The Research And Application Of Infrared Thermography Face Recognition

The face recognition technology is a biometric technology which has a great application value in the direction of security monitoring, access control and data confidentiality, and is the hot spot of the research in the field of pattern recognition. The infrared thermal imaging face recognition is a branch of the field of face recognition and is the focus of biometric technology in recent years, and also is a valuable application. Due to the anti-interference, light source independent, anti-counterfeiting properties of the infrared thermal radiation, it can make up for the defects of the visible light image face recognition. With the increasing infrared face recognition technology, its practical application will be more extensive, will generate the increasing value.This thesis focuses on the infrared Thermal Imaging Face Recognition process to achieve the target of an online infrared face recognition system, and discusses the recognition process of infrared thermal image acquisition, preprocessing, training. A new method to improve existing recognition method is proposed to improve the recognition results.Firstly, the infrared thermal image capturing is discussed, and the theory of infrared thermal imager is studied. Then blood flow data instead of the temperature and spectral data is proposed, and the appropriate pseudo-color coding is used, making the infrared face thermal image quality highly upgraded;Secondly, the infrared face thermal image preprocessing is discussed. The preprocessing includes angle correction and size correction, and the face region extraction and segmentation using the interpolation step of anti-aliasing, etc. Preprocessed image is clear and has a high quality, which has laid a good foundation for the later accurately identification;Then, the recognition and training method of infrared thermal face is discussed. The classic PCA method is studied and be pointed out its inadequacies. And then to its shortcomings, a new optimized method is proposed, and then the effectiveness of the new method is confirmed from the theoretical and experimental results. The new method extracts the features based on Gabor model, and uses the Bayesian classifier to replace classic Euclidean distance classifier, and changes classical algorithm in a projection subspace into Bayesian theory class within the subspace and class subspace, and improves the original classical algorithm which is easy to identify the infrared thermal image falsely with a different time period, different ambient temperatures as a different face of different people;Finally, an infrared thermal image online face recognition system is designed and implemented. The design uses the heuristic rules in software engineering as a guide, and uses the object-oriented design methods. The design of the system architecture is open and flexible, and reflects the characteristics of the object-oriented high cohesion, low coupling, which enhances system framework’s ability to cope with change, and reduces the maintenance costs.