters based on the Density-based Spatial Clustering of Applications with Noise(DBSCAN)algorithm.Then,the Random Sample Consensus(RANSAC)fitting based on parabola model is used to complete the lane line extraction.For different environmental disturbances,an optimization strategy is developed to achieve effective robust detection of lane lines.Finally,Kalman filter is used to track the lane line model to ensure the stability of the system.Experiments show that the proposed algorithm can accurately identify the lane lines in a variety of complex environments,which meet the needs of the automotive auxiliary driving system in terms of accuracy,real-timing and robustness.