一种基于直线模型的道路识别算法研究

A Linear Model Based Road Identification Algorithm

为了提高道路识别的鲁棒性和抗干扰能力 ,提出了一种道路边缘识别算法。该算法的创新点在于采用自然边界作为道路识别的依据 ,同时算法也可以将白色路标作为识别道路边缘的标志。由于道路的自然边界不像车道标志线那样容易受到污染 ,因而使该算法具有了较高的鲁棒性。该算法是采用像素级特征和帧图像特征来识别道路边缘 ,可称为全局道路边界模型。在模型中由于采用了有关的道路约束条件 ,因而使得算法具有较好的抗干扰能力。该算法包括初始检测算法和跟踪算法两个部分 ,其中在跟踪算法中采用感兴趣区域算法 (ROI)来限制检测区域 ,以提高实时性。最后 ,还提出了一个道路边界可信度检测函数 。

In the paper,a road edge identification algorithm is developed .The new idea of this method is touse natural road edge ,as well as the whites trip for road information acquisition .Then atural roade dge will notbee as ilypolluted like the white lane maker do ,soitindicates bettera daptability .Inthealgorithm ,weuseboth the pixelfeature and the framefeaturetoidentify theroadedge ,which is referred to as the global road model.Be cause several road constrainsisused toensure the roadedgede tection ,the algorith misimmune to the influence of thei magedistur bance .The algorithm of theroaded geidentifi cationin cludes two stages:initializationdetection and tracing detection .The in itialization stagedetects the roadedge from the global roadimage . The tracealgorith muses there gion of interest (ROI)to limitdete ctingarea ,which can save muchtime .Inorder to giveamea sure of there liability of the road detecting result,this paper presents a road edge identification esti mation function ,which canesti mate the reliability of the road edge .