CAJ | 학술논문

为兼顾车道识别的实时性和鲁棒性，有序融合了图像中车道线边缘、边缘方向、车道宽度和边缘灰度等特征信息。同时，在识别第1帧车道图像时对整帧图像进行处理；在识别后续各帧车道图像时，根据上一帧图像的识别结果确定动态约束区，并在约束区域内完成识别。实车道路试验结果表明，通过有序融合各种特征信息和合理设置动态约束区能有效识别路面车道线，且车道跟踪具有较好的精确性和可靠性。
위겸고차도식별적실시성화로봉성，유서융합료도상중차도선변연、변연방향、차도관도화변연회도등특정신식。동시，재식별제1정차도도상시대정정도상진행처리；재식별후속각정차도도상시，근거상일정도상적식별결과학정동태약속구，병재약속구역내완성식별。실차도로시험결과표명，통과유서융합각충특정신식화합리설치동태약속구능유효식별로면차도선，차차도근종구유교호적정학성화가고성。
For achieving both the robustness and real-time performance of lane idenfication, the feature in-formation on the edge and its direction and gray scale of lane markings and lane width in the image are sequentially fused. In identifying the first frame of lane image, the whole image frame are processed, while in identifying the fol-lowing frames, dynamically constrained areas are determined according to the identification results of previous frames, within which identification completes. The results of real vehicle road tests show that by orderly fusing vari-ous feature information and properly setting dynamically constrained areas, lane marking identification can be effec-tively achieved with better accuracy and reliability in vehicle lane tracking.