Research on Feature Extraction and Classification Algorithms for Infrared Targets

Infrared target detection techniques have been widely used in infrared alarming and reconnaissance, and the detection task is often done by machine learning. Feature extraction and classification algorithm are two core components of the machine learning. The selection of features and algorithm have a determinative effect on the detection result. Traditional target classification techniques only focus on only limited combinations of features and classification algorithm, which may result in a poor detection result. Based on this, this paper selects the gray features, statistical features, frequency domain features, and graphic features of the infrared target, and compares the three classification algorithms of KNN, Bayes, and SVM to give the optimal combination of features and classification algorithms by experiment.