CAJ | 학술논문

针对目前车道线检测算法识别率不高、弯道检测不准确的问题，提出了一种基于双曲线模型的车道线检测算法．算法首先运用Canny算子对道路边缘进行检测；采用 Hough变换提取道路边界点，并使用扩展的Kalman滤波进行预测跟踪来减小道路扫描范围；最后通过左右车道边界参数与双曲线模型参数进行匹配，利用最小二乘法来求解模型参数，完成车道边界重建．实验结果表明，该算法的准确率达到93．4％，每帧图像的处理速度为87．4ms，有效克服抛物线模型和直线模型在直道与弯道连接处不连续的问题，而且在车道线模糊、对比度较低的情况下也能快速准确的识别出车道线．
침대목전차도선검측산법식별솔불고、만도검측불준학적문제，제출료일충기우쌍곡선모형적차도선검측산법．산법수선운용Canny산자대도로변연진행검측；채용 Hough변환제취도로변계점，병사용확전적Kalman려파진행예측근종래감소도로소묘범위；최후통과좌우차도변계삼수여쌍곡선모형삼수진행필배，이용최소이승법래구해모형삼수，완성차도변계중건．실험결과표명，해산법적준학솔체도93．4％，매정도상적처리속도위87．4ms，유효극복포물선모형화직선모형재직도여만도련접처불련속적문제，이차재차도선모호、대비도교저적정황하야능쾌속준학적식별출차도선．
For the low rate of recognition and inaccurate detection of the curve ,an algorithm of lane line detection based on hyperbola model has been proposed .First the algorithm uses Canny operator to detect the edge of the lane .Then the Hough transform is used to extract lane boundary points .The extended Kalman filter ,with reduced road scanning range ,is utilized for predictive tracking ,which is able to improve the accurace of extraction .At last ,the left and right lane boundary parameters find a perfect match with the hyperbla model by the least squares method ,and lane boundary reconstruction is completed .Experimental results show that the accuracy of the algorithm reaches 93 .4% and the processing speed of each image is 87 .4ms . T he algorithm could overcome parabolic model and discontinuous problem at joint of straight line and curve .It also could quickly and accurately identify lane line in the situation of blur and low contrast lane line .