Motivation: Face recognition is a rapidly growing research area due to increasing demands for security in commercial and law enforcement applications.  Face recognition systems based on visual images have reached a significant level of maturity with some practical success.Despite the success of automatic face recognition techniques in some practical applications under controlled operating environments, recognition based only on visual spectrum remains still limited.Thermal IR imagery is independent of the ambient illumination since the human face and body are the emitters. Thus under poor illumination conditions, where visual face recognition system usually fails, the use of thermal imagery has great advantages. Objectives: The objective of this research is to develop face recognition technology and to combine visual and thermal face recognition . The detailed objectives below may be achieved step by step. At first, comparison of visual and thermal face recognition performance will be evaluated in order to verify which modalities has better performances with manually aligend eye positions in thermal images.  Next, eye detection in the thermal images should be conducted.The eyes are the most important features of faces in visual images. They are used to normalize faces and their surrounding regions are also used for recognition in visual images. Since the eyes in thermal images are not clear compared to visual images, a different approach should be developed. We will address the problem of eye detection and glass regions removal in order to minimize the effect of glasses and achieve a fully automatic recognition. Finally,  the results of visual and thermal face recognition will be fused together to achieve more reliable performance, Technical Approach: As can be seen in Figure 2, the results of visual and thermal face recognition are fused together in order to achieve more reliable performances.  Visual Face Recognition FUSION Better Performance Thermal Face Recognition Figure2. Fusion of visual and thermal face recognition   Figure3 shows that several face images are acquired in the process of tracking people. Then we are going to construct a 3D model that can estimate different poses. Finally, this face will be matched against databases to identify each individual. Figure 3. Video Tracking Based Face Recognition using a 3D model Results: We conducted a detailed evaluation of FaceIt®, a high ranking [by FRVT 2000 and 2002] face recognition product developed by Identix,   in terms of expression, illumination, age, pose and face size, factors that of major influence facial recognition. Comparison of co-registered visual  and thermal images [Acquiredby Equinox Company] resulted in thermal face recogniton has higher performance than visual images when individuals are not wearing glasses. Automatic eye aligning succeeded 95% of the images and 5% of the images were aligned manually to achieve best recognition rates on visual images, while thermal images were all manually aligned without using same eye coordinates from the visual images. Eye detection in the thermal images has been investigting in the thermal images to achieve a fully automatic recognition and combine the results of visual face recognition.  Publications: J. Heo, B. Abidi, J. Paik, and M. A. Abidi, "Face Recognition: Evaluation Report For FaceIt®," to appear in Proc. 6th International Conference on Quality Control by Artificial Vision QCAV03, Gatlinburg, TN, USA, May 2003. This research is being conducted at the IRIS Lab by Jingu Heo under the supervision of Dr. Mongi A. Abidi.

---

Last updated:
Webmaster