【ProAV Lab】Breakthrough - Automatic Tracking Camera: Detecting Individuals

By By Kieron Seth & Seaco Wei, Lumens

August 17, 2022 17192


▶ Introduction


Auto-tracking cameras follow presenters as they move across a classroom, studio, or stage. This eliminates the need for an operator during recording. The key to automatic tracking is subject identification. The person must be clearly identified against any background in order to track flawlessly. Lumens Tracking cameras are equipped with the latest AI technology to work quickly, accurately, and reliably.


▶ Traditional People Identification Methods

Traditionally, tracking cameras have used basic image processing methods to match shapes in the frame with the typical contours of a human face and skin color. Using this method, the detection rate has usually been around 70%. However, in challenging situations, where the camera is shooting at an angle or where the background is not uniform, cameras can often fail to lock onto the presenter or lose tracking during filming.


Fixed Pattern detection


▶ Lumens' new generation AI detection algorithm

Lumens has adopted a new generation of AI intelligent detection technology and uses an NPU (Neural-network Processing Unit) and Machine-Learning to deliver exceptional human identification. It has experienced more than six months of Deep-Learning and tuning. Training has involved the analysis of millions of faces and human-shapes in a variety of shooting angles. The result is considerably greater accuracy, especially with people of different ages and skin tones, and it has the ability to account for any shooting angle and any kind of background.

Artificial Intelligence Deep Learning


Click here Back to Top ↑