People Counting is a must for just about any type of business that runs a physical location. When done right, estimating the flow of people can provide insights that can lead to more streamlined operations and smarter business decisions.
In one of the previous projects, my team and I designed and implemented a People Counting application structured as a "Software as a Service" (SaaS) solution operating a network of Smart 3D Sensors (combination of miniature 3D Camera and Edge AI devices). SaaS is one of the main categories of Cloud computing and is characterised by a software distribution model in which a third-party provider hosts the applications and makes them available to customers over the Internet.
In our application, we developed a highly scalable architecture that consisted of (1) the network of Edge devices connected to the Internet; (2) the back-end of a Cloud-based web application responsible for collecting the data from each Edge device as well as the modules for data storage and analysis for people traffic computation and aggregation; (3) the front-end responsible for statistics visualisation and remote 3D camera calibration.
Our approach for implementation of 3D Sensing included the usage of a miniature 3D camera (Intel RealSense D415) connected by a USB cable to the Edge device. The system proved to be robust enough to detect fast movements and to provide the correct number of people even in situation when they were walking shoulder to shoulder (camera was mounted on the ceiling and as such it provided a very good vantage point).
Detection results could be visualized remotely via a web browser (a sample of the input image, people detection and counting statistics are presented in the slideshow below). It can be seen that the application preserves the identity of the people (locations of detected people are displayed as different colored dots). The blue rectangle indicates the analysis area that will be used by the system to monitor the entrance in the building (counting the entrances / exits).
By 2025 it is estimated that the rise of Edge computing will be responsible for 75% of enterprise-generated data. The move to the Edge will be driven not only by the vast increase in data, but also by the need for higher fidelity analysis, lower latency requirements, security issues and huge cost advantages.
However, as demonstrated by our use case, it is very likely that Edge computing will complement Cloud computing instead of replacing it. As it can be imagined, SaaS can only increase the versatility of an application operating a network of Edge devices capable of running various types of analysis that could be updated and deployed in matter of seconds.
Comments