【作者】 王阿妮；

【导师】 马彩文；

【作者基本信息】 中国科学院研究生院（西安光学精密机械研究所） ， 信号与信息处理， 2009， 硕士

【摘要】 运动目标检测与跟踪技术是计算机视觉领域重要的研究方向之一,在现代工业、农业、国防以及空间技术等领域都有着广泛的应用。然而,在实际应用中,由于单一传感器往往不能提供足够的场景信息,所以,基于多源传感器的协同检测与跟踪技术已经成为当前的研究热点。本文基于红外与可见光模式,研究了多源传感器协同检测与跟踪技术,主要工作有:(1)研究了运动目标检测与提取技术。提出了一种基于自适应背景模型的运动目标检测法,并在此基础上,实现了目标提取。在多种复杂的自然环境下的实验表明,该方法能够较好地完成运动目标的检测和提取。(2)研究了运动目标跟踪技术。详细介绍了运动预测技术,并在此基础上,提出了一种基于扩展卡尔曼滤波器的运动目标跟踪方法,实验证明该方法具有一定的实际应用价值。(3)研究了单源跟踪结果评价因子和多源跟踪结果综合评价技术。首先研究分析了单源跟踪结果,确定了四个评价因子:目标面积惯性、目标宽长比惯性、目标位置惯性和目标速度惯性,并对各种评价因子建立了具体的数学模型。其次在单源跟踪结果评价的基础上,研究了多源跟踪结果综合评价技术,提出了基于灰色关联分析理论的综合评价技术。最后进行了仿真验证,结果证明了所建数学模型的正确性以及所提出方法的有效性。(4)研究了图像配准技术。针对红外与可见光图像,提出了两种图像配准方法,一个是基于子图像边缘相关性的红外与可见光图像配准方法,另一个是基于角点的图像配准方法,实验验证了这两种方法的有效性。更多还原

【Abstract】 Moving target detection and tracking technology is one of the most important research direction in the field of computer vision, and widely used in modern industry, agriculture, national defense and space technology. However, a single sensor cannot provide sufficient information of the scene in practical applications, so multi-sensor collaborative detection and tracking technology has become hot spots now.This paper focus on the multi-sensor collaborative detection and tracking technology based on infrared and visual images. The main work is as follows.(1) Moving target detection and extraction technologies are studied. A moving target detection method based on adaptive background model is adopted to achieve target detection, and on this basis, target extraction is accomplished. Experiments in complex natural environment show that the method can complete the moving target detection and extraction better.(2) Moving target tracking technology is researched. Movement prediction technology is introduced in detail, and on this basis, a moving target tracking method based on extended Kalman filter is brought forward. Experiments show that the method has some practical value.(3) The evaluation factors of single-sensor tracking results and synthesized evaluation technology of multi-sensor tracking results are studied. First, single-sensor tracking results are analysed. Four evaluation factors are chosen: target area inertia, target width length ratio inertia, target displacement inertia and target speed inertia. Specific mathematical models are established for all evaluation factors. Second, synthesized evaluation technology of multi-sensor tracking results is studied based on the single-sensor evaluation. A multi-sensor evaluation method based on the theory of gray relational analysis is proposed. Finally, the simulation is carried out, and the results prove the correctness of the mathematical model and the effectiveness of the method.(4) The image registration technology is studied. Two image registration methods are put forward for infrared and visual images. One is based on the edge relevance of the sub-image, and the other is based on the corner of the image. Experients show that the algorithms are effective.更多还原