Research On Face Recognition Technology In Complicated Environment

Nowadays, Face recognition technology has made considerable progress after yearsof development, and many reliable and classic methods have been commercially used inevery field to serve people. However, the traditional methods are usually used in quite anideal environment such as laboratory, and the methods often need the testers to showenough characteristics of face, such as keeping still sitting posture or the head moving incertain range with small turning.As the development of intelligent surveillance and control system, the work used tobe done by human was expected to be done by computers automatically. So the operationof face recognition becomes necessary and important in daily life. But the environment ofsurveillance videos is quite complicated, one simple image may have many faces,different sitting postures or sheltered parts, meanwhile, the video system which needs thecontinuous recognition to the objectives even more increase the difficulty of theprocessing. In order to resolve the problems above and construct the complete recognitionsystem, this paper proposes some new improved algorithms mainly focus on the facerecognition in complicated environment evolved face detection, feature extraction andrecognition.First, Adaboost+CPCA algorism is proposed based on face recognition algorism ofAdaboost, the improved algorithm adopts the recognition method which combines thecolor information of objectives to improve the precision rate.Second, Feature extraction and recognition technology usually needs to process theunconventional face and feature extraction especially should have the light robustness toadapt to different kinds of light detection. SIFT feature extraction which has the rotationand scaling invariant can be well applied to solve the problems above. A new algorithm isproposed in this paper which combines feature matching and fuzzy classificationanalyzes successfully processes the whole image using SIFT feature extraction with lessuseless features which affects the specific objectives matching. Third, focus on face recognition technology, meanwhile, this paper makes use ofobject tracking technology. This paper combines image detection application andaccomplishment of Gaussian model with the face detection recognition system, theimproved the algorism which is a new reform recognition method for the undetectedobjectives transferred the unknown objectives recognition to the known objectivestracking.Finally, combining the algorithms above, the new automated detection recognitionsystem was constructed which can detect the video image in daily life, this system alsohas better recognition result even when the objectives were partially shaded.