粒子滤波器在图像序列目标跟踪中的应用

Target Tracking Based on Image Sequence Using Particle Filter

粒子滤波器是一种根据带有噪声的观测数据序列估计未知运动状态的技术,它主要用于非线性、非高斯的信号处理系统,它由状态转换模型和观测模型两个部分构成,其基本思想是用一组带权的粒子来表示随机变量的后验概率分布。该文中以图像序列运动目标的位置为未知运动状态变量,相邻两帧图像经过全局运动补偿后的差图像为观测数据,针对室内环境单个步行者的情况,提出了一种简单有效的基于运动检测的状态转换模型和观测模型。实验结果表明,该模型具有良好的跟踪性能。

Particle filter is an inference technique for estimating the motion state from a noisy collection of observations. Generally, it is used in nonlinear and nongaussian signal processing system. Two important components of this approach are state transition and observation models and the basic idea of it is that the posterior density is approximated by a set of discrete samples - particles. Given that the position of the moving target is the unknown motion state and the difference images by ego - motion compensation are the sequent!al sensor data, we build a simple motion model and observation model based on motion detection for a pedestrian in the environment indoors. The experiment results show that the algorithm performs well.