Image analysis – Pattern recognition – Classification
Reexamination Certificate
2000-12-01
2004-06-22
Chang, Jon (Department: 2623)
Image analysis
Pattern recognition
Classification
Reexamination Certificate
active
06754390
ABSTRACT:
FIELD OF THE INVENTION
The invention relates generally to the detection and classification of objects in an image, and more particularly to a method of fusing outputs from multiple detection and classification schemes in order to reduce false alarms.
BACKGROUND OF THE INVENTION
Many minehunting sonar systems employ high-resolution side-looking sonars that are effective at detecting and classifying bottom-located sea mines in the complex littoral environment. However, these shallow-water regions are also filled with natural, biologic and man-made objects that generate mine-like sonar returns which, in turn, cause many false detection alarms. Accordingly, there has been much research and development of detection and classification sonar processing schemes that can reduce alarms while still maintaining a high probability of mine detection and classification.
Since much of the research and development of detection and classification schemes is carried out by independent or even competitive entities, the various detection and classification schemes generally involve significantly different technical approaches. Here, the phrase “significantly different approaches” means that the algorithms used by the schemes are based on different mathematical, geometrical and statistical theories. Each approach effectively constrains its algorithms (i.e., sequence of data processing steps) by predisposing (or prejudicing) the way it views the data representing the object it is trying to classify. This results in feature spaces and class boundaries that are constructed in very different ways by each algorithm. Thus, each algorithm is keying on substantially different characteristics of mines so that the ultimate performance of a given sensor in a given environment is predicated on the choice of a detection and classification scheme.
SUMMARY OF THE INVENTION
Accordingly, it is an object of the present invention to reduce false alarms output by target detection and classification schemes.
Another object of the present invention is to take advantage of the different technical approaches used by a variety of detection and classification schemes in order to reduce false alarms.
A still further object of the present invention is to provide a method of detection and classification that makes simultaneous use of multiple detection and classification schemes.
Still another object of the present invention is to provide a method of reducing the number of false detection alarms produced by a sonar system operating in a littoral region.
Other objects and advantages of the present invention will become more obvious hereinafter in the specification and drawings.
In accordance with the present invention, a method of fusing outputs from multiple detection/classification (D/C) schemes is provided. Each of a plurality of D/C schemes provides output scores for an area of interest. Each output score corresponds to a known location in the area of interest and indicates a degree of confidence with respect to a detection of a target at the known location. The output scores are normalized to form normalized output scores associated with each of the D/C schemes. Each of the normalized output scores is categorized into a group based on the known location associated therewith. The normalized output scores for each group are fused in accordance with a fusion rule.
REFERENCES:
patent: 6125194 (2000-09-01), Yeh et al.
patent: 621556 (1994-10-01), None
Huang et al. “Combination of Multiple Classifiers with Measurement Values.” Proc. 2nd Int. Conf. on Document Analysis and Recognition, Oct. 22, 1993, pp. 598-601.*
Khotanzad et al. “Hand Written Digit Recognition Using BKS Combination of Neural Network Classifiers.” Proc. of the IEEE Southwest Symp. on Image Analysis and Interpretation, Apr. 21, 1994, pp. 94-99.*
Kundu et al. “An Integrated Hybrid Neural Network and Hidden Markov Model Classifier for Sonar Signals.” IEEE Trans. on Signal Processing, vol. 45, No. 10, Oct. 1997, pp. 2566-2570.*
Xiao et al. “Neural Network Classification with IFSAR and Multispectral Data Fusion.” IEEE Int. Geoscience and Remote Sensing Symposium Proceedings, vol. 3, Jul. 6, 1998, pp. 1327-1329.*
Sanderson et al. “Multi-Modal Person Verification System Based on Face Profiles and Speech.” Proc. of the 5thInt. Symp. on Signal Processing and Its Applications, vol. 2, Aug. 22, 1999, pp. 947-950.*
Ciany et al. “Computer Aided Detection/Computer Aided Classification and Data Fusion Algorithms for Automated Detection and Classification of Underwater Mines.” Oceans 2000 MTS/IEEE Conference and Exhibition, vol. 1, Sep. 11, 2000, pp. 277-284.
Chang Jon
Gilbert Harvey A.
Peck Donald G.
The United States of America as represented by the Secretary of
LandOfFree
Fusing outputs from multiple detection/classification schemes does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Fusing outputs from multiple detection/classification schemes, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Fusing outputs from multiple detection/classification schemes will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-3339446