Real-Time Eye Location And Detection Based On Adaboost And Its FPGA Implemention

As the development of the information security industry,iris recognition technology as a potential biological recognition technology,which can replace the traditional identity authentication way and ensure the security of personal information,is getting more and more attention.But based on iris recognition technology of current,people need to actively make eyes in front of the camera in order to collect the iris information,it makes its popularization more inconvenience.At the same time,the current target detection system using computer software to implement,which has some shortcomings,such as,poor real-time performance,high power consumption,not meet in the practic engineering of miniaturization,real-time and low power assumption.Aiming at solving the problem of iris recognition technology popularization and engineering demand,and in view of the advantages of FPGA in parallel processing,this thesis implements the real-time detection system of human eye based on FPGA by means of hardware architecture.First of all,this paper introduces the theoretical basis AdaBoost algorithm,and then briefly discusses the advantages of FPGA in the human eye detection algorithm to implement the design,before the algorithm is analyzed.Based on the above we discusses optimized FPGA implementation algorithms.Then,the design of the human eye detection system is introduced in detail.The OpenCV(Open Source Computer Vision Library)is used to train the face and human eye cascade classifier,and the FPGA logic of the system is designed on the HS0034 C platform of Wuhan Hongshi Technology Limited Company.The CMOS camera controlled by FPGA collects the gray scale images.The state machine calls the corresponding module to process the image,and the cascade classifier is used to detect the image.After the detection is completed,the target is merged,and the real target location is marked,and then we transfer the image to the PC software window for displaying through the USB communication.By the actual testing,the result indicates that our system can identify the location of the human eye accurately,and the image with a resolution of 640*480 can be smoothly display,which can reach 18 FPS at 100 MHz operating frequency.So our design basically achieve the expected target.