The means of 3D object sensing utilized in the PhoXi family of scanners, currently under development by Photoneo (Bratislava, Slovakia), is based on an advanced structured light technique. The PhoXi is able to capture high-resolution 3D data with a single image frame. This is accomplished by the use of so-called multi-image sensing.
First, a little background. Conventional stereo vision uses a pair of cameras to capture images of an object from two different perspectives. The sensor in the PhoXi camera is reported as a custom-design smart CMOS sensor technology from Sony. An algorithm is used to map a given point in the first image to a corresponding point in the second image. Triangulation is then used to compute the 3D position of the given point. It is found, however, that on anything other than a “well textured” object, establishing the desired correspondences can be difficult.
This issue is addressed in the structured light approach by replacing one of the cameras by a pattern projector.
Such a projector emits a well-defined structured illumination. In current systems, the source of the structured illumination is most often a conventional 2D projector. The camera and the pattern projector are located on a well-defined baseline and are focused towards the objects in the area to be scanned. The pattern projector sequentially projects a set of patterns that encode spatial information. The camera captures an image of the scanned area each time the pattern is projected. From the captured images, the algorithm decodes the spatial information. This, in turn, allows determination of the spatial information corresponding to the second image point. With the correspondence established, the algorithm computes an exact 3D position of the object point. Since the correspondence was computed from illumination produced by the system itself, the technique works independently of the object’s texture – thus solving one problem encountered in the prior art.
More than that, the 2D projector in the Photoneo projection system uses coherent laser radiation of a specific wavelength. This, in turn, enables the optical system of the PhoXi projector to be particularly effective in that it delivers essentially all emitted light to the scanning area. The resulting energy efficiency allows PhoXi to use just passive cooling. The high quality glass bandpass filters used in the camera’s optics is claimed to allow the system to deliver good resistance to extraneous light sources and provide an excellent depth of field.
All necessary computations are performed by the system’s embedded processing and control unit. The graphic processing unit is from Nvidia. PhoXi is connected to the user’s computer through GigE using a control application provided along with the PhoXi. The software API is reported as delivering features such as adaptive scanning procedures for complicated objects and the ability to automatic analyze the scanning area and determine the optimal scanning parameters. The API delivers 3D information alongside 2D texture information.
The PhoXi camera capture and processing pipeline is capable of delivering 16 million measurements per second, either at 3.2 megapixels at 5pfs or 0.8 megapixels at 20fps. The company reports that the frame rates can reach 60fps and that the cameras can deliver “a motion-blur free 3D reconstruction.”
The design of the system allows scanners to be configured for a variety of applications through the in-factory ability to change the baseline of the system. The body of the sensor is built from carbon fiber. The resulting rigid structure is claimed to ensure a long lasting calibration. The sensor mount is built from anodized aluminum making it easy to mount to a robotic arm or other moving system.
Photoneo has secured €2.1 million in seed investment. The investment was led by Prague-based venture capital firm Credo Ventures, with the participation of several co-investors. The newly raised capital will allow the company to finish development of the image sensor. Photoneo is working on smart SDK packages to make the development of 3D vision applications “easy even for non-programmers.” -Arthur Berman
Photoneo, Jan Zizka, +421-903-609-136
