Similar to the workings of the human eye, ceiling-mounted Xovis 3D Sensors capture the scene below using stereo vision. The two wide-angle lenses perceive the scenery from different perspectives (left image, right image). The sensor calculates this optical disparity for every pixel of the image. This results in a precise depth map or 3D image of the whole scene. It is calculated up to 30 times per second in real-time. Subsequent processing features deep learning-based person recognition and tracking, allowing to analyze individual paths of each person.
The 3D stereo vision technology is highly robust and resistant to all kinds of external influence, such as shadow, light changes, or heat emission. Xovis 3D Sensors feature a counting accuracy of more than 99% far exceeding the performance of conventional people counting and tracking technologies.
Xovis 3D Sensors, particularly in combination with an AI algorithm, enable customers to handle the most challenging people counting and tracking tasks, such as occlusion and the individual detection of people in heavily cluttered situations.
Xovis co-founder and CPO Christian Studer explains why Xovis’ focus will stay on 3D stereo vision:
“You see devices featuring 3D technologies such as TOF and Lidar, but the future lies with vision sensors that feature AI- capabilities based on Deep Learning. The market demand for detection of the exact behavior of shoppers in brick-and-mortar stores can be targeted best with the new generation of 3D smart sensors.”