A 3D Sensor System that Mimics Human Depth Perception
Many applications in the automation, robotics and quality assurance industries require detailed 3D measurements of a particular object or area. Today, active camera systems are commonly used for such tasks, which measure the depth – and thus the 3D structure of a scene – by emitting light and capturing its reflection. Due to this active illumination, such systems have difficulties coping with bright ambient light, overlapping field of views, large distances, or the presence of reflective surfaces. Also, light-scattering environmental conditions such as fog, steam or smoke can cause disruptions.
With the SP1 stereo vision system, Nerian Vision Technolgoies introduces a new innovative sensor solution, which works passively and thus does not face the same problems. The SP1 uses two standard industrial cameras, which are mounted at slightly different viewing positions. Similar to the human visual system, the SP1 is able to infer the depth of the observed scene by comparing the images from both cameras. This enables the SP1 to provide a depth estimate for each observed camera pixel.
The SP1 gains its performance from a hardware implementation of all image processing operations. This has been made possible by using a programmable logic device – a so-called FPGA – which allows to map of all image processing algorithms into a dedicated logic circuit. This hardware-based design lends itself to massively parallel processing, which leads to a vast performance gain when compared to software-based solutions. This enables the SP1 to process 9 million pixels per second, while at the same time maintaining a small form factor and low power consumption.
When designing the SP1, Nerian deliberately opted for using standard industrial cameras. This allows the user to chose the image sensor, optics and also the camera placement. Each application has different measurement requirements, which calls for a different camera setup. For example, when observing distant objects, it is recommended to use lenses with long focal lengths and to mount the cameras far apart form each other. This setup will lead to a high depth resolution for distant objects, but at the same time it will inhibit the observation of close objects. For making close measurements, one would chose lenses with a short focal length and a narrow camera mounting instead.
Because the user maintains full control over the camera setup, he is able to adjust it according to his needs. Should the measurement requirements change, then it is sufficient to just change the lenses or the camera mounting. After a short manual calibration, the system will again be fully functional. It will then also monitor its own calibration and perform an automatic re-calibration if there are any variations in the camera alignment. This adaptivity allows the SP1 to be used even for applications with unusual or varying measurement requirements.