When discussing IoT, it is often said that every deployment starts with a smart, connected device. These devices contain sensors that are generating data, which is the basic element needed to deliver on business-oriented use cases. There are many types of sensors available, which can detect conditions such as temperature, proximity, acceleration, force and flow. Additionally, control systems can report when a button is pushed, the position of dial or the value of a meter.
But, what about situations where these sensors don’t exist? While it has become easier and less expensive to retrofit existing equipment, it is not always a trivial task. This is an area where the use of optical sensors and image recognition can fill the gap.
First, let’s look at vending machines. Despite great strides being made in developing the next generation of smart, connected vending experiences, there is a still a large fleet of traditional machines in service. Retailers are not ready to retire this equipment, but want to take advantage of the benefits that IoT can offer, including real-time sales reporting and inventory management. Instead of placing sensors on every shelf of the cabinet, a single camera can be added to capture an image of the product as it is vended. Data analytics applied to these images provides product identification by determining the unique characteristics of each package. This not only helps increase accuracy in the event that a product is stocked incorrectly, but can also be used to alert the operator to unauthorized products in the machine. That’s something a simple sensor cannot do on its own.
Image analysis can also help industrial environments, which have a significant installed base of analog gauges and valves. These companies typically have humans taking manual readings, often on paper logs which may never get keyed into an online system for further use. The lack of visibility into operations is compelling these businesses to find a better way of enabling their equipment to participate in the digital world. While some have already deployed cameras to remotely view gauges, it is still a manual effort. It is now possible to automate the processing of these images to determine gauge or valve positions. This dramatically increases the frequency of data collection and enables operations managers to better understand the performance of their equipment.
Another field that can benefit from this approach is parking management. Think about the desire to understand the inventory of spots and the current occupancy rate of a parking lot. Today, these companies are either manually walking the aisles to keep count or just tracking vehicles at the gate. This method is often inaccurate and doesn’t provide a way to guide someone to the nearest available spot. By installing a camera with a wide angle lens, image analysis can tell the difference between an occupied and unoccupied spot and expose this information to both lot managers and patrons.
One factor that has inhibited the use of cameras for these types of IoT deployments has been the cost of both transmitting and storing high definition images in the cloud. However, advances in edge analytics have led to the ability to provision image pattern recognition directly onto gateways and sensors in the field. This dramatically reduces communications costs and improves response times. Only the results of the analysis need to be sent to the cloud so that it can be consumed by dashboards and mobile applications.
As IoT continues to bridge the gap between digital and physical worlds, I encourage you to be creative in your solution architecture, recognizing that more sensors are not necessarily better. Making smart choices and leveraging distributed processing of data will put you on the path to IoT success.
Edited by
Ken Briodagh
