초록 ▼

A system for recognizing objects in images, and using that recognition to solve mission problems, such as identifying, tracking, and reporting movement of mobile military targets of interest. The system incorporates preprocessor and postprocessor with an image correlator, control systems, and feedba

A system for recognizing objects in images, and using that recognition to solve mission problems, such as identifying, tracking, and reporting movement of mobile military targets of interest. The system incorporates preprocessor and postprocessor with an image correlator, control systems, and feedback loops functioning with a real-time operating system. Tracking includes correlating patterns in an image stream so as to monitor image changes that represent changes of location, shape, trends, or other attributes. Reporting may entail visual displays of correlation results, or other data outputs suitable for steering gimbaled sensors, aiming weapons, or for other functions. The instant invention has application in numerous areas where recognition and tracking of patterns and trends are desired in imagery or in other two-dimensional representations of data.

대표청구항 ▼

Having thus disclosed my/our invention and the manner of its use, it should be apparent that incidental changes may be made thereto that fairly fall within the scope of the following appended claims, wherein we claim: 1. A method using an optical correlator, a preprocessor and a postprocessor for p

Having thus disclosed my/our invention and the manner of its use, it should be apparent that incidental changes may be made thereto that fairly fall within the scope of the following appended claims, wherein we claim: 1. A method using an optical correlator, a preprocessor and a postprocessor for performing near real-time correlations between images of a sequence of images possibly containing an object of interest and electronically stored reference filters selected from a plurality of electronically stored reference filters each containing an optically modified reference view of an object of interest comprising: A) storing said plurality of reference filters in a first electronic library, B) electronically capturing an image of said sequence of images as an electronic representation, C) selecting one or more of said reference filters likely to provide at least one correlation peak with said possible said object of interest in an electronically captured said image of said sequence of images, D) using said optical correlator, rapidly comparing, in near real-time, each of said selected reference filters to said electronically captured image of said sequence of images, E) using said post processor to identify those said selected reference filters having a highest correlation peak with said electronically captured image, and, F) developing at least one feedback signal from step E) for predictably controlling and optimizing further selection of said reference filters and said preprocessor performing preconditioning of following electronically captured images of said sequence of images. 2. The method as set forth in claim 1 further comprising: selecting, through use of at least one predictive algorithm, a small plurality of said reference filters likely to develop correlation peaks with an object of interest in said following electronically captured images, storing said small plurality of said reference filters in a second, fast temporary electronic library for fast retrieval. 3. The method as set forth in claim 2 wherein said F) further comprises varying gamma, gain and offset of a said electronically captured image to optimize said correlation peaks. 4. The method as set forth in claim 3 wherein said F) further comprises applying a local region algorithm to optical components of said electronically captured image to optimize said correlation peaks. 5. The method as set forth in claim 1 wherein said C) further comprises selecting one or more reference filters based on exogenous data with respect to a possible object of interest in at least a first electronically captured image of said sequence of images. 6. The method of set forth in claim 5 further comprising selecting following said reference filters for correlation with following said electronically captured images based on a located correlation peak between said object of interest in said one or more reference filters based on said exogenous data and said object of interest in said at least a first electronically captured image. 7. The method as set forth in claim 6 further comprising: H) determining a trend from said correlation peaks in said following electronically captured images, and, I) predictably selecting said following reference filters based on said trend, and storing said following reference filters in a second temporary, fast memory for fast retrieval and correlation with a next said electronically captured image of said sequence of images, and, J) repeating said H) and I) in order to track said object of interest in said sequence of images. 8. The method as set forth in claim 7 further comprising calculating location of said object of interest in a said electronically captured image within about five milliseconds or so. 9. The method as set forth in claim 1 further comprising generating said reference filters by: K) directing a collimated beam of light through an image of a said object of interest applied to a liquid crystal spatial light modulator, L) developing a complex conjugate of a Fourier transform of said image of said object of interest to generate a reference filter of said reference filters, M) developing a complex conjugate of said reference filter, N) performing an autocorrelation between said reference filter and said complex conjugate of said reference filter, O) varying an intensity level of said collimated beam of light to obtain a highest correlation peak, P) associating a selected said intensity level of said collimated beam of light that obtains a highest said correlation peak with a respective said reference filter. 10. The method as set forth in claim 9 wherein said D) further comprises: Q) adjusting an intensity level of a collimated, coherent beam of light directed through said electronically captured image to said selected intensity level of a said selected reference filter to obtain a spatially modulated image with a light intensity level generally that of said selected intensity level, R) performing a Fourier transform of said spatially modulated image to obtain a Fourier transformed, spatially modulated image, S) passing said Fourier transformed, spatially modulated image through a said selected reference filter to obtain a combined image, T) performing a Fourier transform on said combined image to obtain a similarity map indicating degree of correlation between said electronically captured image and said selected reference filter and where correlation with said object of interest is located in said current image. 11. A system employing an optical correlator for identifying and reporting selected patterns in at least some images from an image source and comprising: a plurality of optical reference filters each including a complex conjugate of a Fourier transform of a reference filter of a pattern being sought, a preprocessor receiving said images and selectively enhancing said images by varying gamma, gain and offset to optimize correlation peaks, said optical correlator receiving enhanced said at least some images and at least one optical reference filter of said plurality of said optical reference filters wherein a single image of said at least some images at a time is compared to said at least one optical reference filter to produce a similarly map, a post-processor receiving said similarly map and indicating at least one of a strength and corresponding location of any existing said correlation peaks, and, at least one feedback loop for predictively selecting said reference filter based on selected features of said similarity map. 12. The system of claim 11 wherein said optical correlator comprises: a source of coherent light, a collimating lens receiving said coherent light, a first liquid crystal light modulator upon which said input image is impressed, and receiving collimated said coherent light so that said collimated light is modulated by said input image, a first Fourier transform lens receiving said collimated, coherent light modulated by said input image, a second liquid crystal light modulator receiving Fourier transformed said collimated, coherent light modulated by said input image and upon which said optical reference filter is impressed for producing a product of a Fourier transform of said input image and said reference filter, a second Fourier transform lens receiving said product and re-transforming said product, a multi-pixel detector array upon which a retransformed said product falls and is detected as a similarly map. 13. The system of claim 11 further comprising an adoptive, spatially-variant, constant false alarm rate threshold algorithm in said post processor, and a feedback loop for adjusting said false alarm rate threshold. 14. The system of claim 12 wherein an intensity of said coherent light source is adjustable to a light intensity level of a selected said reference filter, obtaining a spatially modulated image with a light intensity level generally that a selected light intensity level of said reference filter. 15. A system employing an optical correlator for identifying, tracking, and reporting patterns in a stream of input images from an input source, said system comprising: a reference filter store including a plurality of reference filters, each reference filter comprising a complex conjugate of a unique view of an object of interest, at least some of said reference filters comprising an indication of a light intensity level of a coherent light source used to create each of said at least some of said reference filters wherein said light intensity level is a said light intensity level obtaining a highest correlation peak in an autocorrelation process used to create a respective said reference filter, a preprocessor receiving said stream of input images and enhancing for correlation at least some of said input images, an optical correlator receiving an enhanced input image and one or more selected reference filters to produce a similarity map based on an optical comparison between said enhanced input image and said selected reference filter, and further wherein a light intensity level of a coherent light source used in said correlation process to illuminate said input image is adjusted to said indication of a light intensity level of a selected said reference filter being compared with said input image, a post-processor receiving said similarly map and producing correlation outputs that indicate strength and location of at least one located correlation peak in said similarly map, said post-processor further applying predicted filtering in conjunction with said correlation outputs to track at least one parameter of said object of interest, said parameter including at least one location, aspect, and scale of said object of interest, and a feedback loop that enables selection of said one or more reference filters from said filter store responsive to said post-processor. 16. The system of claim 15 wherein said optical correlator comprises: said source of coherent light used in said correlation process, a collimating lens for collimating said coherent light, a first liquid crystal light modulator receiving collimated said source of coherent light used in said correlation process and whereupon said enhanced input image is impressed so that said collimated coherent light is modulated by said enhanced input image, a first Fourier transform lens receiving said collimated coherent light modulated by said input image and providing a first Fourier transform of said collimated coherent light modulated by said input image, a second liquid crystal light modulator receiving said first Fourier transform and upon which is impressed a said reference filter selected by said predictive filter, so that said first Fourier transform is further modulated to become a product of said first Fourier transform and said selected reference filter, a second Fourier transform lens receiving said product and providing a second Fourier transform, a multi-pixel optical detector upon which said second Fourier transform impinges and develops said similarly map as an output. 17. The system of claim 15 further comprising a system-driver-to-reference-logic loop that uses said location of a said located correlation peak from said post processor to drive a pointing system associated with said input source in order to track said object of interest.