Vision RT - Company > Technology

Core Technology

3D SURFACE RECONSTRUCTION

Vision RT’s imaging technology employs stereo vision techniques in the same way as the human visual system perceives depth in 3D. By viewing an object through two eyes (cameras) from different perspectives, the brain is able to derive depth information of a scene from the disparity between the two retinal images.

Computer vision algorithms are able to adopt this approach in order to derive 3D surface information of an object. To accomplish this, the positions, orientations, and optical properties of two different cameras viewing the same object, must be computed. This is achieved through a process known as camera calibration. This involves imaging an object on which a precise pattern has been produced, the dimensions of which are known.

AlignRT^® Calibration Plate

Determination of a 3D Point via Triangulation

The calibration process then derives the orientation, position, and optical properties for each respective camera, with respect to a known reference point.

Once the two cameras are calibrated, corresponding points are determined between both stereoscopic images. Through a process known as triangulation, the actual 3D position of these points may be be computed.

Download AlignRT Core Technology PDF (1.24MB)

In order to compute the 3D surface model of a patient, all possible corresponding points as seen from each data camera (figure 1) must first be identified.

This is achieved by projecting a pseudo-random optical pattern via either a speckle projector (monitoring mode) or a speckle flash (static capture) onto the patient in order to provide uniqueness to all points on the imaged surface (figure 2).

By utilising an advanced matching algorithm, all visible corresponding points may be determined between the two data camera images and a 3D surface model may be computed (figure 3).

AlignRT^® employs two camera units (figure 4), each housing two stereoscopic cameras. 3D surface data can be acquired from each of these pairs of cameras and may be combined to produce a complete model which extends all the way around the patient.

AlignRT^®’s proprietary* technology employs specially designed and configured cameras and high speed 3D reconstruction algorithms which together provide the combination of speed, precision, versatility and robustness, unique to the AlignRT^® product range. This technology is core to all Vision RT's products.

3D SURFACE ALIGNMENT

High speed surface matching algorithms* have been developed for the purpose of aligning two surfaces together. These compute the full six degree of freedom transformation and generate the associated rotations and translations which define the movements required to align the two surfaces (figure 5).

MISCELLANEOUS

The following are some of the other technologies utilised by AlignRT®:

Isocalibration: Proprietary* technique for calibrating the cameras precisely to the treatment iso-centre;
Texture map imaging: Imaging the photorealistic image of the patient and superimposing this on the 3D surface model (figure 6);
Lung modelling algorithm: Used to estimate lung contours from a 3D skin surface of the breast and a medial x-ray;
Gated capture technology: Utilisation of markerless point tracking technology* for gating 3D data capture to a reproducible point in the respiratory cycle.

* AlignRT^® is protected by granted British patents (GB2390792 and GB2407466) and other patent applications relating to Vision RT’s technology which are pending worldwide.

| |