Evolution of Video Surveillance
Is analog video still relevant in casino surveillance? Yes, though it’s declining rapidly as each camera is replaced with an IP camera.
The use of IP-over-Coax devices has become widespread as costs have come down and, in many cases, running new cable is cost-prohibitive. Many jurisdictions have changed regulations to require the removal of analog devices with compliance deadlines. Most video management software companies have provided migration paths from analog to IP to make the transition less painful.
Generally, all storage is built out at 1080P or higher to accommodate this transition. Douglas also tells us that today’s encoding devices for analog to IP recording of analog cameras are far superior to those of the past.
What is the average resolution found in casinos? Today, 1080P with a 6-22mm lens from an 18-20-foot ceiling over table games continues to get the job done. However, 4- or 5-megapixel cameras provide better card identification and make player actions easier to monitor. Below is a comparison of 5MP and 8MP on a poker table.
Combined with greater clarity in imaging, there is also the consideration that the most common codec used in gaming today is H.264, although some are starting to leverage H.265.
The bitrate comparison for a 1080P camera in a casino would be 6 megabits per second with H.264 and roughly 4.5 megabits per second with H.265. The question might be asked, “Why wouldn’t everyone utilize H.265 for the storage and bandwidth savings?” H.265 uses more resources for decoding the video. This requires more powerful work stations and graphics cards to handle the typical workload.
This allows operators to display fewer cameras on each monitor. For cage windows, 4MP has become popular, as it enables the operator to view the entire window while simultaneously zooming in on specific areas, such as tickets and cash.
While fisheye cameras have been around for several years, more recently 12MP cameras capable of 30 frames per second were released. This has enabled fisheye cameras to be gaming-compliant, whereas previously they were used only for forensic searches due to frame-rate limitations.
These cameras are typically scattered throughout the gaming floor, making it easier for operators to track individuals. The digital dewarp feature on these cameras provides a digital PTZ experience that can be performed in live and playback modes and even exported to video. We expect AV1 to enter daily use in the coming years, as it was designed for higher-quality, lower-bandwidth streaming at 4K and 8K resolutions.
What technology advancements are making the most significant impact in casinos? Artificial intelligence and deep learning are the fastest-growing solutions. The most common uses of AI are facial recognition (FR) and license plate recognition (LPR). This alerts casinos to the presence of banned patrons, self-excluded individuals, advantage players, high rollers and other relevant parties.
FR and LPR go hand in hand, as this allows surveillance and security to be notified before the person exits their vehicle by reading the plate upon entry to the parking garage. The fallback would be when they enter the casino and their face is recognized, allowing the operators to know their location and react accordingly.
Object recognition has streamlined the search process, enabling an operator to scan billions of video frames in seconds to find specific results. Some of the more popular use cases are weapons detection, smoke and fire, clothing color (top and bottom), age, gender, hat, mask, bags/backpacks, cellphones, vehicle attributes such as type (Truck, SUV, Van, Bus, Bike, Motorcycle, etc.), color, and even vehicle make and model.
Searching for people based on appearance has also become more effective and popular, as it allows the operator to review multiple servers and cameras to find individuals matching the specified appearance. Currently, the best-performing AI solutions are server-based GPU appliances.
The requirement to infer many frames from HD+ streams results in higher costs and scalability challenges.
The typical casino will use this to mark choke points and critical areas, then use it as a starting point to backtrack through other cameras. Camera manufacturers are working hard to improve edge-based AI by upgrading camera chipsets.
There has been significant improvement over the years, and LPR at the edge is almost as good as server-based AI and easier to scale. Most camera manufacturers offer object recognition that provides most of the classifications listed previously. The most significant benefits of server-based AI beyond accuracy are the data collection and reporting it enables, which opens a whole new world of valuable metrics.
This has allowed surveillance to leverage other budgets, such as marketing, to acquire AI for their properties. During the Covid-19 pandemic, counting people and detecting masks became almost essential. We all hope it never comes back to that; however, the information gathered during that time has improved casino operations by providing a better understanding of these metrics.
—by Max Davis
Max Davis is vice president of business development at Mirasys. With over 20 years in the video surveillance space, Davis started as an installer, then sales, design, and general manager of an integration company, and started the casino vertical in the U.S. in 2014, which has led to over 100,000 cameras being managed by the Mirasys VMS in casinos and gaming with key corporate accounts such as Caesars and MGM.
