Method for occlusion of movable objects and people in augmented reality scenes

[0002] This application claims priority of Provisional patent application 60/284,315 filed Apr. 17, 2001, and is a Continuation in Part of “Augmented Reality-Based Firefighter Training System and Method” Ser. No. 09/525,983 filed Mar. 15, 2000.

[0004] This invention relates to computer graphics, augmented reality, and mixed reality. It defines a way to simulate the occlusion of virtual objects by a real person and/or other moveable objects in an augmented reality graphical application.

[0005] A portion of the disclosure of this patent document contains material that is subject to copyright protection. The copyright owner has no objection to the facsimile reproduction by anyone of the patent document or the patent disclosure, as it appears in the Patent and Trademark Office records but otherwise reserves all copyright works whatsoever.

[0006] In augmented reality, in order to improve realism, it is necessary for real objects to occlude virtual objects. To do this, virtual models of such real objects must be created that correspond in space and time to the location of the real objects. One problem with traditional modeling techniques is that hard edges introduced by polygonal models can be distracting if there is misregistration in augmented reality due to tracking errors and lag.

[0007] The purpose of the invention is to create a method to allow a real person or movable physical object to occlude virtual objects in an augmented reality application using a single tracking data sensor for both human and non-human objects. The invention comprises an orthogonal plane billboarding technique that allows textures with fuzzy edges to be used to convey the sense of a soft-edged 3D model in the space. This is done by creating three orthogonal (perpendicular) planes. A texture map is mapped onto each of these planes, consisting of profile views of the object of interest as silhouettes of how they look from each of these directions. One application for this orthogonal plane billboard is the modeling of a human head and torso.

[0008] FIG. 1 is a diagram of the three orthogonal planes that contain the three texture mapped images of a human head, useful in understanding the invention.

[0009] FIG. 2 is a diagram of a human head and torso, which will be compared to graphical components in FIG. 3; and

[0010] FIG. 3 is a diagram of two sets of orthogonal planes, along with the joint between the two sets, for the human head and torso of FIG. 2.

[0011] In the invention, texture maps representing the movable object are mapped onto three orthogonal planes. A tracking sensor is used to determine the actual position and orientation of the object, as appropriate. When the viewer's viewpoint is perpendicular to a plane, the texture map is opaque. When the viewpoint is parallel to the plane, the texture map is transparent. The texture map is faded between these two extremes as the orientation changes between these two extremes, to accomplish a desirable graphical mixing result that matches the silhouette of the object (or human) while maintaining a softer border around the edge of the silhouette contained in the texture map. The appearance of a soft (“fuzzy”) border is made by fading to transparent the edges of the object silhouette in the texture map.

[0012] In the preferred embodiment, as shown in FIG. 1, a series of three texture maps containing silhouettes 1, 8, and 9 (in this case of a human head, from the side, top, and front, respectively) are shown mapped onto each of three orthogonal planes 10, 11 and 12, respectively. The texture maps may fade to transparent at their edges for a fuzzy appearance to the shape. The orthogonal planes are each broken up into 4 quadrants defined by the intersection of the planes, and the 12 resulting quadrants are rendered from back to front for correct alpha blending in OpenGL of the texture maps and planes, with depth buffering enabled. When a plane is perpendicular to the view plane of a virtual viewpoint looking at the object, the plane is rendered to be completely transparent. A linear fade is used to completely fade the texture map to completely opaque when the plane is parallel to the view plane. This fade from opaque to transparent as the planes are turned relative to a viewer is responsible for the large part of the desirable fuzzy appearance to the shape. The texture maps used to shade the three planes were created from a digital image of a person, then made into grayscale silhouettes, and so match the silhouette of a human user very well. The edges 7 of the human silhouettes in the texture maps were blurred so that they would fade linearly from solid white (which represents the silhouette) to solid black (which represents non-silhouette portions of the texture map) to look better in an augmented reality situation where a little bit of a fuzzy edge is desirable. This fuzzy edge spans what is equivalently approximately 0.5 inches in real distance.

[0013] FIG. 2 depicts a diagram of a human 3 wearing a head tracker 2 on a head mounted display. The virtual representation of human 3 is shown used in the inventive technique in FIG. 3. Item 2 in FIG. 2 is a motion tracker, in this case a six-degree-of-freedom motion tracker that measures the head location and orientation. The position and orientation information from tracker 2 can be applied to orthogonal plane billboards.

[0014] If the object being depicted has a pivot point, such as pivot point 5 in FIG. 3 for a human neck-torso joint, an orthogonal plane billboard torso 4 can be created in approximately the correct place relative to the joint. The torso in this instance may be designed to remain upright, only rotating about a vertical axis. The head in this instance has full 6 degree-of-freedom motion capability based on the data coming from the tracker worn on the head of the user. This allows the head orthogonal plane billboard to be lined up correspondingly with the user's head. The torso orthogonal plane billboard is attached to the pivot point 5 and is placed “hanging” straight down from that point, and has 4 degrees of freedom: three to control its position in space, and one controlling the horizontal orientation.

[0015] The head and torso models, when lined up to a real human, occlude computer-generated graphics in a scene. If augmented reality video mixing is achieved with a luminance key to combine live and virtual images, white head and torso models will mask out a portion of a computer-generated image for replacement with a live video image of the real world. This invention can be applied to any real world movable objects for which an occlusion model may be needed.

[0016] The above technique is also applicable to a movable non-person real world object. If the non-person object has no joints, then such an implementation is simpler since the complexity of coupling the separate head and torso models is avoided. The technique for the single movable real-world physical object is functionally identical to the above method when only the head model is used.

[0017] Although specific features of the invention are shown in the drawing and not others, this is for convenience only, as each feature may be combined with any or all of the other features in accordance with the invention.

[0018] Other embodiments that will occur to those skilled in the art are within the following claims.