Challenge

The value of video for mobile public safety has been validated through the proliferation of video recorders in police cruisers, buses, and metropolitan rail cars. Conventionally, the benefit has been realized through labor intensive manual post-event analysis to support criminal prosecution and protect against frivolous lawsuits. However, with the advent of high-definition mobile cameras and broader distribution, it is impossible to manually monitor and label relevant activity in every deployed camera’s live video stream. In many scenarios, mobile analytics is more challenging than fixed video analysis, because both foreground (people and vehicles) and background are moving. The challenge is to capture and characterize the relevant activity and traffic information (such as a license plate) to provide real-time benefit and protection to public safety officials. Therefore, a mobile analytics solution must be fast, accurate, and scalable as better sensors and cameras are more widely deployed.

Benefits of SIG’s Approach

SIG’s mobile video analytics for public safety leverages SIG’s extensive experience at developing video analysis for airborne military sensors and classifying targets of interest using robust and proven automated target recognition (ATR) methods. SIG’s mobile analytics uses novel and proprietary image processing methods to sort out the relevant foreground activity from background motion. SIG’s analyst-in-the-loop (AIL) technology sorts through an overwhelming amount of imagery to cue a human operator to anomalous activity, uncertain situations, or rule-based behaviors to exploit expert feedback and adapt the algorithms for improved algorithm effectiveness and operator efficiency. SIG’s mobile analytics generates valuable metadata and video tags that enable enhanced forensic analysis. In particular, SIG is pleased to offer a novel automated license plate recognition (ALPR) solution that uses a multi-purpose standard wide-field camera for simultaneous multiple vehicle plate captures, rather than expensive, separate narrow field-of-view ALPR cameras.

Products/Solution

Superior analytics: SIG offers real-time image stabilization, tracking, target classification, and normalcy modeling in a PC-based form factor.

ALPR: wide-field of view multi-purpose color cameras are used to provide robust license plate recognition in multiple scenarios, including in-traffic, roadside scanning, and oblique angle viewing. The solution is robust to variations in state plates and custom plates.

SIG’s mobile analytics platforms can be used with a variety of standard digital camera formats and can be integrated with a wide range of existing mobile platform and back-office configurations.

Technology Summary

SIG’s mobile video analytics for public safety builds on the core capability developed for fixed cameras. SIG’s video analytics algorithms are enabled by SIG’s probabilistic video tracking framework which efficiently and robustly finds and tracks people and vehicles in complex environments. This unique video tracking software is founded on a Bayesian framework that models video pixel statistics for background as well as the appearance, shape, and motion of the objects being tracked. The underlying non-parametric statistical models capture the spatial and temporal pixel dynamics of the scene and the target’s articulations to accurately fit to the observed video and adapt to changing environmental conditions. The Bayesian video tracking framework is implemented with modern approximate inference methods for efficient and scalable computation. A higher-level analytics engine utilizes the output of the probabilistic tracker to detect specific activities of interest. Activities are modeled as specific trajectory events (e.g. a passenger slips and falls on a bus) and as combinations of trajectory events and shape-derived articulations (e.g. a passenger flees from a vehicle under investigation during a traffic stop). In addition, the analytics engine develops time-evolving models of normal aggregate behavior and detects individual behaviors that are anomalous given the learned model. Thus the system augments rule-based alerts by cueing the analyst to behaviors that may be of interest simply because they are unusual. The analytics engine incorporates the full probabilistic output of the underlying tracker to provide a mathematically rigorous confidence that a threatening or suspicious activity was observed.

SIG’s ALPR architecture uses a wide-field license plate detector to distinguish plates from other objects within the scene that contain alphanumeric characters such as traffic signs and advertisements. SIG’s character recognition software uses a novel character detection and extraction approach in conjunction with a Bayesian classification system. SIG’s solution doesn’t force hard decisions on each of the extracted characters, enabling robustness to custom designs and the ability to apply grammer-based rules for probabilistically optimal decisions at the plate level. This approach enables a formal confidence to be computed against local, state, and federal hot-lists.