FlexView

Our FlexView projection system available in the N35 and X36 cameras.
FlexView improves the camera’s voxel resolution by acquiring multiple image pairs with different projected patterns.

Working Principle

With FlexView the pattern inside the projection module can be shifted mechanically by a few micrometers,
which results in a visible pattern movement of a few pixels in the camera images.
Instead of capturing only a single image pair we can then capture multiple image pairs with different
(i.e. shifted) patterns on the scene. Our stereo matching algorithm can combine the information from all
image pairs to generate much more detailed disparity maps and point clouds.

FlexView 1

The original FlexView technique used in the N35 cameras uses only 3 to 8 images which roughly doubles the X, Y and Z resolutions while only sacrificing the additional capture time and very little additional computation time.

FlexView 2

The improved FlexView 2 technique uses a slighly modified projection pattern and is available in the X36 projection modules. In addition to pure random dots the pattern also contains unstructured grey value stripes in regular intervals. The stripes are then used for subpixel interpolation of the stereo correspondences, similar to the sine wave patterns of fringe projection systems. With this method FlexView2 again doubles X, Y and Z resolutions compared to FlexView1 using 8 to 16 image pairs. The newly introduced sequence correlation stereo matching method also greatly reduces the computation time compared to FlexView1.

Applications

In principle you can use FlexView whenever the part you want to measure remains static for the time of the image acquisitions. Using multiple image pairs to reconstruct the scene will also increase the reconstruction robustness on difficult surfaces because the shifted pattern generates additional information on specular, dark or volume scattering surfaces. Many algorithms will profit from an increased point cloud resolution and lower noise. FlexView thus reduces the need for post processing steps and decreases the run time of the entire 3D processing chain.

The integration of the FlexView functionality is a simple as setting the number of image pairs to be used.
Everything else is handled automatically inside our EnsensoSDK.
This allows you to enable the FlexView functionality when switching from N10, N20, N30 or X30 cameras
without modifying existing software by simply updating the camera configuration file.

In addition to the FlexView operation mode all N35 cameras can be operated in single-shot mode
by simply disabling the pattern shifting by software. This again gives you the ability to capture fast moving objects at highest possible frame rates.

X36 projection modules can also run in single-shot mode, but as the X36 pattern is optimized for multi-image acquisition it will deliver slighly less detailed 3D data in single-shot operation. If you intend to use X-series cameras for single-shot measurements only it is therefore recommended to actually stick with X30 models with a pure random pattern.

Mechanics

Mechanically shifting the projection pattern in our case does not involve actual ‘moving parts’. Our mechanics only consists of a piezo actuator pushing the pattern in one direction; without any joints or bearings involved. The pattern holder is made of a special high performance synthetic material and is only bent in a small portion of the material’s elastic range. Thus the pattern shifting basically induces no mechanical wear and the mechanics will operate for millions of shifting cycles without performance degradation.

FlexView also does not make any robustness tradeoffs: all FlexView cameras and projection modules meet the very same vibration and mechanical shock criteria as all other Ensenso cameras.