3 Mission Critical Video Analytic Capabilities Today’s Police Need – And Why

More and more law enforcement agencies worldwide are using video surveillance to extend their monitoring and forensic capabilities. Video camera networks are growing exponentially, from cameras embedded in “smart” streetlamps to doorbell video cameras (for which private citizens may opt to share video feeds as evidence and intelligence with local police departments), the combination of video resources from private individuals and businesses and public infrastructure amount to an extensive system for real-time situational monitoring and evidence collection for post-incident investigations. Even without the significant increase in global surveillance cameras (Omdia estimates that by 2022 the number of installed cameras will surpass 1 billion), police are already struggling to actively monitor feeds in real-time and forensically review high volumes of video evidence: These are major challenges for resource-strapped police departments of all sizes.

Deep-learning based video content analytics software helps overcome these obstacles, by processing video to identify, categorize and index the objects in video footage (such as clothing, bags, vehicles, animals, and other items). Ultimately this drives mission critical analytics for law enforcement enabling them to:

• Easily search and comprehend video for accelerated investigations based on object classification and tracking, face recognition, and license plate recognition
• Seamlessly access video analytics capabilities and evidence in the field via mobile applications
• Attain situational awareness with real-time alerts, empowering officers to respond quickly to developing situations

Forensic Search Accelerates Investigations

Forensic video intelligence solutions enable officers to quickly and accurately conduct searches for objects and events of interest in video, combining distinct filters based on classes and attributes, and behaviors (e.g. speed, path, direction, dwell time, appearance similarity, color, and size). Such search functionality enables officers to review hours of footage in a matter of minutes, with precision. A prime example of this application is the Jussie Smollett case, where the Chicago Police Department was able to quickly review video evidence from hundreds of cameras to trace the movements of the suspects involved in the hoax attack.

Face Recognition

Face recognition – where legally permissible – is a means of pinpointing persons of interest, from suspects to missing persons. Police departments can leverage digital images extracted from video or from external sources to create face matching watchlists. By enabling video search based on specific faces or triggering real-time alerts whenever a face included (or – if configured thusly – excluded) in a face matching watchlist is detected, face recognition has become a critical tool for accelerating investigations and time-to-target.

License Plate Recognition

Another search functionality that is vital to law enforcement is license plate recognition; i.e., technology that detects license plates in video and transcribes the text on the plate. Like face recognition, license plate recognition has both constrained and cooperative access control applications, as well as “in the wild” uses. “In the wild” license plate recognition detects and identifies license plates in video surveillance environments – which means camera selection, placement, and setup that is not dedicated to (and sometimes challenging for) LPR functionalities. Consider the following scenario: an investigator reviews surveillance video in the aftermath of an incident and uses comprehensive video filtering and search to identify a specific car driving away from a crime scene. Using intelligent video surveillance, the operator can capture the license plate details and upload the license plate number to a watchlist for further monitoring. Officers can then search other video feeds for plate matches to gather evidence about the car and – potentially – the driver.

Mobile Access for Extending Video Search to the Field

While video search is a critical capability, many field investigators rely on dispatch officers to pinpoint video evidence and rarely leverage video review personally. In an age where police departments are equipping officers with smart phones and tablets in the field, so they can do their jobs more efficiently and safely, from any location, video content analytics is no exception: Departments can leverage mobile accessible video content analysis to extend video investigation capabilities into the field, so users can conduct video searches on-the-go. Mobile video search helps officers obtain near real-time intelligence in the aftermath of an incident and empowers them to make decisions about how proactively track and apprehend suspects – all without having to return to a crime center to process video footage.

Real-Time Alerting Increases Situational Awareness

Police officers are just as dedicated to preventing public safety or criminal incidents as they are to investigating them post-event. A video analytics solution that offers real-time alerts helps officers proactively respond to situations that could pose a hazard to public safety or indicate criminal behavior. Real-time alerts can be triggered based on:

Object Classification. In terms of object classification, a video content analytics system powered by artificial intelligence (AI) can be leveraged to discern normal patterns and identify anomalies. The Deep Learning-based system is trained to detect, identify and classify objects, and users can create rule-based alerting logic to trigger notifications based on the classified and indexed metadata when certain conditions – or a combination of different object classes and attributes – are met. This can be applied, for example, for detecting activity or human presence in an area that has been pre-defined as off-limits to pedestrian traffic.

 People Counting. Count-based alerts (such as people counting) are another form of notification that can be used to alert operators when the number of pedestrians or vehicles that meet certain attributes exceeds a pre-configured threshold, indicating possible traffic congestion or a breach of security in an off-limits area. These alerts can also be triggered when people counts falls below a threshold, indicating an empty guard post, for example.

 Face Recognition. When police are seeking a specific suspect who is at large, officers can add him or her to a suspect watchlist, compiled of face images extracted from video footage or external sources. They can then configure real-time alerts to trigger notifications when a suspect face match is detected in video feeds. This helps officers quickly locate a suspect, assess the situation and make an informed decision about how to respond. Conversely, police can also identify suspicious people by assembling watchlists of recognized individuals and triggering alerts for detected individuals that are excluded from that list.

Similarly, if officers are looking for a missing person, they can add a reference photo to pinpoint past appearances of that face that have been captured on video. Police can search video based on the approximate area and time the person disappeared to quickly understand where the person was before s/he went missing and configure real-time alerts to trigger an alarm whenever face matches are identified.

 License Plate Recognition. Officers can configure their intelligent video surveillance system to send real-time alerts if matches for license plates of interest are identified across camera networks.

The significance of powerful forensic and real-time analytics for video surveillance cannot be underestimated: For law enforcement, it can truly mean the difference between solving a case or proactively responding to unfolding situations to prevent an accident or a heinous crime. By deploying advanced video analytics technology, law enforcement can ensure optimal efficiency and efficacy in their operations, by maximizing existing video surveillance resources that they already use.

More on video analytics for law enforcement.