Research On Algorithm For Overlooked Infrared Targets Recognition Based On Multiple-spectral Information

Automatic infrared target recognition is a senior intelligent image processing technology focused by domestic and overseas military departments. From 1980s, many researchers devoted themselves to this filed and lots of new theories and algorithms have been developed. However, it still could not be applied in the weapon systems for battlefield environment.This thesis proposes automatic recognition algorithms for airport, water, airplane and ship based on single-spectral and multiple-spectral information respectively. The efficiency and robusticity of algorithms have been demonstrated by the experiment. This thesis is summarized as follow:(1) Analyze the principium of infrared imaging and the differences of the objects consisted of different materials in infrared imaging. The analysis provides support for the related research.(2) Propose the algorithm for airplane recognition based on single-spectral infrared information, in consideration of the relationship between airplane and airport. At first, the algorithm detects the airport in infrared image with complex background and then recognizes the airplanes in the airport region. The algorithm enhances the rapidity and reduces the difficulty of airplane recognition.(3) Propose the algorithm for warship recognition based on single-spectral infrared information, in consideration of the relationship between warship and water. At first, the algorithm detects the water in infrared image with complex background and then recognizes the warships in the water region. The algorithm enhances the rapidity and reduces the difficulty of warship recognition.(4) Propose long wave and middle wave infrared images fusion algorithm in feature level for airplane and warship recognition based on modular neural network. The algorithm integrates the advantages of long wave and middle wave infrared images on the base of effective feature selection and makes the recognition result by multiple-spectral information better than single spectrum information.(5) Propose infrared image and visible image fusion algorithm in decision level for airplane recognition. The algorithm integrates the advantages of infrared image and visible image and makes the recognition result by multiple-spectral information better than single spectrum information.