Image analysis – Pattern recognition – Template matching
Reexamination Certificate
2000-12-20
2004-08-24
Mariam, Daniel (Department: 2621)
Image analysis
Pattern recognition
Template matching
C382S209000
Reexamination Certificate
active
06782131
ABSTRACT:
BACKGROUND OF THE INVENTION
1. Field of the Invention
The present invention relates generally to the field of pattern recognition. More specifically, the present invention relates to a system, method and apparatus for pattern or symbol recognition usable in real-time applications.
2. Background Art
There are a number of applications for, and approaches to, pattern recognition. For example, neural networks can be used for character recognition. But high-volume off-the-shelf semiconductor devices for neural networks are not available. Therefore, implementations that employ neural networks can be relatively expensive and they carry an inherent risk that the parts employed might become obsolete. Also, because of compromises that are typically made to neural network devices in order to lower their cost, they may not meet the performance requirements for some real-time applications. Similarly, Digital Signal Processors (DSP's) can also be used for pattern recognition (e.g., with correlation operations performed in either the time or the frequency domain). But DSP's can also be expensive and may not achieve the performance level required by some real-time applications. Thus, it would be useful to have an alternate approach to pattern recognition that can achieve high performance at a relatively low-cost without significant risk of parts obsolescence.
Many display systems employ cathode ray tubes (CRT's), but new CRT's are not always available to repair or replace these display systems. Therefore, a demand exists to replace these displays with newly designed display systems that utilize flat-panel display technology, such as liquid crystal display (LCD) technology. But the electrical interface to existing CRT-based displays is often incompatible with the requirements of LCD-based displays because of the differences between these two technologies. The primary incompatibility of these two technologies is in the way that the individual pixels of each display type are selected for illumination. LCD technology employs a mechanism for discretely selecting individual pixels, whereas CRT technology employs analog voltage signals to position one or more electron beams on the display surface.
Display processing systems for real-time applications, such as for avionics displays, generally incorporate a hardware vector generation (i.e., line drawing) capability. This hardware provides digital outputs for selecting X (horizontal) and Y (vertical) display screen locations, and digital video to specify the selected intensity and/or color. A separate video on/off output may also be provided. For a calligraphic/stroke type of CRT display, the X, Y, and video signals are converted to analog to directly control the position and intensity of the CRT electron beam(s). The symbol generator hardware then directly refreshes the display by executing its commands at a rate of approximately 60 HZ. For a dynamic display format, the command-list/program should be updated at a rate that is high enough to achieve smooth motion of the symbology. This may require an update rate of 20 or 30 HZ.
For LCD technology—or for a raster type of CRT-based display—the display screen is refreshed from a 2-dimensional frame buffer memory, or “image memory”. An image memory can be an exact 2-D mapping of the pixels of the display screen, whereby each memory location holds a code that determines the color/intensity of its corresponding pixel. The digital X and Y outputs of the symbol generator are then used to address the 2-D image memory. Vectors are written/drawn into image memory instead of directly on the display screen. For real-time applications, a symbol generator is used to update the display format in one image memory while another image memory, which contains the previous update, is simultaneously used to refresh the display screen. Some CRT-based displays are designed to alternate between stroke and raster in a hybrid mode in order to overlay stroke symbology onto raster video.
Avionics display systems often employ separate physical units for the display function and the symbol generator/processor function. Stroke and hybrid CRT displays may utilize an analog interface between the symbol generator and the display unit. If such CRT-based displays are replaced with newer LCD-based units, it may be cost prohibitive to also replace the symbol generator. The new LCD-based display unit should then be designed to function with the existing analog interface. But because LCD technology employs a discrete type of pixel addressing mechanism, this tends to diminish the quality of the displayed symbology.
A conventional approach to the design of a LCD-based display unit with stroke-type analog inputs is to essentially mimic, or simulate, the operation of a CRT-based display. The analog X and Y deflection inputs are converted to digital and used to address a 2-D image memory where the digital/digitized video is written. Thus, the stroke inputs are used to write symbology into an image memory instead of directly onto a CRT display screen. This image memory is then subsequently scanned and read to provide the video for refreshing a LCD screen. While the digitized inputs are used to write into one image memory, a second memory (which stores the prior update) is used to refresh the display. With this design approach, the analog interface contributes errors/anomalies to the displayed image. These result from: 1) induced electromagnetic interference (EMI) on the analog X and Y signals—from external and internal sources; 2) wideband thermal noise generated within the analog circuitry; 3) phase noise in the sampling operation; and 4) quantization noise from the analog to digital (A-to-D) conversion.
All of these noise sources tend to degrade the quality of the displayed symbology. The severity of EMI depends on the individual application. A low-pass pre-filter is preferably employed before the sampling A-to-D conversion operation to prevent high-frequency EMI from aliasing into the bandwidth of interest, but lower frequency EMI will degrade the quality of the displayed symbology. Thermal noise causes small random deviations—that exhibit a Gaussian distribution around the norm—in the digitized samples of the analog deflection inputs. The effect of thermal noise can be minimized by oversampling the inputs (i.e., sampling at a higher rate than is required by downstream operations) and then employing a low-pass digital filter. Phase noise results from jitter in the sampling time, which can occur within the A-to-D converter circuit and also from jitter in the sampling clock. But in this application, a more significant error can result from an inconsistent phase of the sampling clock relative to the clock in the symbol generator (i.e., on different updates of the display format). The display anomalies that result from quantization noise can be minimized by employing a high resolution A-to-D converter (e.g., a 12-bit converter may provide a significantly higher resolution than the pixel resolution of the display).
Display anomalies can be categorized as either static or dynamic. A noise or distortion that is the same on every display update would be categorized as static. Its location would be fixed relative to either the display screen or to a symbol on the display screen (i.e., where the symbol itself might be in motion). An example of a static anomaly would be the distortions to individual symbols that result from aggressive low-pass filtering of the deflection inputs (e.g., resulting in symbols with severely rounded corners). A dynamic anomaly can be a noise or distortion that moves across the screen and/or through the symbology. It can also be located at a specific screen or specific symbol location and be exhibited on either a periodic or random basis. Since humans are adept at detecting motion/change, dynamic anomalies can be more objectionable than static ones. Increasing the amount of low-pass filtering can reduce the dynamic anomalies that result from EMI. However, this would increase the level of static distortion for
Armijo Dennis F.
Honeywell International , Inc.
Luther Kurk
Mariam Daniel
LandOfFree
System, method and apparatus for pattern recognition with... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with System, method and apparatus for pattern recognition with..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and System, method and apparatus for pattern recognition with... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3355067