Municipalities large and small use video for many applications, and most especially for public safety. We often hear stories of how archived video has led to successful location and prosecution of criminals—after the crime was committed. But the goal of security managers and first responders is to counter crimes more quickly, even before they happen, to improve safety and better serve public needs.
Real-time video analytics have been a longtime problem in computer vision applications. The task is quite challenging, with video being an information-intensive media. Systems can be complex, and environmental inconsistencies make video analysis tricky.
With the advent of artificial intelligence, and now deep learning, the opportunity for computer vision applications simply continues to expand. Deep learning technology is enabling a new generation of video analytics, one where people, objects, vehicles, and more can be identified frame by frame—and their movement tracked across frames and cameras.
Giving Security Teams the Tools They Need
It's not uncommon for one small security team needing to simultaneously track multiple activities and locations: cars as they enter a parking garage, hundreds or thousands of visitors as they come into public areas, unauthorized restricted-area breaches. All these events are recorded by potentially hundreds of cameras, if not more. And the analysis of nonstop video streams has to happen in real time, if immediate actions are to be taken.
“People still need to use human resources to identify what’s going on. That doesn’t make sense. Especially if you have one security guy who is responsible for 100 video cameras,” said Felix Song, VP of Video IoT Solution at Gorilla Technology. “We have an edge when it comes to smart analytics. For example, we can identify one person from 5,000 visitors within two seconds.”
But deep learning comes with a cost. “Efficiency is pretty important here, because at the end of the day, the hardware required for deep learning's high computing demand can get really costly,” remarked Dr. Spincer Koh, CEO at Gorilla Technology. “So, CPU usage efficiency means better results at lower hardware costs.”
This is where the muscle of deep learning and video processing at the edge comes in. AI and deep learning technology “piggyback” on big data, and with better algorithms a new generation of video analytics is created. When processing of intensive media is happening all the way from the edge to the cloud, the result is increased accuracy with optimized video processing.
The key to edge analytics is computing power. In partnership with Intel®, Gorilla Technology implemented the OpenVINO™ Toolkit, enabling the company to make significant enhancements to what was already a powerful video solution.
With OpenVINO the company's IVAR (Intelligent Video Analytics Recorder) system, video analytics performance increased by a remarkable 50%. This enabled Gorilla edge devices to handle 1.5X as many video feeds with real-time analytics. And the results are lower deployment costs. More video channels can be analyzed on the same hardware framework, enabling superior response time, operational performance, and asset value, as shown in Figure 1.
Safer Transportation and Better Service
The Gorilla IVAR solution is well suited for public transportation. The Taiwan Railways Administration oversees the country's rail network, operating 300 train stations nationwide. The rail authority's goal was to provide better commuter services while improving station security.
It deployed the first Intel® OpenVINO™ based IVAR system at a busy transit station with more than 17,000 daily travelers and a limited security staff. It now can recognize people on watch lists, monitor footfall traffic, analyze abnormal behavior, detect unlawful intrusions, and more. With IoT sensors built into the IVAR edge systems, authorities can detect fires and intrusions on the tracks or other restricted areas.
Alongside expanding station safety, Gorilla provides better customer experiences. The ticket office can allocate staff as needed—such as in areas where there are long wait times to purchase tickets. Stations can apply platform entry policies if too many people are crowding a platform space, and provide more accurate schedule information.
The administration benefits from significantly lower deployment costs and the public benefits from better customer services. A win-win all the way around.
As Gorilla IVAR provides real-time alerts and smarter services, the railway administration has experienced a 90% decrease in response time. Plus, incidents and complaint rates have dropped by 70%. “Overall crime rates have dropped up to 80% while clearance rates have increased by 50%,” said Dr. Koh.
An Open System
Gorilla is making video surveillance a more cost-effective security tool through an open platform. It runs on standard Intel® processor-based computers, which gives customers a broad range of hardware options. The system supports open-standard codecs and protocols to work with any kind of IP camera as long as they are ONVIF Profile S and RTPS streaming.
Many surveillance systems are cumbersome—with multi-vendor and multi-generation video systems. This makes video and data monitoring a big challenge. IVAR can be easily integrated into existing surveillance systems, enabling numerous cameras in different locations or sites to connect to a central cloud server as shown in Figure 2.
Gorilla's IVAR solution is a software-based comprehensive video surveillance system designed for CPU efficiency. It enables efficient monitoring across areas and activities to immediately detect emergencies and threats. With the creation of watch lists, for example, detection of unusual activities and repeat offenders reduces manual effort and human error. Another feature—intrusion detection—quickly identifies suspects and uses push notifications to alert personnel if an event occurs.
Remote administration allows operators to keep track of cameras and manage multiple feeds simultaneously. Security personnel can efficiently search, view, and monitor premises to increase situational awareness and reduce time to action. By integrating existing surveillance systems, all essential timeline and event data is displayed. And a single dashboard—providing centralized monitoring—brings it all together.
The IVAR solution takes a big step beyond traditional Network Video Recorder (NVR) systems, which typically provide storage, streaming, and playback on a proprietary platform. This limited flexibility makes it a huge challenge for these systems to keep up with rapidly evolving, high-performance technology.
“We can do a lot more today, with a lot less computing power. That’s definitely because of OpenVINO,” said Dr. Koh.
About the AuthorMore Content by Georganne Benesch