Want to know what a car was doing prior to hitting a pedestrian? Tracking a vehicle when it enters a scene, coupled with identifying a jaywalking pedestrian could prove the pedestrian is at fault and not the vehicle. With older systems, law enforcement had to follow each object or person individually, but new security cameras launching in China with smarter chips are assisting law enforcement agencies in the analysis of vehicle and pedestrian traffic with real-time results.
Gone are the days of reviewing hours upon hours of back-to-back video surveillance footage from a traffic camera or waiting for a back-end server to complete analysis to discover critical instances, anomalies or activities such as traffic accidents, parking violations or robberies. Using the hardware and software built into a smart camera, law enforcement will now be able to track multiple objects simultaneously within the scene, avoiding having to review longer segments of captured video.
Using an Intel Atom E3845 processor embedded within a security camera for onboard analysis – a task previously reserved for high-end, back-end servers – unnecessary background footage is automatically ignored or removed while pedestrians, vehicles and bicycles are extracted and categorized in real time.
In just six months, Intel engineers were able to create a reference design, complete with hardware and software, to remove server-side processing, and provide a self-contained camera that is both high-performing and consumes relatively little power. With video recording capabilities of 30 frames per seconds in 1080p resolution, a new type of video analytics is now performed by the camera.
Kedacom, a video conference and network surveillance system manufacturer in China, is the first corporation to adopt Intel’s reference design in its smart IP camera solution, which provides car, people and object differentiation. An Internet protocol (IP) camera is a digital video camera attached to a computer network or the Internet allowing it to send and receive data, often used for surveillance.
With the hardware and software combination, hour-long videos can be “compressed” into a much shorter length of time, sometimes merely a few minutes, as well as into a smaller file. This is accomplished with the camera itself actually analyzing the video, identifying and isolating various objects in the scene, and then overlaying these identified objects in a much shorter video. The resulting video, which is delivered to the back-end servers, shows a static background with the objects overlaid in motion.
According to Mike Wong, business development manager within Intel’s Internet of Things Group in China, previous IP cameras only encoded and compressed video that was filmed; this video was then sent to a centralized server for decoding and analytics. These back-end servers were powered by Intel Core or Intel Xeon processors with network video recorder (NVR) software and transcoding software used to convert different types of video formats including AVI, Motion JPEG, MPEG2/DVD, MPEG/VCD, MPEG4, H.264, H.265 and others.
“By enabling in the camera itself, we conduct metadata extraction and/or object/people/car differentiation in real time in the front end, without the need to transmit the entire video stream backward., just need to transmit the ‘useful’ data/results on demand,” said Wong. “For comparison, previously we needed a Core or even Xeon CPU [on the network video recorder server] or on the backend to decode the compressed and transmitted video feed, then conduct the algorithm, metadata extraction, sorting, filtering, etc.”
The global market for video surveillance equipment is expected to grow by more than 12 percent by the end of this year and revenue is expected to rise to $15.0 billion, up from $14.1 billion in 2013, according to HIS Research’s “Trends for 2014 – Video Surveillance Trends for the Year Ahead” whitepaper. Driving these numbers are a variety of new technologies, analytics, sensors and vendors.
Heng Juem Han, platform architect from Intel China’s Internet of Things group, envisions the smart IP camera technology will be used within shopping malls for crowd monitoring, traffic counting and business intelligence with the back-end servers being freed up to perform data collection and data mining. The technology could eventually become incorporated into consumer video monitoring products, moving beyond the traditional “hot spot” and “zone watching” they currently do.
This content was originally published on the Intel Free Press website.