Aimetis Leaps Forward with Efficient HD Video Analytics

fountain-1342188_1920.jpg

Video surveillance camera technology has consistently improved with HD and 4K cameras being utilized around the world to ensure the clearest footage is available to security teams.

In a recent IFSEC Global report, security professionals from a wide range of authority levels and industry verticals were surveyed regarding their video surveillance needs. When asked to identify the most important features and functions for operational needs, it was revealed that 74% of respondents saw Full HD (62%) or 4K (12%) resolution as one of the top five.

The same survey revealed that a majority of respondents (54%) see video analytics as one of the top features required for their video surveillance system. Both HD footage and video analytics are clearly of the utmost importance to those professionals working day in and day out with their video surveillance system. The problem is that the two features have not been practical to be paired together. Until now.

While recording high-resolution video is highly sought after for its increased level of detail, it comes with a high price in terms of system resource utilization. This tradeoff is magnified when analytics are added to the mix. Aimetis has been at the forefront of research and development of industry-leading video analytics that work in everyday, real-world situations.

We have now introduced Adaptive Analytic Resolution technology that is available in both server-based Outdoor People and Vehicle Tracking (OPVT) and camera embedded Aimetis Outdoor Object Tracking – (AOOT) video analytics. Adaptive Analytic Resolution technology represents a significant breakthrough, delivering accurate and highly efficient analytics with HD cameras.

Why High Resolution Video Analytics are Hard

Given the same camera with the same lens, the higher the resolution of the video stream, the further away you can see objects in the frame. Consider Example #1 below, which contrasts a subject, at the same location within the frame, at three different resolutions.

At 640x360, the resolution typically used for a secondary video analytics feed, you can barely see the person all the way at the back end of the parking lot. At HD resolution of 1280x720 (720p) you can start to see the person and at HD resolution of 1920x1080 (1080p) you can see the person in more detail.

These differences are more dramatically illustrated in the close-up views of the person in question that have been extracted from each frame.

Example #1.png

Example #1

The higher the resolution, the more pixels you have in the frame, triggering a corresponding increase in analytic computation. Herein lies the crux of the trade-off that limits traditional HD video analytic processing. In fact, a 720p frame has four times as many pixels as a 640x360 frame and a 1080p frame has nine times as many pixels as a 640x360 frame (see Example #2).

A typical video analytic algorithm will need to process every single pixel in a frame, giving CPU load a generally linear relationship with video resolution. This means a 1080p frame will consume nine times (and 720p frame four times) the compute resources of a 640x360 frame.

 Example-#2_with-resolution-totals.jpg

Example #2

What does this mean to you? Let’s say you have a server with enough CPU resources to run video analytics at 640x360 resolution on 36 cameras. With the same server, you can run video analytics at 720p resolution on only 9 cameras and at 1080p resolution on only 4 cameras. That means the higher the video resolution, the farther away you can see, but at a significant cost in terms of the number of cameras that you can run on your server.

What is Adaptive Analytic Resolution?

To use HD resolutions to detect objects farther away, but also keep computation costs low, Aimetis developed the new Adaptive Analytic Resolution technology for the OPVT and AOOT video analytics.

Adaptive Analytic Resolution technology is built on the idea that HD resolution is only needed to detect objects in areas at a distance from the camera, and for areas closer to the camera, we can still track objects at a lower resolution (Example #3).

Example #3.png

Example #3

Adaptive Analytic Resolution technology can represent the 720p frame illustrated in Example #3 using only 158,216 pixels (83% fewer pixels than the original 720p frame and 31% fewer than a 640x360 frame) and still preserve enough information to detect and track objects both at a distance and in close proximity to the camera.

What Makes Adaptive Analytic Resolution Unique

Using Adaptive Analytic Resolution technology, Aimetis analytics (i.e. OPVT and AOOT) can detect and track distant objects in HD, but at a fraction of the CPU cost typically required to run video analytics at HD resolutions.

The intelligent Adaptive Analytic Resolution algorithm automatically and efficiently downscales resolution where HD is not required, while accurately detecting and tracking objects as they pass through the high definition scene. It also realizes additional efficiencies by leveraging the primary HD recording feed, without the need for a secondary analytics stream.  

Recall the earlier example of the server with enough CPU resources to run 36 cameras at a video resolution of 640x360. In that example, moving to 720p HD video resolution reduced the number of cameras that can be run on the server from 36 to 9.

However, if you were to use Adaptive Analytics Resolution technology at 720p, you can run 24 cameras on the same server (that is 15 more HD cameras, or 2.7 times the number of cameras compared to using conventional, non-adaptive 720p resolution). At the same time, the Adaptive Analytic Resolution technology at 720p is as effective as a conventional 720p video stream in tracking and classifying objects both far from and close to the camera.

Example #4

As you can see in the Example #4 video, the person walking in the back of the scene, being filmed in 720p Adaptive Resolution, is effectively tracked and classified by the video analytic utilizing only 1.5x the CPU load of a lower resolution 640x360 stream.

With Adaptive Analytic Resolution technology, you can have your cake and eat it too. You can have the precision of Full HD surveillance cameras and the power of Aimetis’ OPVT and AOOT video analytics without breaking your server hardware budget. Never miss a single moment of clarity with Aimetis’ introduction of Adaptive Analytic Resolution technology.

improve video analytics