Walking is both an accessible form of exercise and a necessary mode of transportation for many people. But it’s also a dangerous one, in 2024, the state of California saw more than 11,000 pedestrian accidents resulting in 764 deaths.
Advances in pedestrian recognition technology are at the forefront of making pedestrians safer in a variety of environments, from city streets to industrial sites to anywhere else that people might be walking. And as with many fields, AI is playing a growing role in pushing the boundaries of what pedestrian safety technology can accomplish.
The Basics of Pedestrian Detection Technology
Pedestrian detection technology refers to any system that automatically recognizes that a person is walking in the area of a vehicle’s path. The system then either alerts the driver or takes independent action to prevent a collision.
A pedestrian detection system relies on sensors to scan the surrounding area, using either cameras, radar, LiDAR (a laser-based scanning technology) or some combination of the three. As the data comes in, sophisticated algorithms assess and determine whether a person is in a dangerous proximity to the path of the vehicle.
If a pedestrian is detected, usually the system will send an alert to the driver, either via a console warning or a vocal warning. In some cases, the system may also automatically deploy the brakes if the vehicle gets too close to the detected pedestrian.
How Multi-Modal Sensors Enhance Accuracy
The original forms of pedestrian detection technology tended to use a single form of sensor, often a camera or a series of cameras. And while a camera’s ability to gather visual information is important, there are driving conditions that make camera data unreliable.
More sophisticated versions of pedestrian detection technology are using the multi-modal approach, incorporating other types of sensors to augment camera data and get a fuller picture of what’s happening.
Radar, the use of radio waves to calculate an object’s distance and speed from the sensor, is especially helpful in difficult weather conditions. And LiDAR, which deploys laser beams and is capable of creating fully articulated 3-dimensional maps, is able to tackle low lighting or variable lighting conditions where traditional cameras struggle. By bringing together the data that all these systems deliver, the algorithm can arrive at a more accurate understanding of whether a pedestrian might be in the vehicle’s path.
How Machine Learning and AI Are Upgrading Pedestrian Safety Technology
Any pedestrian detection system is only as good as its accuracy. In some ways, an unreliable pedestrian detection system is worse than none at all, as a driver might rely on the system so much that they aren’t as vigilant in looking for pedestrians on their own. It’s imperative that a pedestrian detection system correctly identify all pedestrians.
But real-world driving conditions present a host of challenges for any algorithm. Driving conditions can vary in the level and type of lighting, various weather conditions, crowded areas, other vehicles, and any objects that might appear near or on the road. Pedestrians may pass behind vehicles, objects or other pedestrians only to reappear abruptly.
Before the advent of AI and machine learning, there were significant roadblocks in front of the creation of a robust pedestrian detection system. Early systems had known issues with certain tricky circumstances and had questionable reliability in challenging conditions.
A machine learning model has the ability to process thousands upon thousands of data sets of every possible variety, filtering and optimizing the algorithm until accuracy reaches near-perfect levels. These models are gradually trained until they far surpass what a conventional algorithm can do.
But AI-based models have another key advantage over traditional pedestrian detection systems: Their ability to predict future behavior. By studying historical data and then comparing that to the real-time data coming in as the vehicle drives, an AI-based pedestrian detection system can predict when pedestrians are likely to enter the roadway or the vehicle’s path and alert the driver proactively.
Integrating Pedestrian Detection Technology Beyond Vehicles
While the most prominent use of pedestrian detection systems has been in vehicles, this technology is starting to spread into other areas as well. This is opening up new opportunities to enhance public safety in hazardous conditions.
One especially promising application is in the field of workplace safety, especially on job sites or in industrial settings. Heavy machinery can pose a significant danger to people walking in the vicinity, and pedestrian detection systems can prevent industrial accidents involving machinery like cranes, forklifts and similar devices.
Yet another development is the concept of smart city infrastructure. Here, instead of alerts coming from the vehicle to alert its driver, the pedestrian detection systems are embedded within traffic structures and other infrastructure.
These alerts can signal both to drivers and to pedestrians, giving either party the chance to avoid a potential collision. In some cases, these systems can even allow for communication between drivers and pedestrians, decreasing the possibility of an accident caused by misunderstanding.
Pedestrian detection technology is a fast-moving field these days, especially when it comes to how AI and machine learning can boost the accuracy and versatility of the algorithm. We’re also seeing a push to apply the technology beyond the traditional application of a vehicle-centered system that alerts a driver. As the technology becomes more sophisticated and widespread, we’ll see fewer and fewer preventable collisions involving pedestrians.