智能目标检测与跟踪系统的设计与实现

为了实现复杂背景下对任意目标的智能检测与稳定跟踪,设计了基于双DSP结构的集视频处理、云台运动控制于一体的实时电视跟踪系统.系统采用模块化硬件结构,并提供直接用户接口可使系统脱离上位机独立工作.同时改进了目标识别算法,提出一种基于区域分割的相关模板提取方法.利用自适应空域滤波器分割出目标区域,再用二次投影法快速提取目标,并以模板评价函数为准则实时更新模板.具有模板自适应性强,目标判断准确,跟踪鲁棒性好等优点.外场实验表明,在目标变化和有物体遮挡干扰的情况下,系统能准确提取任意指定目标,并保持连续稳定的跟踪.

Design and implementation of target intelligent detecting and tracking system

To achieve the intelligent detection and steady tracking to an arbitrary target in the complex background, a real-time tracking system was designed, in which two DSPs were used in order to detach the data processing function from the control function. The hardware structure was modularized and a Direct User Interface (DUI) was designed so that the system could work independently without the upper computer. Furthermore, a new algorithm for extracting correlation template was proposed: Firstly, applied an adaptive special filter to partition the target area, and then extracted the target accurately using twi-projection method. The template was updated according to its evaluative function so that it had good adaptability and robust capability to the target. The field experiments show that although target is changeful or covered, the system can still extract it accurately and keep tracking it steadily.