Three Generations of Optical Trackers

MicronTracker is the first of a third generation of optical trackers.

First generation trackers use infrared light emitting diodes (IREDs) as targets, observed by infrared-only light sensors. This approach has a number of significant drawbacks, the main one being that power needs to be provided to the IREDs, either via unreliable and movement-limiting wires, or via batteries which add weight and bulk and require periodic replacement. Many first-generation trackers also cannot distinguish between individual IREDs illuminating simultaneously, and thus require pulsing them one at a time in sequence. Pose measurement errors frequently occur when the marker is in motion, since the positions of the targets making up the marker are measured at different times.

Second-generation sensors emit infrared light from a ring surrounding each infrared camera lens and use balls or discs as targets. The balls or discs are coated with a retro-reflective material containing small spheres that mirror the light back to the lenses, causing the targets to appear as bright spots in the infrared images. This approach eliminates the need for wires to bring power to the targets, improving usability and reliability. On the other hand, specially coated targets are required. To eliminate accuracy degradation, these targets need to be replaced frequently at a substantial cost and inconvenience.

Third generation trackers use available light in the visible spectrum to detect and pinpoint high-contrast targets printed or painted on any suitable surface. Standard video lenses and sensors are used to capture the images, which are then fed into the host computer for processing. Computer vision software detects the presence of target patterns, and then pinpoints the location of the feature point at the pattern's center.

The fundamental differences in approach provide our 3rd generation product family, MicronTracker, with many significant advantages over earlier generations. These advantages will be explained in the next few sections.

Pinpointing Reliability: Centroid vs. Xpoint

Both 1st and 2nd generation trackers use bright infrared spots to mark target locations. Target centers are pinpointed by computing the intensity centroid of the spot's image. A smudge or a partial occlusion affects the centroid's position, leading to a significant measurement error. Alarmingly, this condition is not detected and the user is not alerted.

MicronTracker computes target locations at the intersection of 4 high-contrast regions, called an Xpoint. Each of the four black/white boundary lines independently serves to pinpoint the location of the target. Portions of the Xpoint region hidden by smudges or occlusions are disregarded, with no significant impact on accuracy.

Marker Identification Reliability

Bright spots in an image do not contain any geometrical information other than their center's location. Xpoints have both a location and an orientation. This additional discriminating characteristic greatly reduces erroneous mismatches between targets seen on left and right images, and misidentifications of markers, where the characteristics of the observed targets are matched against templates. Finally, misleading bright reflection spots are more often naturally found in an operating environment than Xpoints.

Closer and Smaller

The intensity of the light reaching 1st and 2nd generation cameras from their infrared targets falls with the square of the distance the light travels. A target 4 times farther from the camera shows 16 times dimmer in the image, making it barely detectable in comparison. This problem makes it difficult to design earlier generation cameras to have a close, yet deep enough, FOM. In turn, obtaining a given level of accuracy at a larger distance from the camera requires that the lenses be placed proportionally farther apart, increasing the size and weight of the camera.

With available light coming from sources far behind them, 3rd generation tracker cameras may be placed much closer to the FOM without sacrificing FOM depth. This allows the lenses to be placed close together, leading to a smaller, lighter, construction. Such a camera fits unobtrusively into cluttered environments where space is at premium. It does not require large arms or tripods for support. In fact, it can be mounted directly into the patient holder/table, eliminating the need for a separate reference marker. It is easier to move, store and ship. Peopleare less likely to accidentally bump into it during an operation or experiment.

Easier to Position

Positioning the camera at the outset to eliminate the need for repeated repositioning during the measurement session is often a major challenge with trackers of earlier generations. MicronTracker's FOM starts only 20cm in front of the camera. The camera can be placed right in front of the volume to be measured, eliminating the difficulty of estimating imaginary boundaries 1-2 meters away. Better still, MicronTracker provides visible light video images, showing the user exactly what the camera sees and where the visibility boundaries are. If tracking of an instrument is lost, no guessing at the reason is necessary - a quick peek at the images is all it takes.

Reduced line of sight interruptions

With their small size and close FOM, MicronTracker cameras are positioned such that it is much less likely people or objects will interfere with the line of sight to the markers being tracked. Furthermore, the small camera size, low cost, and excellent software support make multi-camera configurations both practical and affordable. By overlapping the FOMs of different cameras, tracking continues even when targets are hidden from one of the cameras.

Easy targets

Being simply visible patterns, targets can be made in many different fashions to provide the most convenient fit to their intended usage. They can be printed on throwaway paper, printed on reusable, sterilizable, plastic sheets, or painted directly on instruments. Lightweight and flat, they can be attached directly to skin and bone and eliminate the need for cumbersome patient attachments.

Unlimited number of tools/markers

Its suprior identification reliability allows MicronTracker to easily distinguish between markers only slightly different from each other. MicronTracker's unlimited marker templates database allows each camera to discriminate between hundreds of different tools. Better yet, the database may be updated at run-time, allowing new marker templates to be added simply by presenting them to the camera and assigning them a name. Highly efficient algorithms and a fast in-memory data reporting allow MicronTracker to concurrently track dozens of markers at full frame rates.

Access to images and lower-level controls

MicronTracker's programming interface, containing over 200 well-documented functions, puts the tracking operations under the control of the application programmer. Low-level data elements, such as images and objects identified in them, may be directly accessed, and various attributes of the tracking operation can be manipulated. All processing is activated directly by the application. While the "native" API is provided in C, it is object-oriented and can be easily accessed from high-level languages, such as C++, Visual Basic, or Python, through available wrappers. Demonstration applications in source code format and a developer's manual help programmers quickly integrate MicronTracker into their product.

FOM to size

Enlarging the FOM is simple: add more cameras. Multiple cameras can be added while maintaining complete software compatibility with a single camera configuration. Registering the individual FOMs together is as simple as momentarily placing a single marker in the region where they overlap.

With its small size and closer FOM, MicronTracker can operate within limited spaces, such as inside the bore of an imaging scanner. MicronTracker is even MRI-compatible.

Augmented reality support

3-D geometry can be accurately overlaid on the video images to provide the user with visual coordinate registration accuracy feedback, hidden locations of interest, an object cutting/placement plan, or any other useful spatial information. Since the stereo video images themselves define the coordinate system, the augmented reality overlay does not suffer from the annoying visual motion lag common to other augmented reality presentations (click on the images to the right to see video clips).

Affordability

Perhaps the most important advantage 3rd generation trackers have over older designs is low price. Due to its design simplicity and the high volume production of its standard hardware components, MicronTracker's manufacturing costs are far lower than the more custom infrared hardware and specialized signal processing electronics used in older generations. These savings are passed directly on to customers.

Not only is the tracker itself more affordable, printed/painted markers reduce or eliminate per-use cost, making MicronTracker-based guidance systems more attractive to its intended market.

A long upgrade path ahead

The start of a new generation of pose-tracking technology, MicronTracker is only at the beginning of an exciting evolution path. The continuously improving cost/performance ratio of consumer video electronic parts will enable faster, higher-resolution image sensors, allowing targets to shrink and FOMs to expand. Periodic software updates will enable novel system capabilities and increase its flexibility in meeting the needs of a growing range of applications.