Infrared And Visible Target Recognition Based On Fuzzy Support Vector Machine

With the rapid development of science and technology, digital image recognitiontechnology in the civilian, military and other fields have achieved a very wide range ofapplications. This paper combined the advantages of infrared sensors and visible sensors.An infrared and visible target recognition system is constructed based on fuzzy supportvector machine theory.Fuzzy support vector machine has become a complete theoretical system, which isbased on statistical theory, the VC dimension theory, etc. As a new theory, thisalgorithm can help to reach a quite highly accuracy in target identification.This paper is investigated on the basis of theory of digital image processingtogether with fuzzy support vector machine theory. Infrared and visible light targetrecognition system on the basis of FSVM is studied and designed in more detail. Thecontents of this research mainly includes its image pre-processing module, featureextraction module, database management module, training module of recognition modeland target identification module.Development process for each module target recognition system problems at workon the basis of previous studies makes a number of practical and effective method fortargeted improvements, including: image pre-processing for infrared target gray degreecharacteristics, based on the OTSU algorithm to ask thresholds to be adjusted so that thethreshold after adjustment is closer to the target area of gray, improving segmentationresults; K-Means algorithm is necessary to optimize and improve the selection strategyof the initial cluster centers for some inherent problems, and the clustering results innoise area presents a regional consolidation strategy for processing; Based on LIBSVM,system adds a fuzzy membership function by improving the existing classificationfunction and builds FSVM recognition algorithm which has been programmed;Considering the real-time requirements, the system uses C++language development anddesign and achieves the development of FSVM infrared and visible target recognitionsystem by the use of Visual C++compiler running environment.Through experiments, the recognition speed and other content recognition accuracy of the infrared and visible FSVM target recognition systems has been verified.Experimental results show that FSVM system has good identification accuracy. Systemhas good practical value.