基于道路环境的车载双目系统立体匹配算法  

Stereo Matching Algorithm for Vehicle Binocular System Based on Road Condition

作　　者：张方义 赵刚 刘海英 姜永昌 于志豪[2] ZHANG Fang-yi;ZHAO Gang;LIU Hai-ying;JIANG Yong-chang;YU Zhi-hao(College of Transportation 1, Qingdao 266510, China;College of Mechanical and Electronic Engineering,Shandong University of Science and Technology, Qingdao 266510, China)

机构地区：[1]山东科技大学交通学院,青岛266510 [2]山东科技大学机械电子工程学院,青岛266510

出　　处：《科学技术与工程》2018年第36期78-86,共9页Science Technology and Engineering

基　　金：山东省自然科学基金面上项目(ZR2017MEE072)资助.

摘　　要：为提高车载双目系统户外工作性能,提出了一种车载双目系统立体匹配算法。在代价计算阶段,提出了一种3bitCensus变换,该变换提高了对局部特征的描述能力并具有较强的鲁棒性;在代价聚合阶段,提出了基于边缘截断的自适应窗口聚合算法,该算法能够根据局部特征和边缘信息自适应的调节聚合窗口,提高了算法匹配精度;在视差计算阶段,利用跳跃式视差计算方法,进一步提高了计算效率。实验证明:基于3bit-Census变换+基于边缘截断的自适应聚合+跳跃式视差计算的立体匹配结构能够有效提高双目系统的鲁棒性、匹配精度及计算效率,适用于车载双目系统。In order to improve outdoor performance of vehicle binocular system, a stereo matching algorithm for vehicle binocular systems was proposed. In the stage of cost calculation, a 3bit-Census transformation was proposed, which improved the capability of describing local features and had strong robustness. In the phase of cost aggregation, an adaptive window aggregation based on edge truncation was proposed. The algorithm can adjust the aggregation window according to local features and edge information, and improved the matching accuracy of the algorithm. In the phase of parallax calculation, the calculation efficiency was improved by using leap parallax calculation method. The experimental results show that the local stereo matching structure of “3bit-Census transformation & An adaptive window aggregation based on edge truncation & Leap parallax calculation” can effectively improve the robustness, accuracy and calculation efficiency, and is suitable for vehicle binocular systems.

关 键 词：双目系统 立体匹配 鲁棒性 匹配精度 计算效率 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]