Video Surveillance
description
Transcript of Video Surveillance
Video SurveillanceCapturing, Management and Analysis of Security Videos.-Abhinav Goel-Varun Varshney
Introduction-why Surveillance at IIIT ?
Person Identification• Face Detection & Recognition
Crowd Patterns • Group Leader
Traffic Patterns/Crowd Recognition• Time when the area is crowded
Trend/Life Patterns• Hairstyle Patterns, Clothing
Patterns, Smoking habits etc.
Object Detection• Suspicious abandoned objects
Activity Detection• Coverage of Accidents,
Thefts ,vehicles arrived, rumbling crowd.
For the implementation of all the above and more analysis schemes the biggest need of the hour is : A scalable ,efficient and a robust system.
That is where we jump in !
Video Surveillance System : Design Objectives
System Necessity:
“Design of a system for the information retrieval application on security videos.
It should act as a platform for deployment and experimentation of various video analysis algorithms(as explained earlier) on large scale.”
Robust
Modular
Scalable
?
?admin
Robust: Should be resistant to crashes.Reliable. For eg. In case of Failure : Error Reporting to admin
Scalable : More external storage devices can be easily added.
Modular: System should be divided into smaller parts so that chances of failure reduce. Layered Architecture.
Layer-1
Layer-2
Layer-3
. . .
System Design : Modules
Continuous Camera Capture
Any camera can be installed which can be integrated with OpenCV.
The video capturing goes on continuously for 24 Hrs.
Storage DevicesVideo
Processing Station
Web ServerContinuous
Camera Capture.
Central Control
Server(CCS)
Web Server
•Hosts the interactive front end which has a user friendly interface to access the ‘Video Processing Results’.
•Sends a request a to the CCS to send the captured videos (which are temporarily stored here) to the ‘Processing Station’.
•Sends a request to CCS to search for a user requested video.
Video Processing Station
•Responsible for receiving videos from Web Server and running various Video Processing Algorithms.
•Sends a request to CCS to store the result in a suitable available storage device.
CCS
•Master Controller of the whole System.
•Accepts /Sends requests to other stations.
•Stores the meta-data corresponding to the current state of the system .(storage devices available, processing state etc.)•Stores complete meta-data of the processing results and their location.
System Process
Continuous Incoming Frames
Segmentation Process
Further large scale Video Processing
No of white pixels=n. If there are sufficient number of boxes with total white pixels greater than a threshold => activity frame.Capturing continues till a threshold amount of 'non-activity' frames are found.
Segmentation Process
System Process
Comparison with other techniques: It is better than Using a static image and doing background subtraction as the activity is studied at a box level.
User End:
1)A robust ,scalable system.
2) User can track all the activity videos by using a timeline.
3) A detailed census of the traffic is also available to the user along with a log of the recent activity.
4) Further features like a gallery of the faces captured in a shot are also visible(usage of OpenCV face detector)
System DemoSee Demo Video :(link on the website) or
Challenges and Statistics
Stress on System Stability: Robust & Easily Scalable.
Suitable Segmentation algorithm which can particularly capture distance activity changes
User End : Easy and efficient search of Videos
Determination of suitable thresholds to distinguish between ‘active’ and ‘static’ frames. Capturing+Storage+Processing+Control
Efficient System
Test: 14 Hrs continuous capture with web camera
No system crash ,17 activity videos,2 corrupted
videos
Capturing + Processing
No Storage scalability Less stable
Start: Naive one Server Design : All on one PC
System Crashes Delay in Processing
Technologies Used & Future Work
Front End: Mod_Python, Ajax, JavascriptDatabase : MysqlVideo Processing Tool : OpenCVProgramming Environment : Linux
This formulated system can be deployed easily at any Linux Environment with the above support.
•Installation of the external surveillance camera•Implementation of more Video Processing Algorithms to reduce Human Intervention!
-> Easy due to stable platform
Thank You