采用地平面约束的双目PTZ主从跟踪方法

在视频监控领域，包含PTZ（pan-tilt-zoom）相机的双目主从系统可以同时获取跟踪目标的全景信息和高分辨率信息，因此得到了广泛研究与应用。针对智能视频监控的需求，提出了一种基于地平面约束的双目PTZ主从跟踪方法。该方法分为离线标定和在线实时跟踪两个阶段。离线阶段，利用两相机不同视角间的目标匹配关系计算地平面所诱导的单应矩阵，提出了一种从两相机同步视频流中自动估计单应矩阵的方法，该方法与传统方法相比具有不需要标定物和人工干预的优点，然后采用匹配特征点的方法估计相机的主点和等效焦距。在线实时跟踪时，通过单应变换建立主从相机之间的坐标关联，并利用离线阶段标定的主点和等效焦距估计从相机控制参数，从而实现主从跟踪。与其他算法相比，该方法可以应用于宽基线的情形，能够适应目标深度的变化，满足了实时性的要求。室内、外场景的多组实验验证了所提方法的有效性。

Master-slave tracking algorithm using dual PTZ cameras based on ground plane constraint

Master-slave camera system had been widely used in visual surveillance domain because both large-view information and high-resolution local-view information of the tracked object could be obtained at the same time. For the widely demanding of intelligent visual surveillance, a master-salve tracking algorithm using dual PTZ(pan-tilt-zoom) cameras based on ground plane constraint was presented. The proposed method could be divided into two stages: offline calibration and online tracking. In offline stage, the view-to-view homography induced by the ground plane was computed using object matching from two cameras. A novel method that automatically estimated homography from synchronized video streams of the two cameras was presented. Compared with traditional methods, the proposed method didn't need calibration tools, and without the need of manual intervention. Then principal point and equivalent focal length were estimated using matched features. In online stage, the calculated homography transformation was used to establish coordinate correspondence of the master and slave cameras. With the estimated principal point and equivalent focal length, the control parameters of the salve camera could be computed. Therefore, master-salve tracking could be realized. Compared with other methods, the proposed algorithm can be used in the case of wide baseline. The proposed method can also adapt to the depth change of tracked target. Meanwhile, the requirements for real time master-salve tracking can be meeted. Quantitative results in the indoor and outdoor scene show good performance of the proposed approach.