移动通信模式下压缩感知视频检测技术及其目标重构  

Video Detection and Object Reconstruction by Compressive Sensing in Mobile Communications

作　　者：詹瑾[1,2] 赵慧民[1,2] 傅仁轩 肖茵茵[1,2] 李春英[1,2] 林智勇[1,2] ZHAN Jin ZHAO Huimin FU Renxuan XIAO Yinyin LI Chunying LIN Zhiyong(School of Computer Science, Guangdong Polytechnic Normal University Gnangzhou 510665, China Guangzhou Jiesai Science Technology Co., Ltd., Guangzhou S 10310, China Guangzhnu Key Laboratory of Digital Content Processing and Seturity Technology, Guangzhou 510665, China)

机构地区：[1]广东技术师范学院计算机科学学院,广东广州510665 [2]广州市数字内容处理及其安全性技术重点实验室,广东广州510665 [3]广州杰赛科技股份有限公司,广东广州510310

出　　处：《移动通信》2017年第11期44-51,共8页Mobile Communications

基　　金：国家自然科学基金(No.61672008;No.61571141);广东省自然科学基金(No.2016A030311013;No.2015A030313672);广东省应用型科技研发专项项目(No.2016B010127006;No.2015B010131017);广东省教育厅国际科技合作项目(No.2015KGJHZ021);广东省科技计划项目(No.2014A010103032)

摘　　要：为了解决移动通信视频监控的目标追踪问题,提出一种新的空间域视频压缩感知模型,该模型首先通过测量矩阵获取视频少量样本值,然后通过该样本值同时重构运动目标、背景和视频序列,最后通过视频序列估计得到一个置信图,可以进一步提高目标的重构质量。大量的实验证明,该模型与典型的空域检测技术比较,能够降低视频检测的数据量,并有效地重构视频目标,且对运动干扰具有更好的鲁棒性。Video detection and object reconstruction need to process a huge amount of data. In order to deal with the object tracking of the video monitoring in mobile communications, a novel video compressive sensing model （VCSM） based on spatial domain was proposed in this paper. In this model, a small number of sampled videos are obtained by means of the measurement matrix firstly. Then, the moving object, background and video sequences are reconstructed based on the samples. Finally, a confidence map is derived by estimating the video sequences to further enhance the construction quality if the object. Massive experiment results show that, compared with the detection techniques based on spatial domain, the proposed model can reduce the video data amount to be detected, effectively construct the video object and have good robustness.

关 键 词：压缩感知 视频检测 目标重构 鲁棒性 

分 类 号：TN929[电子电信—通信与信息系统]