基于码本模型和压缩跟踪算法的遗留物检测.pdfVIP

11 VIDEO APPLlCA恶TFZ525 γ，理军费 [本文献信息]张辉，徐伟，谢正光，等.基于码本模型和压缩跟踪算法的遗留物检测[J] .电视技术，2014 ， 38(23). 基于码本模型和压缩跟踪算法的遗留物检测 张辉，徐伟，谢正光，蒋小燕，马文莹 (南通大学电子信息学院，江苏南通 226019) [摘 要]针对一般遗留物检测算法复杂度高和跟踪效果不理想的问题，提出了一种基于码本模型和压缩跟踪算法相结合的 遗留物检测方法。首先通过码本模型建模和适时匹配背景更新算法来获取静止目标区域信息:然后利用稀疏测量矩阵对静 止目标区域的多尺度特征进行降维，得到分类器的正负样本;最后用朴素贝叶斯分类器对提取的特征进行分类，当分类器响 应最优时得到当前帧中跟踪到的目标位置，即使目标被部分遮挡，也能实现对遗留目标的准确跟踪。实验结果表明，该方法 不仅简单高效、实时性好，而且可以消除由物体短暂停留而带来的干扰。 [关键词]遗留物检测;码本模型;背景更新:压缩跟踪;实时 [中图分类号] TN91 1. 73 [文献标志码】 A Abandoned Object D毗ction Based on Codebook Model and Compressive Tracking Algorithm ZHANG Hui , XU Wei , XIE Zhengguang , JIANG Xiaoyan , MA Wenxuan ( College 01 Electronics and I阶， nnation ， Na脚ng Unive，哟， Jia咿u Nantong 226019 , China) [ Abstract] The general abandoned object detection algorithms proposed are either highly complex or weak in tracking abandoned object Therefore , an abandoned object detection method based on codebook model and cornpressive tracking algorithm is proposed. FirsÙY , the stationary target area in- formation is obtained through codebook background modeling and timely matching background updale algorithm. Then , a veηsparse measurernent rna- trix is used to reduce the dimensions of multi - scale irnage features , which 缸e acted as the positive and negative sarnples in the classifier , for the sta- tionary t盯get area. At last , those features extracted above 缸e classified via a naive Bayes classifier. As a result , the tracking location is found in the currenl frarne 切th 由ern缸irnal CI甜甜er response. Even when partia Ily covered , the abandoned object can still be accurately tracked. The results show Ihal 出e proposed method is nol only simple , efficienl