双目协同多分辨率主动跟踪方法

Multi-resolution active tracking algorithm using two-camera collaboration

PTZ(pan-tilt-zoom)相机由于其具有可变视角和可变分辨率能力,在视频监控领域得到了广泛的应用。该文针对智能监控的需求,提出了一种基于双目PTZ相机的多分辨率主动跟踪方法。该方法分为离线标定和在线协同跟踪两部分。离线标定部分,提出了一种基于图像特征匹配的单目自标定和基于目标运动信息的双目自标定方法,该方法操作简单,无需标定物,在最大程度上减小了对人工干预的依赖,在此基础上推导了系统所具有的两个重要性质;在线协同跟踪部分,设计了一种分段静止的协同跟踪策略。通过实际监控场景下的视频实验,验证方法的有效性和可行性。实验结果表明,该方法可以在复杂环境下有效的主动跟踪目标,在智能监控领域具有较广泛的应用前景。

PTZ （pan-tilt-zoom） cameras have been widely used in visual surveillance domain due to their ability to cover wide field of view and large scale range. For the large demand of intelligent visual surveillance, a novel multi-resolution active tracking method was presented based on two PTZ cameras. The proposed method consists of two stages： off-line calibration and on-line cooperative tracking. In off-line calibration stage, a novel method was presented that combines feature based single-PTZ camera self-calibration and object motion based dual-PTZ cameras self-calibration. The proposed approach doesn′t require calibration tools and manual operation. Two important properties of proposed system were also inferred based on above calibration method. In on-line tracking stage, a piecewise cooperative tracking strategy was designed. The verification of proposed algorithm framework was implemented on real outdoor surveillance environment. Experimental results show that the proposed method can track the moving object effectively. The proposed framework can be used in intelligent visual surveillance with wide application prospect.