软件工程系或计算机系毕业论文--运动目标检测跟踪基于卡尔曼滤波.docVIP

大连理工大学本科毕业设计（论文） 基于卡尔曼滤波器的运动目标检测与跟踪 Moving Objects Detecting and Tracking based on Kalman Filter 学 院（系）： 软件学院 专 业： 学 生 姓 名 学 号： 指 导 教 师： 评 阅 教 师： 完 成 日 期： 大连理工大学 Dalian University of Technology 摘 要 运动目标的检测和跟踪一直都是国内外计算机视觉领域研究的热点问题之一。如何在检测和跟踪过程中保证实时性、稳定性、高效性目前仍然是研究的重点。 本文在分析和参考现有研究成果和理论的基础上，设计并且实现了单个摄像头且摄像头静止情况下，对多个刚性运动目标实施实时、稳定、高效的检测与跟踪。主要内容如下： 在目标检测方面，采用了应用广泛的背景减除法，鲁棒较高，计算复杂度较低，准确度较高。首先我们通过平均背景法构建了背景图像，然后用当前图像进行差分，接着进行去噪处理，阈值化后得到前景目标，通过查找目标轮廓得到目标重心。 在目标跟踪方面，本文采用卡尔曼滤波器来实现系统中目标跟踪的任务。根据实际试验情况和参考文献确定了卡尔曼滤波器的参数以及具有预测稳定性的系统状态转移矩阵和观测矩阵，继而通过卡尔曼滤波器的预测和校正跟踪出运动目标的轨迹。 本文试验的结果证明了本文所采用算法的有效性，可以应用在实际场景中实时检测和跟踪多个运动目标。 关键词：目标检测；目标跟踪；背景差分；卡尔曼滤波器 Moving Objects Detecting and Tracking based on Kalman Filter Abstract Moving objects detecting and tracking has always been one of the hottest issues of computer vision research field. How to ensure the real-time surveillance, stability, high efficiency is still one of the most important issues of moving objects detecting and tracking process. In the analysis and reference to the existing research results and theoretical basis, this paper presents a moving object detecting and tracking method which is under the circumstance of one stationary camera and real-time surveillance. The main contents are as follows: In terms of detecting moving object, this paper uses the most popular background subtraction method to build background and uses the difference between image and background to get foreground objects. By finding contour, get centroids of foreground objects, then de-noising at the last. In terms of tracking moving object, method of Kalman filtering is discussed. According to requirement of system and circumstance, parameters of Kalman forecast are designed, system state vector, state transfer matrix and observation matrix are also determined. According to the prediction and correction