基于多路径粒子滤波目标跟踪

视觉跟踪是机器人自主导航、智能监控系统的关键技术.在粒子滤波跟踪算法中,设计粒子滤波器时,粒子大小是固定不变的,而目标与镜头之间是相对运动的,目标图像大小可能超出粒子范围,使得跟踪跟踪目标丢失.为了解决了上述问题本文提出了基于小波变换图像的粒子滤波方法.这里做法是在图像中目标上采集粒子,然后采用小波变换的分解算法和重建算法处理目标附近区域图像,再将采集到的粒子分成三部分,分别在原图和两种小波图像中传播.这样可以增强粒子滤波的鲁棒性,提高跟踪精度.

Target tracking based on multi-path particle filter

Visual tracking is a key issue for autonomous navigation,intelligent monitoring system and so on.While a Particle Filter for tracking is designed,the size of the target in image may be changed,witch could cause the target image out of given range of the particle.And the tracked target will be lost since the size of the target image keeps changing.A method for tracking a mobile target with visual image,particle filter with wavelet image(PFWWI),is proposed to solve the above problem in this paper.Here,the method is based on the wavelet transform incorporated in traditional Particle Filter,and particles are made to move in original image and the image processed by wavelet transform.The results show that the improved approach using multiple particles can enhance the robustness of tracking algorithm and improve tracking accuracy.