The Security Industry Association’s Public Safety Working Group and Retail Security Interest Group have collaborated to develop detailed information on recommended technologies supporting incident monitoring and investigations in the retail sector. Here you will find a list of security solutions aiming to improve security for retail owners and assist law enforcement in conducting investigations against petty theft, organized retail crime and other criminal behavior against the retail sector.
There are some pressing security and retail industry issues that so far remain unresolved; these are areas where technology is ahead of the industry. One example is integrated systems. While most security managers would agree that security systems should be fully integrated (e.g., intrusion, access control and video), most systems today are still stand-alone systems. End users are rapidly replacing “closed” appliance-based solutions with platforms linking security devices for scalability, agility and elasticity.
CCTV vs. Video Surveillance
Video surveillance and other object identification sensors may be used, especially to support law enforcement and loss prevention investigators. In general, CCTV implies an analog, proprietary, closed circuit producing images at lower resolution than HD video. Today’s use of digital video surveillance cameras, often referred to as IP video surveillance cameras, have sensors capable of 720p HDTV and 1080p HDTV standards.
Video analytics solutions are a set of computerized vision algorithms that automatically analyze live or recorded video streams without the need for human intervention. The output of these algorithms results in actionable data for retail stores. Adding video analytics to a surveillance network allows the retail and loss prevention professional to be more effective in the detection, prevention, response to, and investigation of incidents captured by the surveillance system.
Video content analysis (VCA) is commonly used in support of security and perimeter protection, traffic monitoring and asset protection.
VCA can be used to detect specific objects or behaviors. Based on the application, detection of those objects or behaviors can be used to initiate alerts, object/feature tracking or other responses to the detected object or behavior or to exclude (i.e., filter) additional information about those objects or behaviors. Some common VCA functions include:
- Target type detection – Identification within video data of objects such as people, vehicles and static objects
- Event type detection – Identification within video data of behaviors such as moving, remaining stationary, crossing a line, occupying a space or being part of a crowd
- Filter by color – Identification of objects of a specified color
- Filter by size – Identification of objects based upon size
- Filter by defined time Ranges – Identification of objects or events based upon the time of detection
- Search on selected cameras or group of cameras – Detection of objects or events within the field of view, such as a store entrance
- Search for similar targets – Identification within a video data set of objects having similar attributes
These functions are most commonly used to support the following analyses:
- Accurate, wide-ranging statistical data related to people and vehicles.
- Multiple viewing options for statistical analysis of traffic volumes (people/vehicles), including numerical charts and user-friendly graphs to enable traffic comparisons, aggregates and identification of traffic trends.
- Advanced visualization options of heat map and target path to analyze movement trends and motion patterns of people through stores, enabling comprehension of highest customer activity and dominant foot traffic paths
Anonymous Location of People
In addition to the previously described “heat” or people activity map, retailers are currently using WiFi access points to anonymously detect where people are gathering in a store or mall or even outside the building. A real-time location system, two-way communication and analytics in coordination with existing physical security systems including access control, video surveillance, intercoms, intrusion detection and emergency management systems can be life-saving measures in the event of mass casualties or active shooter scenarios.
Recognition Level Examples
The video surveillance sensors can provide these increasing (1=lowest recognition level; 5=highest) levels of forensic evidence suitable for use in investigations, especially useful in retail store environments:
- General object recognition differentiating a person, animal or vehicle using video surveillance, thermal imaging, infrared (IR) depth light detection and ranging (LIDAR) sensors
- The number of people entering a portal or crossing a boundary during a given time period with any of the above sensors
- A fallen person detected by visible light, thermal imaging sensor or IR depth sensors or “skeletal recognition” systems using a human fall down algorithm
- Recognition of gender through video surveillance and a facial recognition algorithm
- Small population (less than 5,000 facial images) “face-matching” processes video surveillance data for face locations in a frontal pose, then process using facial recognition algorithms
Large population (greater than 5,000 facial images) face recognition algorithms also are used by social media services like Facebook, which locates faces in frontal posse in images and processes image biometric data through collection, analysis, automatic or manual tagging and storage. Since the processing is done primarily on static digital images, often at low resolution, this category is generally not considered as a verification tool.
Electronic Article Surveillance (EAS)
EAS is meant to deter theft in retail stores and may be integrated into the video surveillance system. If a non-purchased item leaves a store, the radio-frequency identification antenna at the store exit will give an alarm linked to the exit cameras, enabling high-resolution pictures to be transmitted to security guards.
Social Media Monitoring Solutions
Social media monitoring for threatening keywords and gang language has become essential for both law enforcement and protection of soft targets.
Social media monitoring applications scan public social media messages and alert the retailer or law enforcement when specific keywords or phrases are used in a particular location. The application contains keywords in a “library of interest” and permits crime analysts, public safety professionals and retailers to add tailored keywords for specific events such as sales, concerts and sporting events. Taking it a step further, should one of the keywords or phrases be used in a social media account library of interest, the geofencing social media software sends an alert to designated security guards or law enforcement.
 IEEE Library: Fusion of Domain-Specific and Trainable Features for Gender Recognition From Face Images