Articles

DETECTION AND IDENTIFICATION FOR VIDEO SURVEILLANCE SYSTEMS (DIVISS)

Intelligent video surveillance, or video analytics, is the next generation of security applications. Whether it is a crowded airport, a subway platform, a nuclear plant, an office high-rise, or a vigorously patrolled border, there is an acute need to be aware of any and all suspicious activity on the premises. AUG Signals has developed unique multi-sensor video registration and enhancement technologies to provide better detection and identification capabilities for video surveillance.
Video Motion Analysis (VMA) DIViSS has the capability to filter out moving objects that are inherent to a particular scene, without flagging it as possible security breach by placing a “virtual fence” around such disturbances. Rain, falling leaves, even moving cars will not trigger detection events, providing that these are intrinsic to the monitored environment, while objects and entities foreign to the milieu will be automatically detected. This allows robust, 24/7 outdoor system operation. The decision as to which objects should be filtered out can be programmed by the end user (i.e. the operator instructs the system which objects to ignore – this can be based on several characteristics, such as size, shape, velocity, etc.) or self-learned. The VMA feature allows DIViSS to learn the scene, either upon being prompted by the operator or independently. This involves the system briefly switching to a “learning mode” and then continuing normal operation. The VMA also provides the exact segmentation of each object, the number of moving objects, and the direction of movement. Static Detection DIViSS can detect and flag an object that was added (e.g., object left unattended) or removed (i.e. object missing) from the scene. This operation is performed automatically, in real-time, includes rotation and scaling factors determination, and does not require any prior knowledge of the object/image. CFAR Target Detection Target detection is a fundamental capability for all intelligent video surveillance applications. The false alarms inherent to this process are a problem that reduces the value of target surveillance. The implementation of AUG Signals’ cutting-edge multi-Constant False Alarm Rate (CFAR) detection technologies vary the detection threshold as a function of the sensed environment, significantly reducing the false alarm rate.  Sequential frame registration – camera stabilization Sequential images from individual sensors are registered by a global motion estimation method that achieves accurate image registration and provides the viewer with a smooth spatiotemporal evolution of events. Correct alignment of sequential images serves two purposes: Stabilizes camera(s) in order to eliminate possible sequential frame distortion due to camera vibration in outdoor environment or camera pan-tilt-zoom operation Provides aligned frame sequence for multi-frame deblurring and super-resolution enhancement using information fusion Sequential image registration is performed on a subpixel-level accuracy. The registration operations are performed on grayscale versions of the original frames to reduce computational complexity (to allow for real-time implementation) while maintaining almost the same level of accuracy. Fusion and enhancement – multi-frame fusion While other commercially-available products lack the ability to accurately register complex motions of multiple objects due to their inability to enhance images beyond single-frame based sharpening and de-noising, AUG Signals’ unique multi-frame registration and fusion techniques enable multi-frame based image enhancement (super-resolution capacity). Super-resolution increases spatial resolution of video images on a frame by frame basis to meet end user needs (e.g., assist online investigations and increase screening capabilities on request). Based on correctly aligned sequential frames, AUG Signals’ super-resolution algorithm obtains super-resolution image by performing multi-frame fusion. The algorithm is able to increase the spatial resolution far beyond the sensor’s resolution, facilitating the recognition/identification of fine details of an object (e.g., fine labels or other features that are smaller than the video sensor resolution).
Original	State of the art technology	AUG Signals multi-frame fusion
AUG Signals has also developed multi-frame blur reduction algorithms to compensate for the loss in resolution of moving objects that occur due to finite exposure time of cameras:
Original	State of the art technology	AUG Signals multi-frame blur reduction technology
Fusion and enhancement – multi-sensor fusion Recent advancements in sensing technologies have allowed the development of a wide range of cameras. These cameras range from light sensors (e.g. CCD, CMOS) to infrared (IR) sensors, electromagnetic radar sensors, thermal sensors, etc. These different types of sensors provide complimentary scene interpretation and surveillance capabilities. Multi-sensor capabilities allow video surveillance to be adapted for applications with less than ideal environments; for example, IR cameras can be used in low-light situations while CCD can be employed when sufficient lighting is available. Additionally, AUG Signals has developed the necessary technology to allow various sensors to be used simultaneously. The fusion of data streams coming in from multiple sensors provides a synergistically more comprehensive and accurate representation of the scene and enhanced object detection. In real time automatic implementation, computational complexity is one of the major requirements of multi-sensor registration algorithms. The state-of-the-art spatial domain registration methods are all computationally expensive. AUG Signals’ multi-sensor registration algorithm is very efficient in estimating the motion parameters between the reference image and each of the other images, achieving real time implementation while outperforming the other existing frequency domain registration methods.
Reference image from sensor 1 Reference image from sensor 2 Registered images
Video content authentication The security and protection of the content of video recordings are crucial for most video surveillance applications. AUG Signals’ watermark-based authentication techniques provide tampering tracking and analysis and ensure content integrity by imposing an encoded, invisible tag (i.e. the “watermark”). Cropping or otherwise modifying the image (including frame deletion and object removal and substitution) alters the watermark, which allows the user to easily identify and pinpoint image tampering as well as estimate the degree of tampering. In addition to incorporating the most advanced security measures, AUG Signals’ video content authentication distinguishes malicious attacks on content integrity from incidental distortions. AUG Signals’ video content authentication results are invariant to image format conversion, features low computational complexity and storage requirements, and introduces no significant delay in real time streaming applications.

AUTOMATIC IMAGE REGISTRATION (AIR)

Toronto, August 2008: AUG Signals, in collaboration with Akoostix Inc. and GeoSpectrum Technologies have selected by DRDC Atlantic to execute the Land Acoustics and Seismic Surveillance Literature Review. The purpose of this contract is to support DRDC Atlantic’s Land Acoustics and Seismic studies for Surveillance Operations (LASSO) project. The consortium will perform a rigorous investigation of acoustic and seismic sensors and tracking capabilities relevant to operational surveillance.

For more information, please contact the Project Manager, Dr. Abhijit Sinha at abhijit@augsignals.com

Toronto, July 2008: AUG Signals has secured a Defence Industrial Research (DIR) Program project entitled “Enhanced Military Capabilities with Rapid Covariance Equalization Change Detection.” Sponsored by the Department of National Defence, this two year project aims to develop a software product for macro-level change detection (i.e. detection and quantification of change) of SAR, IR, and electrooptical images. The innovation will be to develop Covariance Equalization (CE) change detection technique that will reduce the false changes detected because of the often unavoidable image misregistration. The resulting change detection technology can be applied strategically, operationally or tactically, with the focus on automatic and efficient processing of large volumes of data to flag regions of change at a macro-level for further analysis.

In operational theatres such as Afghanistan, adversaries make use of the local population as a means of concealing their activities from allied surveillance. For example, the IED (Improvised Explosive Device) system of supply and deployment is disguised amongst the day-to-day activities of the local population. Analysis of individual changes is of limited value in wide-area surveillance since there are too many changes occurring and specific changes of importance are difficult to identify. Macro-level change detection instead aims to take a high-level view and establish a normal range for the number of changes in an area of interest over a period of time. If the number of changes moves outside this band of normalcy, it is an indicator of activity in the area of interest and a possible precursor of an attack, a supply operation, or a movement of insurgents. Pattern recognition techniques in conjunction with human intelligence can be used to identify the nature of the change and can lead to a risk/threat evaluation.

The resulting capabilities will include identifying changes in enemy resources or activity during reconnaissance, detecting suspicious activity during surveillance, providing scouting ability for a moving convoy, and supporting situational awareness when protecting garrisons or high-value targets.

For more information, please contact the Project Manager, Dr. Ting Liu at tliu@augsignals.com

Toronto, March 2008: AUG Signals is pleased to announce we have secured a contract with Defence Research and Development Canada entitled “Flexible Association Algorithm Development”.  The objective of this work is to develop data association algorithms that show robustness in terms of identification estimates when faced with association errors. Target to measurement association – denoted as association problem in tracking literature – is an essential component and probably the most challenging part in target tracking. Errors in data association can cause higher and sometimes unaccounted uncertainty in target kinematics. It can lead to early termination, coalescence and delayed detection of tracks.

The developed algorithm will take into consideration the association related uncertainties in identification and will be able to recover from identification errors once association uncertainties are over.

For more information, please contact the Project Manager, Dr. Abhijit Sinha at abhijit@augsignals.com.

Toronto, September 2007: AUG Signals has been chosen to develop an Intelligent Video Surveillance System in partnership with InfoWrap Systems Ltd., an Israeli-based video surveillance company. Funded by the Canada-Israel Industrial Research and Development Foundation (CIIRDF), the objective of the project is to develop an intelligent, multi-sensor video surveillance system, to be launched late next year.

Detection and Identification for Video Surveillance Systems (DIViSS) integrates several unique and powerful capabilities into a video surveillance platform, revolutionizing how video cameras can be applied to surveillance applications. These include motion detection, automatic detection of abandoned objects, resolution enhancement, object identification, content protection, and integration of multiple sensors. Features are provided through a simple interface, allowing operators to easily use these advanced analysis functions.

Potential applications include: surveillance, identification of suspicious and/or abandoned objects in crowded areas, monitoring of confidential materials, traffic monitoring and many more.

For more information, or to request a demonstration of the system, please contact Tatyana Litvak.

Toronto, April 2007: AUG Signals is pleased to announce we have secured the Canadian Space Agency’s support for our new project, entitled “Conflation Capabilities for Spaceborne Remote Sensing: Automatic Co-registration and Vector Image Alignment.”

The objective of this work is to advance and expand the existing registration software of AUG Signals. Specifically, novel technologies will be developed to automatically register vector and image data, EO and SAR data at a user defined high registration accuracy.

The uniqueness of the technology lies in its simultaneous utilization of multiple common feature image pairs generated from originally different data sources. Due to the different image characteristics of the different sensors, automatic registration of EO and SAR is always a challenge. Poor accuracy of the geo-information of the vector data and/or image data causes large discrepancies between vectors and the same features presented in the image data when the two datasets are directly overlaid. AUG Signals’ new technology solves these problems by identifying correct control point pairs in the multi-layer feature image pairs. This will lead to novel approaches and solutions that will exploit currently available techniques resulting in increased registration/conflation performance for images from multiple sensors.

These conflation capabilities offer a significant value-added tool for registration – one of the fundamental tools for data analysis – and can be applied to a variety of diverse markets, from agriculture, forestry, and environmental applications to natural resources, topographic mapping, military and defence, etc.

For more information, please contact the Project Manager, Yifeng Li.

Toronto, February 2007: AUG Signals has been awarded a CSA (Canadian Space Agency) contract to develop a polarimetric constant false alarm rate (CFAR) detector using Markov chains. This is an innovative technique that will result in a new class of CFAR detectors, which may either stand alone or complement traditional CFAR detection performance. The anticipated CFAR detection improvement is expected to be as much as 30%.

Target detection is a fundamental analysis tool essential for almost all remote sensing applications. Unlike conventional CFAR detectors that have been designed to detect objects described by point data or with low image resolution, a high-performance Markov chain CFAR detector will make optimal use of the higher resolution information provided by RADARSAT-2. The Markov chain CFAR detector will exploit the finer resolution and selective polarization provided by RADARSAT-2 in order to yield greater detection performance.

This new technology can be used for a variety of military and defense applications such as coastal surveillance, ship detection, detection of land targets, iceberg monitoring, etc. In addition, Markov chain CFAR detector can also be employed in commercial applications, anything from agriculture (compliance monitoring, crop monitoring) and forestry to sea ice monitoring.

Markov chain CFAR detector will solidify Canada’s position as a global leader in remote sensing data provision, applications, and innovative technology.

For more information, please contact the Project Manager, Dr. Chuhong Fei.

Toronto, February 2007: AUG Signals has been chosen to lead an Intelligent Drinking Water monitoring project in partnership with EPCOR Water Services, University of Toronto, Public Health Agency of Canada, Queen’s University Centre for Water and the Environment, and University of Calgary. Funded by Precarn, the objective of the project is to develop an Intelligent Situation Assessment Unit (ISAU) that will detect drinking water contaminants by using inputs from multiple sensors and public health syndromic surveillance data. The water quality analysis will be communicated to water system operators, providing them with early warning of contamination and allowing for an immediate response.

Emerging sensor and intelligent systems technologies create an opportunity for drinking water distribution monitoring to be significantly enhanced. Current methods of laboratory sampling can be improved upon, reducing detection times and the costs of water monitoring. This is particularly true in smaller municipalities and rural regions, where water treatment procedures may not be as sophisticated and water testing facilities and expertise may be limited.

The fusion of sensor and non-sensor information pertaining to water quality is an entirely new application of intelligent systems and may lead to revolutionary methods of detecting drinking water contamination. The project establishes Canada as a leader in drinking water monitoring technology research and innovation.

For more information, please contact the Project Manager, Xia Liu.

Toronto, January 2007: AUG Signals has been awarded a Canadian Space Agency-supported contract to transfer AUG Signals’ automatic registration technology for temporal hyperspectral and polarimetric images from its original application of earth observation to applications for the defence industry and military. The technology will support the use of multiple types of sensors, including infrared, electro-optical, etc.

For more information, please contact Tatyana Litvak