Video detection analytics has matured tremendously over the last few years. Two decades ago you could activate a very crude motion detection analytic, with the main goal of extending storage capacity. These analytics were ‘reasonable’ at best, and not reliable for accurate detection, especially outdoors.
With more computational power, it became possible to use video analytics for real time detection and alerts, using more advanced algorithms. These algorithms were able to model what it saw as the ‘background’, and to distinguish moving objects from the known background, tracking their path, and determining very roughly whether the objects were persons or vehicles. But complicated environments (trees, bushes, car lights, spiders) still yielded significant false alarms.
With advancements in machine learning algorithms, detection accuracy was further enhanced, to the point where analytics could be deployed with success in most indoor and outdoor environments - as long as alarms were verified by a human operator (locally or remotely).
And with the drive towards continuous improvement, we are now at the point where reputable manufacturers/developers are using artificial intelligence and/or deep learning technologies, to reduce false alarms even further. Computational power is getting even better, and when using GPU’s for higher processing capabilities, it is expected that in a few years’ time, these detection analytics will equal (or surpass) the level of detection accuracy of the human eye.
We cover various aspects of video monitoring as a service, with specific focus on outdoor video detection and monitoring, in a 9-part series. Watch out for the next post!
For more information or demonstration of the system, please Book A free Demo.