Market is flooded with depth sensor that runs stereo vision or time-of-flight (also known as ToF), two different approaches to retrieve height information to construct a depth model for counting. But just as Tesla moved to a full vision approach in 2021, computer vision is a serious contender in the mix.
Building a depth model and counting simultaneously is resource intensive for a device that draws less than 15-watts. Besides, ARM-based system-on-chip (also known as SoC) today are built with neuro engine dedicated to performing deep learning algorithm, further freeing the computation resource for wider application.
Advance surveillance cameras nowadays run AI on the edge (hyperlink to AI solution) and accurately detects people, but with enough dataset fed to the algorithm, recognition of wheelchair, bike, luggage, or strollers are in sight. Such object recognition capability is limited in many depth sensors but are valuable to provide business intelligence data for retailers and onboard transportation service providers.
One of the challenges in retail counting is the exclusion of staff counting. Similar to the approach for object recognition, the algorithm can be trained to recognize specific patterns on the employee shirt, therefore eliminate them in counting.
This approach can be further transferred to onboard transportation. Detect door status with a specific door mark.
VIVOTEK opts for fisheye design to leverage the wide field-of-view. With integration with the camera SDK, dewarp is available to provide surveillance monitoring, making it a dual application device.
This approach can be further transferred to onboard transportation. Detect door status with a specific door mark.
Kelly Huang
Senior Product Marketing Representative
Kelly Huang is a member of VIVOTEK product marketing team based in Taipei. He is responsible of translating technical knowledge to product messaging, and often the other way around. Prior to product marketing, he worked as a technical trainer, enabling global salesforce and customer base with comprehensible technical insight. Native to Taiwan, he grew up in Thailand and has a confusing unisex name that requires explanation in every self introduction.