WO2016095696A1 - 一种基于视频大纲的监控视频可伸缩编码方法 - Google Patents

一种基于视频大纲的监控视频可伸缩编码方法 Download PDF

Info

Publication number
WO2016095696A1
WO2016095696A1 PCT/CN2015/095913 CN2015095913W WO2016095696A1 WO 2016095696 A1 WO2016095696 A1 WO 2016095696A1 CN 2015095913 W CN2015095913 W CN 2015095913W WO 2016095696 A1 WO2016095696 A1 WO 2016095696A1
Authority
WO
WIPO (PCT)
Prior art keywords
video
outline
scalable
cost
key
Prior art date
Application number
PCT/CN2015/095913
Other languages
English (en)
French (fr)
Inventor
彭力
Original Assignee
江南大学
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 江南大学 filed Critical 江南大学
Publication of WO2016095696A1 publication Critical patent/WO2016095696A1/zh

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/30—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using hierarchical techniques, e.g. scalability
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00—Information retrieval; Database structures therefor; File system structures therefor

Definitions

  • the present invention relates to video information compression techniques.
  • control terminals with multi-channel video surveillance on the car at home and abroad can not actively analyze and report the abnormal behaviors inside and outside the car, such as the driver dozing off, the elderly and the pregnant women fall, etc., by setting up on the bus.
  • the camera on the camera quickly collects image and video information, and can perform key frame extraction, motion target analysis, video information compression, image fusion reconstruction, and finally can locate and track key parts such as the interior of the car and the luggage compartment, using artificial intelligence technology. Identify the number of people on board, the gender of the passengers and the approximate age to better manage and control everything in the car.
  • Scalable coding is a solution that can well adapt to the diversity of the terminal and the dynamic heterogeneity of the network. It only needs to encode the high-quality video source once, get the highest-level video stream, and extract the specific part from the code stream.
  • Basic quality code stream when the terminal does not support high-level code stream or network congestion, it can transmit only the basic quality code stream to the terminal.
  • the existing video outline technology cannot perform scalable browsing of the outline information of the surveillance video, and the protection of the key part of the fidelity is not satisfactory. Therefore, it is of great theoretical significance and application value to explore the new scalable coding technology of video information based on video outline technology and apply it to the vehicle intelligent monitoring system.
  • the main purpose of the invention is to solve the detection, identification and alarm of the bus accident by image video technology.
  • the key task is to propose a practical video outline technique that generates an outline video and preserves the original information to the greatest extent possible by analyzing the complete video, extracting, mapping and reconstructing the main moving objects and background images.
  • the outline video removes the space-time domain redundancy of the original video, the length is greatly shortened, thereby reducing the burden of encoding, transmission, and storage.
  • the outline video retains the main motion information of the original video, which can be used to quickly browse and retrieve the video in the face of the increasing amount of video data. How does the video outline technology improve the key details in a limited storage space?
  • the video compression method has a large calculation amount, a long operation time, poor stability and feature extraction performance, and cannot meet the reliability and rapid real-time requirements of the vehicle during operation. Changing the time-space mapping relationship can result in a more compressed video with a higher degree of compression.
  • the outline video will inevitably have many trajectories of breaks and loss of detail, and some of them can express the time details of key information such as entering, going out, meeting, etc. Affects the ability of the outline video to restore original video information. How to take into account the details and improve compression efficiency is also an important issue.
  • the coding process of the whole scheme is as follows: firstly, the coded surveillance video is analyzed with the video outline technology based on detail retention, and the moving object region information and the moving object mapping information are obtained; the ROI region in the video coding is determined based on the object region information, based on the object
  • the mapping information determines a stripe type in the video encoding, and the present invention encodes the motion object mapped to the initial outline video and its background into a P slice, and initializes the motion object not mapped to the outline video and its background code as a B slice; Then, based on the coding scheme, the original surveillance video is subjected to ROI-based scalable video coding, and the moving object flag bit and the motion region flag bit are encoded and combined into an object flag bit; finally, a scalable code stream containing the object flag bit is generated. Thereby, the scalable transmission and the scalable browsing of the surveillance video code stream are realized, and the structure is shown in FIG. 2.
  • the biggest advantage of the motion object-based video outline is that the motion object itself and its motion trajectory are monitored, analyzed and displayed as a complete unit. Therefore, the real-time, integrity and accuracy of the motion object extraction is the key to the system implementation.
  • Optical flow method, frame difference method and background subtraction method have been applied to the video outline system as the object extraction method, but considering the complexity of the algorithm, the extraction accuracy and the integrity, the present invention compromises the mixed Gaussian model based on pixel points. Background modeling method extraction.
  • Moving object tracking is the basis for determining trajectories, behavioral analysis, and determining mapping functions.
  • Common methods in recent years include the mean shift by Fukunag and the particle filter proposed by Gordon et al.
  • Mean Shift is a deterministic algorithm with low computational complexity; the particle filter is a statistical algorithm with high accuracy.
  • the present invention adopts the Mean Shift algorithm with low algorithm complexity.
  • the present invention proposes a video outline technique based on detail retention, which is shown in Figure 3.
  • the core idea is to find and process the key details based on the object, and then introduce the detail cost and the non-detail cost to replace the original motion loss, to update the traditional cost function, and obtain the minimum value of the new cost function through the simulated annealing algorithm.
  • the division of detail cost and non-detail cost in the motion cost can focus on the key details, while the sliding window processing technology for key frames can solve the mutations such as motion trajectory break in the outline video. problem.
  • Figure 1 is a scalable video outline framework.
  • Figure 2 is a mass scalable coding framework.
  • Figure 3 is a block diagram of a surveillance video scalable coding based on detail retention.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Data Mining & Analysis (AREA)
  • Databases & Information Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Compression Or Coding Systems Of Tv Signals (AREA)
  • Closed-Circuit Television Systems (AREA)

Abstract

公开了一种基于视频大纲的监控视频可伸缩编码方法，对待编码的监控视频使用基于细节保留的视频大纲编码方法进行大纲分析，对主要的运动对象和背景图片提取、映射和重建，生成一个大纲视频，得到运动对象区域信息和运动对象映射信息；基于对象区域信息确定视频编码中的ROI区域，基于对象映射信息确定视频编码中的条带类型；然后对原始监控视频进行基于ROI的可伸缩视频编码，同时将运动对象标志位和运动区域标志位进行编码并合并为对象标志位；生成含有对象标志位的可伸缩码流，以实现监控视频码流的可伸缩传输和可伸缩浏览。

Description

[根据细则37.2由ISA制定的发明名称]　一种基于视频大纲的监控视频可伸缩编码方法 技术领域
本发明涉及视频信息压缩技术。
背景技术
目前国内外已经出现了车载带多路视频监控的控制终端，但多数只是监视，不能主动对车内外异常行为状况主动分析报警，比如司机打瞌睡、老人小孩孕妇摔倒等，通过架设在大巴车上的摄像头快速采集图像视频信息，能够进行关键帧提取、运动目标分析、视频信息压缩、图像融合重构，最终可以对车内、行李舱等关键部位进行定位和跟踪，利用人工智能技术，可以识别上车人数、乘客性别和大致年龄，进而更好地管理和控制车内一切情况。
随着视频应用的广泛普及，监控视频数量和清晰度的大幅增加造成的海量数据使得其浏览和存储成为非常棘手的问题。一方面，完整浏览长达几天的视频会耗费大量时间和人力，快进又容易导致重要信息的遗漏；另一方面，监控视频的大部分帧中并没有运动对象，很多浏览时间会被白白浪费，即使出现了运动对象，也往往分布非常稀疏，每帧的画面空间存在很大冗余。因此，视频的时空域冗余没有得到合理消除，视频编码的目标也由单纯追求高压缩率转向了适应更加多元化的网络带宽和不同终端的处理显示能力。可伸缩编码是一种可以很好适应终端多样性和网络动态异构性问题的方案，只需要对高质量视频源进行一次编码，得到最高层视频流，从该码流中提取特定的部分作为基本质量码流，当终端不支持高层码流或者网络拥堵时，就可以只传输基本质量码流到终端。现有的视频大纲技术无法对监控视频的大纲信息进行可伸缩浏览，对关键部分保真度的保护也不尽人意。因此，探索新的基于视频大纲技术的视频信息的可伸缩编码技术，并将其应用到车载智能监控系统中，具有重要的理论意义和应用价值。
发明内容
本发明主要目的是通过图像视频技术解决大巴车突发事件的检测识别和报警。关键任务是要提出一种实用的视频大纲技术，通过对完整视频的分析，对主要的运动对象和背景图片提取、映射和重建，生成一个大纲视频并尽最大可能保留原始信息。一方面，由于大纲视频去除了原始视频的时空域冗余，长度大大缩短，因此减轻了编码、传输和存储负担。另一方面，大纲视频保留了原始视频的主要运动信息，可以方便的对视频进行快速浏览和检索面对日益增长的海量视频数据，视频大纲技术如何在有限的存储空间内尽可能提高关键细节的保真度、如何满足用户越来越灵活的查询浏览需求、如何提高编解码的效率、如何进一步降低存储空间和传输网络的压力，这些问题已成为多媒体领域的热点问题，具有深远的研究意义和广泛的应用前景。
在对传统视频大纲技术、传统可伸缩视频编码技术和基于ROI的可伸缩视频编码技术的研究与 分析的基础上，提出一种新的基于视频大纲的监控视频可伸缩编码技术。它是一种基于细节保留的视频大纲技术，对原始视频的关键部分和非关键部分采用不同的浓缩强度，并且相应地更新视频大纲中代价函数的计算方法，这样将有益于在相同存储代价下最大可能地表现关键信息，有望大幅提高图像压缩效率，整体监控视频压缩结构见图1。
如何在相同存储代价下最大可能地表现关键信息，大幅提高图像压缩效率这是项目的主要解决的难题。目前视频压缩方法计算量大、运算时间长、稳定性与特征提取性能差，不能满足车载运行时的可靠性、快速实时性要求。改变时空域映射关系可以获得更高压缩程度的浓缩视频。但是一段浓缩视频的长度一旦确定，大纲视频不可避免的会存在很多运动轨迹的断裂和细节丢失，其中不乏一些能够表现进入，走出，相遇等关键信息的时间细节，这种情况在很大程度上影响了大纲视频还原原始视频信息的能力。如何将细节考虑进去，提高压缩效率也是一个重要问题。
整个方案的编码流程如下：首先对待编码的监控视频用基于细节保留的视频大纲技术进行大纲分析，得到运动对象区域信息和运动对象映射信息；基于对象区域信息确定视频编码中的ROI区域，基于对象映射信息确定视频编码中的条带类型，本发明将映射到初始大纲视频中的运动对象及其背景编码为P slice，将初始化没映射到大纲视频中的运动对象及其背景编码为B slice；然后基于该编码方案对原始监控视频进行基于ROI的可伸缩视频编码，同时将运动对象标志位和运动区域标志位进行编码并合并为对象标志位；最终生成含有对象标志位的可伸缩码流，从而实现监控视频码流的可伸缩传输和可伸缩浏览，结构见图2。
基于运动对象的视频大纲最大的优势就是运动对象本身及其运动轨迹被作为一个完整的单元进行监测、分析和显示，因此运动对象提取的实时性、完整性和准确性是系统实现的关键。光流法、帧差法、背景消减法都曾被作为对象提取方法应用于视频大纲系统，但是综合考虑算法复杂度、提取精确度和完整性，本发明折中选择基于像素点的混合高斯模型背景建模方法提取。
运动对象跟踪是确定轨迹，行为分析和确定映射函数的基础。近年来常见的方法有由Fukunag提出的均值飘移和Gordon et al提出的粒子滤波器，其中Mean Shift属于确定性算法，计算复杂度低；粒子滤波器属于统计性算法，准确性高。但是考虑监控视频大纲分析对实时性的要求，本发明采用算法复杂度较低的Mean Shift算法。
基于视频压缩，本发明提出了一种基于细节保留的视频大纲技术，结构见图3。其核心思想是找到基于对象的关键细节进行处理和标记，然后引入细节代价和非细节代价代替原有的运动损失，来对传统代价函数进行更新，经模拟退火算法得到新代价函数的最小值。在本方法中，运动代价中细节代价和非细节代价的划分能够对关键细节进行有重点的保留，而对关键帧的滑动窗口处理技术则能很好的解决大纲视频中运动轨迹断裂等突变性问题。
附图说明
图1是可伸缩视频大纲框架。
图2是质量可伸缩编码框架。
图3是基于细节保留的监控视频可伸缩编码结构图。

Claims (3)

  1. 本发明特征是要提出一种实用的视频大纲技术，通过对完整视频的分析，对主要的运动对象和背景图片提取、映射和重建，生成一个大纲视频并尽最大可能保留原始信息。
  2. 在对传统视频大纲技术、传统可伸缩视频编码技术和基于ROI的可伸缩视频编码技术的研究与分析的基础上，提出一种新的基于视频大纲的监控视频可伸缩编码技术。它是一种基于细节保留的视频大纲技术，对原始视频的关键部分和非关键部分采用不同的浓缩强度，并且相应地更新视频大纲中代价函数的计算方法，这样将有益于在相同存储代价下最大可能地表现关键信息，有望大幅提高图像压缩效率。
  3. 本发明其核心思想是找到基于对象的关键细节进行处理和标记，然后引入细节代价和非细节代价代替原有的运动损失，来对传统代价函数进行更新，经模拟退火算法得到新代价函数的最小值。在本方法中，运动代价中细节代价和非细节代价的划分能够对关键细节进行有重点的保留，而对关键帧的滑动窗口处理技术则能很好的解决大纲视频中运动轨迹断裂等突变性问题。
PCT/CN2015/095913 2014-12-15 2015-11-30 一种基于视频大纲的监控视频可伸缩编码方法 WO2016095696A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201410782676.0 2014-12-15
CN201410782676.0A CN105898313A (zh) 2014-12-15 2014-12-15 一种新的基于视频大纲的监控视频可伸缩编码技术

Publications (1)

Publication Number Publication Date
WO2016095696A1 true WO2016095696A1 (zh) 2016-06-23

Family

ID=56125883

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2015/095913 WO2016095696A1 (zh) 2014-12-15 2015-11-30 一种基于视频大纲的监控视频可伸缩编码方法

Country Status (2)

Country Link
CN (1) CN105898313A (zh)
WO (1) WO2016095696A1 (zh)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108873762A (zh) * 2017-05-11 2018-11-23 上海航天电子有限公司 空间实验用多路视频采集实验控制器
CN110719438A (zh) * 2019-08-28 2020-01-21 北京大学 一种数字视网膜视频流与特征流的同步传输控制方法
CN116744006A (zh) * 2023-08-14 2023-09-12 光谷技术有限公司 基于区块链的视频监控数据存储方法

Families Citing this family (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109151469B (zh) 2017-06-15 2020-06-30 腾讯科技（深圳）有限公司 视频编码方法、装置及设备
CN109168032B (zh) * 2018-11-12 2021-08-27 广州酷狗计算机科技有限公司 视频数据的处理方法、终端、服务器及存储介质
CN110427865B (zh) * 2019-07-29 2023-08-25 三峡大学 高电压禁止区域人类行为视频特征图片提取与重构方法
CN110717422A (zh) * 2019-09-25 2020-01-21 北京影谱科技股份有限公司 基于卷积神经网络的交互动作的识别方法和系统
CN111933156B (zh) * 2020-09-25 2021-01-19 广州佰锐网络科技有限公司 基于多重特征识别的高保真音频处理方法及装置
CN112804188B (zh) * 2020-12-08 2021-11-26 鹏城实验室 一种可伸缩视觉计算系统

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101262568A (zh) * 2008-04-21 2008-09-10 中国科学院计算技术研究所 一种产生视频大纲的方法和系统
WO2012019417A1 (zh) * 2010-08-10 2012-02-16 中国科学院自动化研究所 在线视频浓缩装置、系统及方法
CN102395029A (zh) * 2011-11-05 2012-03-28 江苏物联网研究发展中心 一种支持视频可伸缩浏览的视频编解码方法和装置
CN102905200A (zh) * 2012-08-07 2013-01-30 上海交通大学 一种视频感兴趣区域双流编码传输方法及系统
WO2013178172A1 (zh) * 2012-08-30 2013-12-05 中兴通讯股份有限公司 一种监控视频摘要的方法及装置
CN103596011A (zh) * 2013-11-20 2014-02-19 北京中星微电子有限公司 图像数据的存储处理方法和装置

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103108160B (zh) * 2013-01-24 2016-08-03 中国联合网络通信集团有限公司 监控视频数据获取方法、服务器和终端
US10165227B2 (en) * 2013-03-12 2018-12-25 Futurewei Technologies, Inc. Context based video distribution and storage
CN105306945B (zh) * 2014-07-10 2019-03-01 北京创鑫汇智科技发展有限责任公司 一种监控视频的可伸缩浓缩编码方法和装置

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101262568A (zh) * 2008-04-21 2008-09-10 中国科学院计算技术研究所 一种产生视频大纲的方法和系统
WO2012019417A1 (zh) * 2010-08-10 2012-02-16 中国科学院自动化研究所 在线视频浓缩装置、系统及方法
CN102395029A (zh) * 2011-11-05 2012-03-28 江苏物联网研究发展中心 一种支持视频可伸缩浏览的视频编解码方法和装置
CN102905200A (zh) * 2012-08-07 2013-01-30 上海交通大学 一种视频感兴趣区域双流编码传输方法及系统
WO2013178172A1 (zh) * 2012-08-30 2013-12-05 中兴通讯股份有限公司 一种监控视频摘要的方法及装置
CN103596011A (zh) * 2013-11-20 2014-02-19 北京中星微电子有限公司 图像数据的存储处理方法和装置

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108873762A (zh) * 2017-05-11 2018-11-23 上海航天电子有限公司 空间实验用多路视频采集实验控制器
CN110719438A (zh) * 2019-08-28 2020-01-21 北京大学 一种数字视网膜视频流与特征流的同步传输控制方法
CN116744006A (zh) * 2023-08-14 2023-09-12 光谷技术有限公司 基于区块链的视频监控数据存储方法
CN116744006B (zh) * 2023-08-14 2023-10-27 光谷技术有限公司 基于区块链的视频监控数据存储方法

Also Published As

Publication number Publication date
CN105898313A (zh) 2016-08-24

Similar Documents

Publication Publication Date Title
WO2016095696A1 (zh) 一种基于视频大纲的监控视频可伸缩编码方法
CA2967495C (en) System and method for compressing video data
US9846820B2 (en) Method and system for coding or recognizing of surveillance videos
WO2016173277A1 (zh) 视频编码方法、解码方法及其装置
CN112991656A (zh) 基于姿态估计的全景监控下人体异常行为识别报警系统及方法
CN102395029B (zh) 一种支持视频可伸缩浏览的视频编解码方法和装置
WO2018150083A1 (en) A method and technical equipment for video processing
CN110324626A (zh) 一种面向物联网监控的双码流人脸分辨率保真的视频编解码方法
CN102663375B (zh) H.264中基于数字水印技术的主动目标识别方法
Wang et al. Intermediate fused network with multiple timescales for anomaly detection
CN110290386B (zh) 一种基于生成对抗网络的低码率人体运动视频编码系统及方法
CN111586412B (zh) 高清视频处理方法、主设备、从设备和芯片系统
Gan et al. Video object forgery detection algorithm based on VGG-11 convolutional neural network
Ji et al. Tam-net: Temporal enhanced appearance-to-motion generative network for video anomaly detection
CN103826125A (zh) 用于已压缩监控视频的浓缩分析方法和装置
CN111787218A (zh) 一种基于数字视网膜技术的监控相机
CN101237581A (zh) 基于运动特征的h.264压缩域实时视频对象分割方法
CN103957423A (zh) 一种基于计算机视觉的视频压缩和重建方法
Hu et al. UAV image high fidelity compression algorithm based on generative adversarial networks under complex disaster conditions
Nandini et al. Automatic traffic control system using PCA based approach
Ouyang et al. The comparison and analysis of extracting video key frame
CN112866715B (zh) 一种支持人机混合智能的通用视频压缩编码系统
Verma et al. Intensifying security with smart video surveillance
CN110674347B (zh) 视觉屏蔽双层ap视频摘要生成方法
Zhao et al. Research on human behavior recognition in video based on 3DCCA

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 15869204

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 15869204

Country of ref document: EP

Kind code of ref document: A1