by Stephen Callender, Vice President of Product Management at CrowdVision
It’s unlikely that you’ll find another team who like to nerd out over sensor technology as much as we do. There are always new advancements to discuss. Our customers who run large facilities and smart city initiatives want their crowd analytics systems to be futuristic and effective, and we love that, too.
Practically speaking though, there are more important topics to cover first. You have operational needs that aren’t being met, or you have a list of insights that could have a positive impact on the way you manage crowding. Let’s talk about that.
The reason why sensor tech is not where we want to start is because a one-size-fits-all sensor doesn’t exist. Designing sensor systems for crowd analytics solutions requires an understanding of where each one fits in the landscape of sensor technologies and, most importantly, a willingness to learn about them and the boldness to use them.
Sensor Agnostic Approach
Since our next conversation will be a little different, I’ll tell you why this is the best approach and why this method would improve the crowd analytics industry.
In my role, roughly half of my time involves talking to existing and potential customers. Each one has a problem to solve, but at the core they are often the same: If we understand where crowds are in our facility, then we can give them better information and we can plan our operations better.
In current times, there’s one more to add to that: We need data to keep our people safe.
We also talk about a number of constraints. Sometimes customers ask for such detailed information about guests in their facility that it would violate just about every privacy regulation that has been published – great for the operator, bad for the public. It’s a non-starter. More commonly, there are budgetary constraints or architectural and design constraints. Other times a decision maker in the company simply doesn’t like or trust a specific type of sensor, which is also a fair and valid constraint.
The variety of requirements and constraints that we come across on a daily basis are what frame our sensor agnostic approach.
Sensor Agnostic System Design
Recently, our team was asked to design an analytics system that measures the activity of pedestrians and vehicles along the front of a very busy, very large facility. We received a wish list of a dozen different analytics that involved analytics about crowding, vehicle activity and the interactions between people and vehicles.
Because of the large scale, the number of sensors that needed to be deployed and the networking and computing infrastructure to support it had to be considered. There were also many different types of objects and activities that needed to be identified. This is where the sensor agnostic design really shines.
The Design Process
First, look at the physical space and carve it up into discrete parts, then list the required analytics by each. Vehicles sizes needed to be measured, and the customer wanted to know which lane each one was in. LiDAR technology, which is today famously used to give sight to self-driving cars and help them see other vehicles and pedestrians, was a perfect fit for this. We chose a model with a 360° horizontal and 95° vertical field of view model with high resolution, so a minimal number of units could capture all the desired vehicle activities, with the added benefit of viewing people as they load and unload from each one.
We added license plate recognition cameras to the design – not to record and store license plate numbers, but to assign an anonymous unique ID when a vehicle passes by. This helped us understand how many unique vehicles visited.
Finally, we recommended leveraging the existing CCTV camera system with object recognition software to tell us the difference between a taxi and a personal vehicle. This allows the customer to count their customers as they enter and exit the facility, as well as maximize their CCTV investment.
It will take fewer than 25 sensors to deliver all of the analytics on the customer’s wishlist, while reducing the construction cost, time and materials.
For our colleagues in crowd analytics, enabling the sensor agnostic method takes a considerable amount of engineering and development. The zone analytics software that is open to ingesting data from multiple sensor technologies has taken years to engineer. However, it has afforded us the flexibility to find new and creative ways to say yes to our customers.
Testing and Adopting New Sensor Technologies
Sensor makers often contact us about their products, which has caused us to develop a testing and adoption process. There are the qualities to look for:
- Accuracy and capture rate
- Installation requirements
- Production capabilities (meeting demand)
- Whether it produces a new type of data that customers are looking for, or it simply replaces a sensor we already use by being better or less expensive
- Cost per unit
Our relationships with sensor makers have been positive. We work closely with some of them and do our part to provide feedback to their engineering teams about how they can improve upon the above categories for the crowd analytics industry.
Sensor makers can also focus more time on research and development while we focus on applying their technology in the field.
With so many great options to use today, sensor agnostic crowd analytics systems foster more creativity and flexibility in solving problems for customers. It’s the best way for facility operators, smart city technologists and those operating in other crowded spaces to select a single provider for rolling out any size of system.
I’m looking forward to the next time we talk, when we can dig into the problems you’re facing then follow the sensor agnostic system design approach to find the best solutions.
Footnote: Callender has worked in the crowd analytics space since 2017, having previously focused on design and entrepreneurship in the software industry.