DLP 3D Machine Vision

DLP^® technology enables 3D Machine Vision capabilities by providing single or multiple camera 3D image capture, utilizing the same camera(s) employed for other Machine Vision functionality. 3D Machine Vision is a crucial component of advanced autonomous and semi-autonomous robotic systems used in the manufacturing, security, medical, environmental, and scientific domains. 3D Machine Vision, combined with advanced motor controls, capable mechanical systems, and comprehensive software, enables precise, adaptive, and “intelligent” robotic systems which are aware of their environment and/or target. The ability to recognize target configurations and precisely maneuver in 3-space (perhaps with detached mobility) provides valuable capabilities beyond traditional mechanical systems.

DLP technology facilitates 3D Machine Vision by utilizing a DMD (Digital Micromirror Device) as a spatial light modulator.  The DMD achieves high quality, flexible, and speedy sequential pattern illumination of the subject field.  Typically, solid state (LED) illumination is used to provide monochrome or multi-color, high brightness illumination within the visible to NIR wavelength range.

A synchronized camera of sufficient resolution, sensitivity, and capture speed is required to complete the 3D sensing loop.  The resolution of the camera/camera lens should be commensurate with the DMD resolution for best measurement quality in x, y, and z (depth) dimensions. Both the projected patterns and the camera imaging field should match very closely. This overlapping area is where the 3D vision field occurs.

The 3D measurement depends on the principle of geometric triangulation. This requires a certain amount of baseline offset between the pattern projection lens and the camera lens, with both lenses aimed at the subject field. The DMD projection unit and sensor camera optics co-alignment must be mechanically precise and robust so that measurement calibration can be established and maintained. The Machine Vision functionality also requires precise localization of the field of view with respect to the robotic unit and its tools.

The chosen measurement algorithm determines the type and number of patterns used. Patterns may be either binary (1-bit) or multi-bit (gray scale). The measurement algorithm (multi-scale binary, phased mult-bit depth, or other method), and the type and number of patterns used will affect the speed, resolution, and accuracy of the measurements.
The measurement algorithm is implemented in software and executed on a suitably capable processor (embedded or PC-based). The 3D measurement process produces a point cloud (set of x, y, z measurements), which can be analyzed and utilized in any number of ways to provide a precision 3D visual field to the robotic system. In addition to the 3D point cloud measurement function, the Machine Vision system may also analyze shapes, edges, textures, colors, and other parameters revealed by visual measurement and analysis.

The sequence of patterns required for a full measurement take a finite period of time to occur. The relative movement of the Machine Vision system (robot) and objects in the field of vision during the measurement interval must be considered and accommodated in order to avoid measurement errors. Faster pattern rates allow for fewer errors due to this relative motion.

The DMD is part of a chipset, and must be utilized with its chipset devices. The LED illuminators (and drivers) require sufficient power and thermal considerations, which can be considerable.