Research On Infrared Image Target Feature Extraction And Classification

Automatic target recognition (ATR) in infrared image is an important research subject in the field of pattern recognition, which plays a key role in precision-guided weapons and has wide application prospects. Feature extraction and classification of targets in infrared image are difficult points and key techniques in ATR. The study of feature extraction and classification will improve efficiency and accuracy of the infrared system, which is of great importance to increase survival probability of our army.This thesis mainly deals with the feature extraction technologies in infrared image and target multi-classification methods. As presupposition and preparation of feature extraction this paper introduces infrared image segmentation algorithm based on threshold in brief. Meanwhile, simulation experiments and performance comparison have been done. For feature extraction, firstly, linear feature extraction algorithms using Principal Component Analysis (PCA),Independent Component Analysis (ICA) and nonlinear algorithms using kernel functions are studied. Secondly, in the light of the disorder of the features extracted by ICA, an ordered extraction algorithm based on evaluation factor minimum is brought forward. Experiments show that the algorithm can select the features which are more different among different classes and achieve high classification accuracy just with fewer features. For target multi-classification, firstly, minimum distance (MD) classifier, nearest neighbor (NN) classifier, K nearest neighbors (KNN) classifier and SVM (support vector machine) classifier are discussed in detail. Secondly, for the ambiguous problem of KNN classifier used in multi-classification, a new multi-classification method based on Hadamard Error Correcting Output Code combining with KNN is proposed. The experimental results show that this method has the ability of anti-jamming and increases the target classification rate further.