中文摘要：

为解决复杂道路交通场景中的误检问题,提高跟踪效率,提出了一种基于离散小波变换(DWT)和卡尔曼(Kalman)滤波的多运动目标跟踪算法。本文利用DWT对背景和序列图像进行三层分解,仅将低频子图进行背景差分提取了运动目标。在跟踪阶段,建立区域-目标模型,以检测信息作为观测值并利用Kalman滤波进行位置预测;通过匹配预测值和观测值,建立包含稳定目标、丢失目标和新出现目标的三层目标链,并对稳定目标进行Kalman最优估计。采用在线计算获取各运动目标的初始位置和速度,使预测值一开始就接近观测值。通过对实际视频序列进行检测跟踪实验和对比,分析了算法性能。实验结果表明,本文方法具有较高的抗噪声能力、跟踪实时性和准确性。

英文摘要：

To solve error detection and enhance track efficiency in complex road traffic scenes,a new method for tracking multi-object is proposed based on discrete wavelet transform(DWT) and Kalman filter.This paper adopts the three-level DWT to decompose background and image sequences,and detects moving objects only using background difference among those low frequency sub-images.In tracking step,region-object model is built,and the positions of objects in next frame are predicted with the Kalman filter.Three-level ...