Optimization Design Of Face Detection Algorithm For Android Platform

Face detection technology is an indispensable technology in computer vision and pattern recognition,and it is a key step in face image processing system.With the rapid development of mobile internet.the application of the technology is more and more widely which is mainly used in authentication,intelligent security and remote video conference,etc.Intelligent terminal which is convenient to carry and has the advantages of simple operation and being widely used is spreaded rapidly in the world.Face detection softwares on the mobile devices have broad application prospects.However,because of the limited processing power of mobile devices,it is not ideal to perform face detection on mobile devices.Therefore,based on the improvement of existing face detection algorithm,an optimized design of face detection algorithm for Android platform is implemented in this paper.Through comparing the two traditional face detection algorithms which are the face detection algorithm based on skin color segmentation and Adaboost face detection algorithm,the advantages and disadvantages of the two face detection algorithms are analyzed in this paper.The traditional Adaboost algorithm is improved in this paper which puts forward a new two-threshold search method to reduce training time of Adaboost classifier to shorten the time of face detection and combines the face detection algorithm based on skin color segmentation and improved Adaboost algorithm.Then,the improved face detection algorithm is transplated into Android platform and Java code in application layer calls the face detection algorithm of the OpenCV class library written with C++ through JNI.Finally,the face recognition system obtains images from camera on the Android platform and realizes face recognition by calling the OpenCV library through face recognition algorithm in the local library.In this paper,the CMU-PIE standard test set and the Caltech Computational Vision Group Archive-Faces are used to test the improved algorithm on the Android platform.After the test,the detection rate of the paper is over 90%,the error detection rate is lower than 9%,and the test time of 24*24 pixels is lower than 1300ms,and the performance indicator of the system is verified.Compared with the traditional Adaboost algorithm,the detection rate of improved algorithm is increased by 1.2%,and the detection time is shortened by 45.8%.At the same time,in a different light intensity,face under the condition of existence angle deflection or obstructions,the effect of face detection is effective.