The revolution of artificial intelligence, the increase of processing power and the reduced size of components have taken edge computing to the unprecedented level, it is now. Instead of a bulky installations consisting of a wide variety of devices, all you need now is an embedded, smart camera system and a modem or router for power and ethernet on site. Edge computing offers the same and more functionality, but being compact and far more easy to install and maintain. AVUTEC has risen to this challenge to move from edge computing towards edge integration by moving more and more functionality to the edge computing device. Let’s see what this means.
Why edge computing
“Edge computing is the technique whereby the analysis and processing of data produced by a device is done by the device itself or by a device close to it.”
Although the advantages of edge computing are becoming more known, a short recap can do no harm. Edge computing is sustainable while it consumes less power and needs far less network bandwidth. It is the perfect solution for outdoor areas with low or no Ethernet coverage. In areas where coverage is dense, edge computing proves it worth during periods of Ethernet failure. And last but not least, as video analysis runs on the device, edge computing is faster and more accurate, which is essential in time and meta-data critical applications.
So far technical developments have made edge sensor devices increasingly deployable. With the implementation of edge integration, AVUTEC stretches the possibilities of edge computing by integrating as much functionality in a single device as nowadays is possible.
Taking the leap to edge integration
“Edge integration turns our camera sensor systems into standalone, multi-functional units, that process and analyse video on board, communicate the results and integrate seamlessly in any ecosystem”
A definition of the edge integration is a beginning, but a use-case will illustrate the concept far better. Let’s jump right into it. In a smart city in the Netherlands Gatekeeper-X sensors are installed to automate tax collection and to support crowd management. All boats that are taxable, have a unique license plate at their sides. The smart camera systems read the license plates and in parallel visually classifies boats into categories and tracks them. Meanwhile their ecosystem count boats on the sometimes crowded waterways, based on the real-time metadata information reported by the Gatekeeper-X sensors. Alarms and alerts are generated when boat count hits the threshold. Captured license plates and boat categories are uploaded to the city’s registration platform while authorized staff can check the stored and privacy proof images. People that are shown on the images, are blurred by the Gatekeeper-X for privacy regulations, even before they are transmitted.
For this project a custom classification network was trained and implemented. And the AVUTEC ANPR engine was adjusted to read the customised license plates on the boats. Throughout the training period, the internal storage capacity of the Gatekeeper-X camera system recorded all relevant images. It acted like an image collector to build the dataset used for the AI network training. After training, during operation, internal storage buffers images and recognition results in case of Internet connection failures.
This use-case clearly shows the capabilities which true edge integration has to offer. All features and functionality implemented in a smart camera system, with no further use of server capacity. Edge integration offers a camera sensor system that is designed to be widely applicable in a variety of visual AI projects, yet equipped with the hardware and software to be the flexible and future-proof camera it needs to be.