基于动态感兴趣区域车道线分段拟合算法

提出实时视频中基于动态感兴趣区域及分段拟合的车道线的检测算法,动态调整感兴趣区域（ROI）,缩小处理空间。采用大津算法（OTSU）动态提取感兴趣区域灰度阈值,并将该值作为多梯度Sobel边缘检测中的灰度阈值以提高边缘检测精度,利用改进的并行快速细化算法骨架化边缘图像,利用基于广度优先最短路径算法去除毛刺,最后再将图像划分近景和远景区域。在不同区域,采用直线或者曲线分段拟合,提高拟合精度。模拟实验结果表明,背景不太复杂时,一帧图像处理时间约为15ms;而背景较复杂时,处理时间约为35ms,能满足实时性。

Drive Line Segment Fitting Algorithm Based on Dynamic Region of Interest

Proposes drive line detection algorithm based on dynamic region of interest and fitting algorithm.Dynamically adjusts of region of interest（ROI） and reduces the processing space.Using Otsu algorithm（OTSU） dynamically extracts gray level threshold,and makes it as the gray level threshold of the multi-direction Sobel gradient edge detection to improve the accuracy,uses Zhang-Suen skeleton the edge images,then removes burrs and divides the image into two regions.In different regions and conditions,uses line or curve fitting to improve precision.Simulation result shows that a frame with simple background,the processing time is about 15ms,while with complex background,is about 35ms,it can meet the real-time requirements.