基于参考点配准法的全景泊车系统研究

为了提高驾驶行车的安全性,研究并改进了一种全景泊车系统,由鱼眼摄像头,图像处理单元和显示器三大部分组成.鱼眼摄像头采集汽车四周的超广角图像;在OpenCV环境下通过鱼眼摄像头标定,桶型矫正和图像的透视变换技术形成汽车前后左右的4幅平面俯视图;提出了一种参考点配准的方法,可以根据平面俯视图得到汽车的360°全景图像,并在显示器上显示.与特征点匹配法相比,参考点配准方法受环境的影响很小且对硬件性能要求较低,因此,更适用于嵌入式系统实现.

Study of panoramic parking system based on reference points registration method

In order to improve the safety of driving,the panoramic parking system is studied and improved. The system consists of three components,which are fish-eye cameras,an image processing unit and a displayer. First, fish-eye cameras are used to capture the ultra-wide-angle images around the car. Then four vertical views around the car are obtained by the fish-eye camera calibration,the bucket correction and the perspective transformation based on OpenCV. Reference points registration method is proposed to obtain a 360° panoramic bird's eye view, which is showed on displayer,according to the vertical views. Since the proposed method is hardly affected by the environment and not dependent on hardware performance heavily,it is more suitable for embedded system implementation compared with the feature points matching method.