基于压缩感知和粒子滤波的改进絮体跟踪算法  

AN IMPROVED FLOC TRACKING ALGORITHM BASED ON COMPRESSIVE SENSING AND PARTICLE FILTER

作　　者：谢昕[1] 李慧萍[1] 胡锋平[2] 

机构地区：[1]华东交通大学信息工程学院,江西南昌330013 [2]华东交通大学土木建筑学院,江西南昌330013

出　　处：《计算机应用与软件》2015年第1期244-246,250,共4页Computer Applications and Software

基　　金：国家自然科学基金项目(61272197);江西省自然科学基金项目(20132BAB201027);江西省教育厅科技项目(GJJ13362);江西省研究生创新专项基金项目(YC2012-X017)

摘　　要：针对水处理过程中混凝絮体的跟踪问题,提出一种融合压缩感知与粒子滤波的絮体跟踪算法,即采用压缩感知技术提取絮体的图像特征,并以此进行单帧图像检测,得到检测值;同时通过粒子滤波实现非线性非高斯状态空间模型的絮体位置的最优估计,采用最优估计值和检测值进行数据关联,从而确定各个粒子的航迹以实现对絮体跟踪。实验结果表明该算法实现了絮体的实时跟踪及沉降速度的计算,有效地解决了获取图像特征时运算量大、效率低等问题,保证了跟踪的精度及效率。Aiming at the problem of flocculate tracking in water treatment process, we propose an floe tracking algorithm which combines compressive sensing （CS） with particle filter （PF）, that is, the feature of floes image is extracted with CS technology, and is used for singleframe image detection to get the detection value ; Meanwhile, PF is employed to realise the optimal estimation of flocs positions in non-linear and non-Gaussian state space model, and then the data association in regard to the estimated optimal value and the detection value is made, so as to determine the trajectory of each floe particle to achieve floes tracking. Experimental results demonstrate that the method realises the real-time tracking of floes and the setting velocity calculation, and effectively solves the problems of heavy computation and low efficiency in image features extraction, and thus guarantees the accuracy and efficiency of flocs tracking.

关 键 词：压缩感知 粒子滤波 絮体跟踪 沉降速度 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]