Computer Vision News - May 2021

335 Active-Vision by SwRI reaction so that agencies know immediately when something is wrong and can take action. Dedicated sensing hardware does exist that could do a similar job, but it is cost-prohibitive both to purchase and particularly to maintain. The Active-Vision system is cost-effective and needs little to no configuration. The team just finished an 18-month internal research project building out its capabilities to detect vehicle- based behaviour, including wrong-way driver detection, speed detection, and tracking GPS coordinates in real-time for vehicles within the field of view. The system integrates with existing ActiveITS software to facilitate getting messages out to agencies as quickly as possible. For example, an event can be automatically generated that says there is a wrong-way driver at a specific location. At that point, the travelling public can be notified via digital signs on the highways, and roadside units can be alerted. This can all be done autonomously in real-time, which will be a gamechanger in terms of saving lives. “We’re really happy with where it’s at and we’re looking forward to bringing that to the field and getting agencies using it,” Dan says enthusiastically. The system uses well-established computer vision algorithms primarily focused on detection and tracking. “It’s a straightforward application of computer vision, object detection, tracking, camera geometry, and global data,” David explains. “ We’re using a single-shot detection algorithm. We have a couple of different variants of that. You can imagine that if you’re getting object detection and object tracking, a lot of the other analytics fall out of that, such as the position and velocity estimation.” “It’s a straightforward application of computer vision, object detection, tracking, camera geometry, and global data.” David Chambers