Final Report Summary - SMARTENC (Smart Video Encoders for Wireless Surveillance Networks)
The objective of the SMARTENC project is to jointly design video encoders and analytics embedded in networked surveillance cameras. The joint design aims at improving the video analysis performance in both embedded and central analytics engines and enhancing the video compression quality.
SMARTENC project is motivated by a few key observations on surveillance networks that can be redesigned to improve the efficiency and reduce the cost of such systems. Firstly, the video compression hardware used in cameras is decoupled from and not aware of the embedded video analytics, e.g. Event detection, object identification and tracking, face detection and recognition. Secondly, video compression quality is optimized for human viewers, not for automated video analysis. And lastly, video is usually compressed at a fixed bitrate regardless of the network topology and its time-varying characteristics.
Project outlined research goals in four main work packages:
• Test Environment Creation
• Development of Smart Video Encoding Techniques
• Network Adaptive Video Streaming Algorithm Design
• Integration and Evaluation
The main achievements of the project are as follows:
• Developed a video acquisition, processing, analysis, compression, streaming and display pipeline software named as: “Object Oriented Video Processing Architecture (OVA). OVA also became an integrated part of the ITEA2 (Information Technology for European Advancement) funded SPY (Surveillance Improved System) project. The researcher and the host organization also developed a camera platform called SONEC (Smart Open Networked Camera) that will be a perfect medium for the SMARTENC project field tests.
• Designed and developed real-time embedded video analytics and processing algorithms and integrated to the OVA architecture. These algorithms are now an integral part of host organization camera product line. The algorithms are:
- Moving object detection for land and aerial surveillance. The developed algorithms are presented and published in IEEE conference proceedings.
- Visual object tracking algorithms based on Harris features and correlation filters.
- Digital video stabilization based on image projections. The developed algorithm has been filed as an international patent application.
- Panoramic video generation algorithm for distributed aperture cameras.
- Algorithm to detect changes in a scene captured in different times of the day.
- Human detection for surveillance. Developed jointly with the FP7 funded SUBCOP (Suicide Bomber Counteraction and Prevention) project.
• Implemented video encoding algorithm for improving efficiency of embedded coders streaming real-time video over the network. This study is carried in collaboration with teams from researcher’s former employee Texas Instruments. The international collaboration resulted in a conference publication and a patent application.
The project was very fruitful for the researcher and the host organization. Innovative IR thermal cameras are added to the host organization product portfolio, patent applications are filed and work is presented in academic conferences. This shows the practicality of the project proposal and transfer of researchers’ expertise to the host organization. Project was also a success for the career development and reintegration of the researcher. Alignment of the project and host company product map improved the impact of the project.
SMARTENC project is motivated by a few key observations on surveillance networks that can be redesigned to improve the efficiency and reduce the cost of such systems. Firstly, the video compression hardware used in cameras is decoupled from and not aware of the embedded video analytics, e.g. Event detection, object identification and tracking, face detection and recognition. Secondly, video compression quality is optimized for human viewers, not for automated video analysis. And lastly, video is usually compressed at a fixed bitrate regardless of the network topology and its time-varying characteristics.
Project outlined research goals in four main work packages:
• Test Environment Creation
• Development of Smart Video Encoding Techniques
• Network Adaptive Video Streaming Algorithm Design
• Integration and Evaluation
The main achievements of the project are as follows:
• Developed a video acquisition, processing, analysis, compression, streaming and display pipeline software named as: “Object Oriented Video Processing Architecture (OVA). OVA also became an integrated part of the ITEA2 (Information Technology for European Advancement) funded SPY (Surveillance Improved System) project. The researcher and the host organization also developed a camera platform called SONEC (Smart Open Networked Camera) that will be a perfect medium for the SMARTENC project field tests.
• Designed and developed real-time embedded video analytics and processing algorithms and integrated to the OVA architecture. These algorithms are now an integral part of host organization camera product line. The algorithms are:
- Moving object detection for land and aerial surveillance. The developed algorithms are presented and published in IEEE conference proceedings.
- Visual object tracking algorithms based on Harris features and correlation filters.
- Digital video stabilization based on image projections. The developed algorithm has been filed as an international patent application.
- Panoramic video generation algorithm for distributed aperture cameras.
- Algorithm to detect changes in a scene captured in different times of the day.
- Human detection for surveillance. Developed jointly with the FP7 funded SUBCOP (Suicide Bomber Counteraction and Prevention) project.
• Implemented video encoding algorithm for improving efficiency of embedded coders streaming real-time video over the network. This study is carried in collaboration with teams from researcher’s former employee Texas Instruments. The international collaboration resulted in a conference publication and a patent application.
The project was very fruitful for the researcher and the host organization. Innovative IR thermal cameras are added to the host organization product portfolio, patent applications are filed and work is presented in academic conferences. This shows the practicality of the project proposal and transfer of researchers’ expertise to the host organization. Project was also a success for the career development and reintegration of the researcher. Alignment of the project and host company product map improved the impact of the project.