CAJ | 학술논문

提出一种基于边缘分布和特征聚类的车道标记线检测方法。首先采用可变窗口计算车道标记线局部灰度阈值，结合图像梯度提取出有效边缘。然后按照不同工况下车道标记线的边缘分布特性，提取特征点。最后对特征点进行聚类处理，将离散的特征点归类为不同的直线段。测试结果表明，该方法可取得较高的车道线识别率，有效排除误识别，准确表示车道线方向信息。
제출일충기우변연분포화특정취류적차도표기선검측방법。수선채용가변창구계산차도표기선국부회도역치，결합도상제도제취출유효변연。연후안조불동공황하차도표기선적변연분포특성，제취특정점。최후대특정점진행취류처리，장리산적특정점귀류위불동적직선단。측시결과표명，해방법가취득교고적차도선식별솔，유효배제오식별，준학표시차도선방향신식。
A novel lane marking detection method based on edge distribution and feature clustering is pro-posed in the paper. Firstly the local grey level threshold of lane markings is calculated with variable window and combined with image gradient the effective edges are extracted. The feature points of lane markings are then extrac-ted according to the edge distribution characteristics of lane markings under different scenarios. Finally, clustering processing is conducted, in which the discrete feature points are classified into different straight line segments. Test results indicate that the method proposed can achieve high identification rate, effectively eliminate erroneous extrac-tion and accurately show the directional information of lane markings.