论文部分内容阅读
近些年来,泊车辅助逐渐成为汽车辅助驾驶领域的研究热点。其中,如何由车尾摄像头得到场景鸟瞰图(Bird-View)以辅助泊车已成为具有挑战性的课题。本文提出了一种基于场景分类的bird-view方案,该方案既能给出接近真实情况的地面俯视图,又能提供车后障碍物的详细分布与距离,从而为驾驶员泊车入位提供便利。通过帧间滤波和算法优化,帧间抖动和实时性问题得到解决,使得该方案具备在实际中应用的条件。文中给出了详细的设计思路、方案对比以及实验分析,以验证所提方案的优越性。
In recent years, parking assistance has gradually become a research hotspot in the field of vehicle assisted driving. Among them, how to get a bird-view from a rear camera to assist parking has become a challenging issue. In this paper, a bird-view scheme based on scene classification is proposed, which can not only give the ground view close to the real situation, but also provide the detailed distribution and distance of obstacle behind the car to facilitate the parking of drivers . Through the inter-frame filtering and algorithm optimization, inter-frame jitter and real-time issues are resolved, making the program has the practical application of the conditions. The article gives a detailed design ideas, program comparison and experimental analysis to verify the superiority of the proposed program.