基于V-视差的障碍物检测改进方法

针对传统V-视差方法提取的道路信息与实际道路信息偏离较大,造成障碍物漏检率相对较高的问题,提出一种改进的V-视差障碍物检测方法。在详细分析传统V-视差检测方法的原理和存在的缺陷基础上,提出了极大值约束、较小值抑制约束、距离约束和偏移约束四种条件结合的道路信息提取算法。在采用最小二乘法拟合道路线段后,通过计算道路线段的斜率和截距,并结合视差图检测障碍物。实验结果表明,较传统方法具有较好的障碍物检测效果,且具有一定的实时性和工程实用价值。

Improved method of obstacle detection based on V- disparity

The road information extracted by traditional V-Sdisparity method always has relatively large difference to the practical situation which may cause the higher false dismissal probability. An improved method of obstacle detection based on V-Sdisparity is presented. The theory and deficiency of the traditional method is analyzed. Firstly the road information extraction algorithm was presented by combining maximum constraint, MIN restrain constraint, distance constraint and excursion constraint, the road segment information was fitted with the least square method, and then the slope and intercept was calculated, finally the obstacles were detected by combining the disparity map. The experimental results show that the method has a better detectionSeffectSthan the traditional methods and it has some real time and practical value in engineering.