Historically, there have been various approaches to problems relating to the detection of small, weak, or hidden objects, substances, or patterns embedded in complex backgrounds. One approach has been to use low-dimensional sensor systems that attempt to detect a clean signature of a well-known target in some small, carefully chosen subset of all possible attributes, e.g., one or a few spectral bands. These systems generally have difficulty when the target signature is heavily mixed in with other signals, so they typically can detect subpixel targets or minority chemical constituents of a mixture only under ideal conditions, if at all.
Another approach has been to employ high-dimensional sensor systems that seek to detect well-known (prespecified) targets in complex backgrounds using Principle Components Analysis (PCA) or similar linear methods to construct a representation of the background. Orthogonal projection methods are then used to separate the target from the background. This approach has several disadvantages. The methods used to characterize the background are typically not real-time algorithms, they are relatively slow, and they must operate on the entire data set at once; hence, they are better suited to post-processing than real-time operation. Also, the appearance of the target signature may vary with the environmental conditions — this must be accounted for in advance, and it is generally very difficult to do.
When operating in high-dimensional pattern spaces, massive quantities of data must be managed, which requires hundreds of millions of computations for each pixel. Thus, the need to compress massive quantities of data for storage, download, and/or real-time analysis becomes increasingly important and equally elusive.
The present invention is a system for the rapid compression of hypersensor datasets that contain objects, substances, or patterns embedded in complex backgrounds. A hypersensor is a sensor that produces as its output a high-dimensional vector or matrix consisting of many separate elements, each of which is a measurement of a different attribute of the system or scene under construction; an example is a hyper-spectral imager. Hypersensors based on acoustic or other types of signals, or combinations of different types of input signals, are also possible.
The Compression of Hyperdata with ORASIS (Optical Real Time Spectral Identification System) Multisegment Pattern Sets (CHOMPS) is a collection of algorithms designed to optimize the efficiency of multispectral data processing systems. ORASIS is a method for finding endmembers in a hypersensor image by selecting representative exemplars in a prescreening process, in nearly real time, without assuming any a priori knowledge of the scene. CHOMPS is then used to optimize the processing efficiency by employing two types of algorithms: focus searching algorithms and compression packaging algorithms.
The focus searching algorithms reduce the computational burden by reducing the number of comparisons necessary to determine whether or not data is redundant. Compression is realized by constructing the dataset from the exemplars defined in the prescreening operation, and expressing those exemplars in wave space with the necessary scene mapping data, or further processing the exemplars and expressing the exemplars in terms of endmembers to facilitate the efficient storage, download, and the later reconstruction of the complete dataset with minimal deterioration of signal information.