System and method for object tracking
A hand-manipulated prop is picked-up via a single video camera, and the camera image is analyzed to isolate the part of the image pertaining to the object for mapping the position and orientation of the object into a three-dimensional space, wherein the three-dimensional description of the object is stored in memory and used for controlling action in a game program, such as rendering of a corresponding virtual object in a scene of a video display. Algorithms for deriving the three-dimensional descriptions for various props employ geometry processing, including area statistics, edge detection and/or color transition localization, to find the position and orientation of the prop from two-dimensional pixel data. Criteria are proposed for the selection of colors of stripes on the props which maximize separation in the two-dimensional chrominance color space, so that instead of detecting absolute colors, significant color transitions are detected. Thus, the need for calibration of the system dependent on lighting conditions which tend to affect apparent colors can be avoided.
1. Field of the Invention
The present invention relates to computer vision systems, and more particularly to a system in which an object is picked-up via an individual video camera, the camera image is analyzed to isolate the part of the image pertaining to the object, and the position and orientation of the object is mapped into a three-dimensional space. A three-dimensional description of the object is stored in memory and used for controlling action in a game program, such as rendering of a corresponding virtual object in a scene of a video display.
2. Background of the Invention
Tracking of moving objects using digital video cameras and processing the video images for producing various displays has been known in the art. One such application, for producing an animated video version of a sporting event, has been disclosed by Segen, U.S. Pat. No. 6,072,504, the disclosure-of which is incorporated in the present specification by reference. According to this system, the position of a tennis ball during play is tracked using a plurality of video cameras, and a set of equations relating the three-dimensional points in the court to two-dimensional points (i.e. pixels) of digital images within the field of view of the cameras are employed. Pixel positions of the ball resolved in a given digital image can be related to a specific three-dimensional position of the ball in play and, using triangulation from respective video images, a series of image frames are analyzed by a least-squares method, to fit the positions of the ball to trajectory equations describing unimpeded segments of motion of the ball.
As described in some detail by Segen, once a three-dimensional description of position and motion of an object has been determined, various methods exist which are well known in the art for producing an animated representation thereof using a program which animates appropriate object movement in a video game environment.
Stated otherwise, Segen is concerned with determining the three-dimensional position of an object in motion from a plurality of two-dimensional video images captured at a point in time. Once the three-dimensional position of the “real” object is known, it is then possible to use this information to control a game program in any number of different ways which are generally known to game programmers.
However, the system of Segen relies on a plurality of video cameras for developing positional information about the object based on triangulation. Moreover, the detected object of Segen is a simple sphere which does not require information about the orientation (e.g. inclination) of the object in space. Thus, the system of Segen is not capable of reconstructing position and orientation of an object, whether moving or at rest, from a two-dimensional video image using a single video camera.
It is common for game programs to have virtual objects formed from a combination of three-dimensional geometric shapes, wherein during running of a game program, three-dimensional descriptions (positions and orientations) of the objects relative to each other are determined by control input parameters entered using an input device such as a joystick, game controller or other input device. The three-dimensional position and orientation of the virtual objects are then projected into a two-dimensional display (with background, lighting and shading, texture, and so forth) to create a three-dimensional perspective scene or rendition by means of the rendering processor functions of the game console.
As an example, there can be “virtual object” that forms a moving image in a game display corresponding to how one moves around the “real” object. To display the virtual object, the calculated three-dimensional information is used for fixing the position and orientation of the “virtual object” in a memory space of the game console, and then rendering of the image is performed by known projection processing to-convert the three-dimensional information into a realistic perspective display.
However, in spite of the above knowledge and techniques, problems continue to hinder successful object tracking, and a particularly difficult problem is extracting precisely only those pixels of a video image which correspond unambiguously to an object of interest. For example, although movement of an object having one color against a solid background of another color, where the object and background colors vary distinctly from one another, can be accomplished with relative ease, tracking of objects, even if brightly colored, is not so easy in the case of multi-colored or non-static backgrounds. Changes in lighting also dramatically affect the apparent color of the object as seen by the video camera, and thus object tracking methods which rely on detecting a particular colored object are highly susceptible to error or require constant re-calibration as lighting conditions change. The typical home use environment for video game programs demands much greater flexibility and robustness than possible with conventional object tracking computer vision systems.
SUMMARY OF THE INVENTIONIt is a principal object of the present invention to provide an object tracking system suitable for video game programs which overcomes the aforementioned disadvantages of the conventional art, enabling tracking of the position and orientation of an object or prop which serves as an input device for effecting action in the game program.
Another object of the invention is to provide alternative interfaces for games, wherein rather than using a conventional joystick, the user can stand in front of a video camera connected to the game console device, and by physically moving or manipulating an object within view of a single camera, cause a corresponding action to occur in the game.
Yet another object of the invention is to provide methods for mapping two-dimensional information of a discriminated pixel group belonging to the manipulated object to a three-dimensional space, to provide a three-dimensional description of the position and orientation of the object in three dimensions from a single video camera image.
Yet a further object of the invention is to provide a three-dimensional description of an object which includes a rotational component of the manipulated object.
A still further object of the invention is to provide techniques for the selection of object colors which maximizes one's ability to discriminate, on the basis of color transitions, the pixel groups from a video image which belong unambiguously to the manipulated object and which provide the needed information for deriving a description of three-dimensional position and orientation of the object.
The above and other objects, features and advantages of the present invention will become apparent from the following description when taken in conjunction with the accompanying drawings in which preferred embodiments of the present invention are shown by way of illustrative example.
BRIEF DESCRIPTION OF THE DRAWINGS
The game console 60 constitutes a component of an overall entertainment system 110 according to the present invention which, as shown in
The main memory 114, vector calculation unit 116, GIF 112, OSDROM 126, real time clock 128 and input/output port 124 are connected to the MPU 112 over a data BUS 130.
Further connected to the BUS 130 is an image processing unit 138 which is a processor for expanding compressed moving images and texture images, thereby developing the image data. For example, the image processing unit 138 can serve functions for decoding and development of bit streams according to the MPEG2 standard format, macroblock decoding, performing inverse discrete cosine transformations, color space conversion, vector quantization and the like.
A sound system is constituted by a sound processing unit SPU 171 for generating musical or other sound effects on the basis of instructions from the MPU 112, a sound buffer 173 into which waveform data may be recorded by the SPU 171, and a speaker 175 for outputting the musical or other sound effects generated by the SPU 171. It should be understood that the speaker 175 may be incorporated as part of the display device 80 or may be provided as a separate audio line-out connection attached to an external speaker 175.
A communications interface 140 is also provided, connected to the BUS 130, which is an interface having functions of input/output of digital data, and for input of digital contents according to the present invention. For example, through the communications interface 140, user input data may be transmitted to, and status data received from, a server terminal on a network. An input device 132 (also known as a controller) for input of data (e.g. key input data or coordinate data) with respect to the entertainment system 110, an optical disk device 136 for reproduction of the contents of an optical disk 70, for example a CD-ROM or the like on which various programs and data (i.e. data concerning objects, texture data and the like), are connected to the input/output port 124.
As a further extension or alternative to the input device, the present invention includes a digital video camera 190 which is connected to the input/output port 124. The input/output port 124 may be embodied by one or more input interfaces, including serial and USB interfaces, wherein the digital video camera 190 may advantageously make use of the USB input or any other conventional interface appropriate for use with the camera 190.
The above-mentioned image processor 120 includes a rendering engine 170, a main interface 172, an image memory 174 and a display control device 176 (e.g. a programmable CRT controller, or the like).
The rendering engine 170 executes operations for rendering of predetermined image data in the image memory, through the memory interface 172, and in correspondence with rendering commands which are supplied from the MPU 112.
A first BUS 178 is connected between the memory interface 172 and the rendering engine 170, and a second BUS is connected between the memory interface 172 and the image memory 174. First BUS 178 and second BUS 180, respectively, have a bit width of, for example 128 bits, and the rendering engine 170 is capable of executing high speed rendering processing with respect to the image memory.
The rendering engine 170 has the capability of rendering, in real time, image data of 320×240 pixels or 640×480 pixels, conforming to,.for example, NTSC or PAL standards, and more specifically, at a rate greater than ten to several tens of times per interval of from {fraction (1/60)} to {fraction (1/30)} of a second.
The image memory 174 employs a unified memory structure in which, for example, a texture rendering region and a display rendering region, can be set in a uniform area.
The display controller 176 is structured so as to write the texture data which has been retrieved from the optical disk 70 through the optical disk device 136, or texture data which has been created on the main memory 114, to the texture rendering region of the image memory 174, via the memory interface 172, and then to read out, via the memory interface 172, image data which has been rendered in the display rendering region of the image memory 174, outputting the same to the monitor 80 whereby it is displayed on a screen thereof.
There shall now be described, with reference to
As shown in
As noted above, the essential information which must be provided is a three-dimensional description of the object, which in the case of
Initially the pixel data input from the camera is supplied to the game console 60 through the input/output port interface 124, enabling the following processes to be performed thereon. First, as each pixel of the image is sampled, for example, on a raster basis, a color segmentation processing step S201 is performed, whereby the color of each pixel is determined and the image is divided into various two-dimensional segments of different colors. Next, for certain embodiments, a color transition localization step S203 is performed, whereby regions where segments of different colors adjoin are more specifically determined, thereby defining the locations of the image in which distinct color transitions occur. Then, a step for geometry processing S205 is performed which, depending on the embodiment, comprises either an edge detection process or performing calculations for area statistics, to thereby define in algebraic or geometric terms the lines, curves and/or polygons corresponding to the edges of the object of interest. For example, in the case of the cylinder shown in
The three-dimensional position and orientation of the object are calculated in step S207, according to algorithms which are to be described in association with the subsequent descriptions of preferred embodiments of the present invention.
Lastly, the data of three-dimensional position and orientation also undergoes a processing step S209 for Kalman filtering to improve performance. Such processing is performed to estimate where the object is going to be at a point in time, and to reject spurious measurements that could not be possible, and therefore are considered to lie outside the true data set. Another reason for Kalman filtering is that the camera 190 produces images at 30 Hz, whereas the typical display runs at 60 Hz, so Kalman filtering fills the gaps in the data used for controlling action in the game program. Smoothing of discrete data via Kalman filtering is well known in the field of computer vision and hence will not be elaborated on further.
In
As shown in
Referring now to
To determine the length, center point, etc. of the pixel group 307 in accordance with the geometry processing step S205 discussed above, known area statistics calculations are used. Area statistics include the area, centroid, moment about the X-axis, moment about the Y-axis, principal moments, and the angle of principal moments, which typically are used for calculating moments of inertia of objects about a certain axis. For example, to determine the moments about the X and Y axes, respectively, if each pixel making up the pixel group is considered to correspond to a particle of a given uniform mass m making up a thin homogeneous sheet or lamina, then the moments about x and y axes of a system of n such particles (or pixels) located in a coordinate plane are defined as follows:
The center of mass of this system is located at the point (x, y) given by
Further, assuming the lamina is of a shape having a geometric center, such as the rectangle in the case of
Because the image 305 is already taken to be in the X-Y plane, the X-Y position of the center point p can be derived directly from the image. Also, the theta θ quantity is taken directly from the image simply by knowing any line l, determined in accordance with the geometry processing step S205 described above, which runs along the longitudinal axis of-the pixel group 307 corresponding to the cylinder 301. Typically, a longitudinal line l passing through the center point p is used for this purpose.
Determination of the phi φ quantity requires some additional knowledge about the pixel width w in at least two different locations w1 and w2 wherein the ratio of the width quantities w1: w2 provides a value which can be used for determining φ. More specifically, if the cylinder 301 is inclined so that the top end thereof is closer to the camera 190 than the lower end of the cylinder, then, since the lower end of the cylinder is at a greater distance from the camera 190, the pixel width quantity w2 of the image will have a greater value than the pixel width quantity w1, and vice versa. The ratio w1:w2 is proportional to the inclination φ of the cylinder 301 in the Y-Z plane, and therefore the phi quantity φ can be determined from this ratio. Typically, for better accuracy, a plurality of equidistant measurements of pixel widths between ends of the pixel group 307 are taken, and averaging is performed to determine the ratio w1:w2.
Determination of the depth quantity Z can be done in different ways. However, it is important to recognize that the size and number of pixels making up the two-dimensional pixel group 307 are affected both by the inclination-of the object in the φ direction as well as by the actual distance of the physical object from the video camera 190. More specifically, as the object inclines in the φ direction, the apparent length of the object as seen by the video camera tends to shorten, so that the length l of the pixel group shortens as well. However, at the same time, as the object moves farther away from the camera along the Z-axis, the apparent size of the object overall, including its length l, also becomes smaller. Therefore, it is insufficient simply to look at the length l alone as an indicator of how far away from the camera the object is. Stated otherwise, the depth quantity Z must be determined as a function of both and φ.
However, if the phi quantity φ has already been determined and is known, a phi-weighted value of l, which we may call lφ, can be determined, and the pixel length of lφ in the image, which changes as the object is moved closer or farther from the camera while assuming that φ stays constant, then can be used to determine the depth quantity Z since lφ will be proportional to Z.
Another method for determining depth Z is to count the total number of pixels in the pixel group 307 corresponding to the object. As the object gets closer to or farther from the camera, the number of pixels making up the pixel group 307 increases or decreases respectively, in proportion to the depth quantity Z. However, again, the number of pixels in the pixel group 307 is also affected by the inclination in the phi φ direction, so the number of pixels N must first be weighted by phi φ to result in a weighted quantity Nφ which is used for determining the depth quantity Z based on a proportional relationship between Nφ and Z.
Yet another advantageous method for determining the depth quantity Z is to use the average width Wavg of the rectangle, which is calculated as the sum of a given number of width measurements of the rectangle divided by the number of width measurements taken. It should be clear that the average width of the pixel group is affected only by Z and not by the phi-inclination of the cylinder. It is also possible to determine phi φ from the ratio of the total length of the pixel group to the average width (i.e. l:Wavg), and moreover, wherein the sign of the phi-inclination can be determined based on whether w1 is greater or less than w2.
In
The prop according to the second embodiment, similar to the first embodiment shown in
As shown in
It should also be realized that, unlike the cylinder in the previous embodiment, the shape and size of the circular pixel group are not influenced as a result of the phi φ angle of inclination. More specifically, even if the object overall is tilted in the phi direction, the sphere 309 and the pixel group 311 will retain their general shape and, unlike the length of the cylinder 301, will not become foreshortened as a result of such tilting. Therefore, an advantage is obtained in that the total number of pixels of the pixel group making up the circle in the image can always be related proportionately to the depth quantity Z and, for determining Z, phi-weighting as in the previous embodiment is not required.
Determination of inclination of the object in the theta θ direction is done directly from the image, just as in the previous embodiment, by determining the angle theta θ between a center longitudinal line of the pixel group 307 corresponding to the cylinder 301 and the Y-axis.
Determining the angle of inclination in the phi φ direction is handled somewhat differently than the previous embodiment. More specifically, such a quantity can be determined by knowledge of the depth quantity Z, determined as described above, and by the length l between the center point of the circle 311 and the center point of the pixel group 307 which corresponds to the cylinder 301. For any known and fixed depth quantity Z, the length 1 (as viewed from the perspective of the camera) becomes shorter as the object is tilted in the phi φ direction. Therefore, if the Z quantity is known, it is possible to determine, simply from the length l, the degree of inclination in the phi φ direction, and it is not necessary to calculate a relative width quantity or ratio of widths, as in the embodiment shown by
As in the embodiment shown in
According to this embodiment, a pixel group making up the cylinder is extracted from the image to provide a two-dimensional line along which to look for color transitions. To determine the quantities Z, θ and φ, positions are determined at which color transitions along any line l in the longitudinal direction of the cylinder 301 occur.
More specifically, as shown in
D={square root}{square root over ((ΔCr)2+(ΔCb)2)} (4)
By selecting colors which maximize the value of D (to be explained in more detail later), it is possible to select a threshold Dt at which only color transitions above a certain separation where D>Dt are considered to correspond to the color transitions of the stripes S1, S2 and S3. Accordingly, the pixels along the line of the cylinder are filtered, using such a threshold, in order to find the large color transitions corresponding to the stripes S1, S2 and S3.
As shown in
It shall next be explained how knowledge of l1 and l2 provides sufficient information for determining the quantities of Z, θ and φ, necessary for describing the position and orientation of the object in three dimensions.
First, since the line l defined by the pixels running along the length of the cylinder has already been determined, and since the camera is assumed to face normally to the X-Y plane, the angle θ is taken directly as the angle between the longitudinal line of the cylinder and the Y axis, basically in the same manner as the preceding embodiments.
For determining the angle of inclination in the phi φ direction, the ratio of the lengths l1:l2 is used. For example, in the case (as shown) in which the cylinder is inclined in the direction toward the camera 190, with the upper end of the cylinder being closer to the camera than the lower end, the length l1 will appear longer to the camera 190 (since it is closer) than the length l2. It will also be appreciated that, although the apparent lengths l1 and l2 will also be affected by the overall distance Z of the object from the camera 190, the ratio of these lengths l1:l2 will not change and therefore this ratio provides a constant indication of the inclination of the cylinder 301 in the phi φ direction.
For determining the depth quantity Z, a procedure similar to that of the first embodiment is employed, wherein a phi-weighted quantity lφ of the total length l (l=l1+l2) is determined for giving Z. More specifically, the influence of the inclination angle φ on the total apparent length l of the object is first determined, and then the total length, properly weighted by the influence of φ, is proportional to the distance (or depth quantity) Z of the object from the camera 190.
Stated more simply, φ is determined from the ratio of l1 and l2, and once phi φ is known, the total depth quantity Z can be determined from the sum of l1 and l2.
There shall now be described, in connection with
Each of the tracking methods described above can be used to obtain five of the six degrees of freedom of the objects. The only one missing is the rotation of the cylinder about its axis. Information about the rotation of the cylinder would seem difficult to obtain because cylinders are symmetric in rotation about this axis. The approach taken by the present invention to obtain this rotational component is to add a helical stripe SH that goes around the cylinder 301 exactly once. As the cylinder 301 is rotated, the height of the stripe SH will correspond to the rotation angle.
More specifically, as shown in
In addition to the helix H, a center line l of the pixel group corresponding to the cylinder is determined as described previously. Also the overall length l of the pixel group is determined.
For obtaining a degree of rotation of the cylinder, various heights h (only h1 and h2 are shown for simplicity) each of which define the distance between one end of the cylinder and the point p where the center line intersects the helix are determined.
As shown on the right-hand side of
The specific quantity used for determining rotation is the ratio of the detected height between the lower end and the point on the helix to the total length l of the pixel group. This ratio gives a number from 0 to k (where k=hmax/l) which maps directly to a range of from 0 to 360 degrees. Thus, additional information with respect to the object and orientation of the cylinder 301 in three-dimensional space can be provided. Such information can be used to control the rotation of a virtual object, for example, when displayed in a game program.
Next, with respect to
By contrast,
According to the present invention, the color of the stripes S1, S2 and S3 and the color of the cylinder 301 are chosen in such a way as to maximize stripe detectability for the video camera. Color-based tracking is notorious for its difficulties due to changes in lighting, which cause the apparent color to vary. As a result, if one is attempting to detect a certain color of blue corresponding to an object, for example, under certain lighting conditions it is possible for the color of blue, as perceived by the camera, to vary to such a degree that accurate detection of the object is made difficult. In the present invention, by looking for color transitions instead of absolute colors, a more robust tracking solution can be attained. For example, in the embodiment of
As discussed above, video cameras capture data using the two-dimensional chrominance color space shown in
More specifically, as shown in
Although blue and orange have been described as an example, it will be appreciated that-any other color pairs, for example green and magenta, which also possess a large separation in the chrominance color space may be used. In other words, the method provides a general criteria whereby colors may be selected using their chrominance signals Cr and Cb in such a manner to maximize their separation in the color space.
More specifically, a generally applicable method for the selection of colors, as well as for calculating distance between any two colors, is performed in such a way that the distance between two colors is calculated as a distance projected onto a certain diameter-spoke of the color wheel. First, a given diameter-spoke on the color wheel is selected having a certain angle of orientation θ. By choosing the angle of orientation of the selected diameter on the color wheel, it is possible to select the color transitions one wants to detect. For example, if green is (1, 1) and magenta is (−1, −1), the diameter of the spoke should be set at an orientation θ of 45 degrees. Then the color separation distance is calculated simply by projecting the colors onto the 45 degree line. In this manner, for the case of green and magenta, the computed distance is exactly the same as the Pythagorean distance D discussed above, however with a diameter-line orientation of 45 degrees, the distance between blue and orange is zero, because they both project to the origin. This tells us that, for a selected diameter line of 45 degrees, green and magenta are the optimal colors for detection, since they possess the maximum separation in the color space for this diameter.
Thus, for any given diameter angle of θ, which can be chosen from 0 to 180 degrees, the separation between two colors (Cr1, Cb1) and (Cr2, Cb2) may be calculated according to equation (5) as follows:
D=[Cr1 cos θ+Cb1·sin θ]−[Cr2·cos θ+Cb219 sin θ] (5)
The distance calculation shown by equation (5) can therefore also be used for setting the threshold Dt based on a predetermined orientation defined by the angle θ. For example, if the color transitions for the object were in fact green and magenta, the general distance calculation above can be used for threshold setting, while fixing the angle θ of this equation at 45 degrees.
Herein have been described several methods for determining the position and orientation of a real object manipulated in front of a video camera, by mapping the two-dimensional image information of the object captured by the camera to a three-dimensional space, wherein a three dimensional description including position and orientation of the object may be used to control action in a game program.
Although one clear example of controlling a game program is to have a “virtual object” that forms a moving image in a game display corresponding to how the “real” object is moved or positioned, it will be appreciated that the three-dimensional information can be used to control game programs in any number of different ways foreseeable to persons skilled in the art. For example, a “theremin” like musical effect can be achieved wherein changes in the position and orientation of the manipulated object could be used to influence volume, tone, pitch, rhythm and so forth of sounds produced by the sound processor. Such a musical or rhythmic sound effect can by provided in combination with visual effects displayed on the screen of the game console, for enhancing the experience perceived by the game player.
It shall be understood that other modifications will be apparent and can be easily made by persons skilled in the art without departing from the scope and spirit of the present invention. Accordingly, the following claims shall not be limited by the descriptions or illustrations set forth herein, but shall be construed to cover with reasonable breadth all features which may be envisioned as equivalents by those skilled in the art.
Claims
1. An object tracking system comprising:
- an input device configured to detect two-dimensional input pixel data from a prop device; and
- a multiprocessor unit configured to calculate three-dimensional position and orientation data associated with the prop device from the two-dimensional input pixel data.
2. The object tracking system of claim 1 wherein the multiprocessor unit further comprises a memory configured to store the three-dimensional position and orientation data associated with the prop device.
3. The object tracking system of claim 1 wherein the multiprocessor unit further comprises an image processor configured to execute operations for rendering the three-dimensional position and output data associated with the prop device.
4. The object tracking system of claim 1 further comprising a monitor for displaying action in a game program caused by rendering the three-dimensional position and output data associated with the prop device.
5. The object tracking system of claim 1 further comprising the prop device configured to provide input data.
6. The object tracking system of claim 1 wherein the input device is a camera.
7. The object tracking system of claim 2 wherein the memory further comprises a display rendering region.
8. The object tracking system of claim 2 wherein the memory further comprises a texture rendering region.
9. A machine readable medium having embodied thereon a program being executable by a machine to perform a method for tracking and discriminating pixel groups, the method comprising:
- receiving pixel data from an input device;
- segmenting color whereby the color of each pixel is determined and the image is divided into two-dimensional segments of colors;
- defining edges of an object in the image; and
- calculating three-dimensional position and orientation data of the object.
10. The machine readable medium of claim 9 wherein the method further comprises localizing color transitions whereby distinct color transitions are defined.
11. The machine readable medium of claim 9 wherein the method further comprises filtering the three-dimensional position and orientation data.
12. The machine readable medium of claim 9 wherein defining edges of an object employs an edge detection process.
13. The machine readable medium of claim 9 wherein defining edges of an object employs performing calculations for area statistics.
14. The machine readable medium of claim 9 wherein a definition of an edge of the object is algebraic.
15. The machine readable medium of claim 9 wherein the definition of an edge of the object is geometric.
16. The machine readable medium of claim 11 wherein the filtering comprises Kalman filtering.
17. A method for tracking and discriminating pixel groups comprising:
- receiving pixel data from an input device;
- segmenting color whereby the color of each pixel is determined and the image is divided into two-dimensional segments of colors;
- defining edges of an object in the image; and
- calculating three-dimensional position and orientation data of the object.
18. The method of claim 17 further comprising localizing color transitions whereby distinct color transitions are defined prior to defining the edges of the object in the image.
19. The method of claim 17 further comprising filtering the three-dimensional position and orientation data.
20. A system for tracking and discriminating pixel groups comprising:
- means for receiving pixel data from an input device;
- means for segmenting color whereby the color of each pixel is determined and the image is divided into two-dimensional segments of colors;
- means for defining edges of an object in the image; and
- means for calculating three-dimensional position and orientation data of the object.
Type: Application
Filed: Aug 26, 2004
Publication Date: Feb 3, 2005
Inventor: Richard Marks (Foster City, CA)
Application Number: 10/928,778