چكيده انگليسي :
Design and Hardware Implementation of a Face Detection System Mohammad Reza Mahmoodi mr mahmoodi@ec iut ac ir Date of Submission Department of Electrical and Computer Engineering Isfahan University of Technology Isfahan 84156 83111 Iran Degree M Sc Language FarsiSupervisor Dr Sayed Masoud SayediAbstractFace detection has numerous applications in robotics surveillance systems and humancomputer interaction HCI In the first part of this thesis a skin detection algorithm isproposed which has a remarkable detection rate The algorithm first conservatively extractsthe initial seed the points which are related to human skin with high probability Thenother skin pixels are extracted in a multi step multi feature diffusion method In addition aface detection algorithm is designed which uses the color information of the image in askin classification preprocessing step to eliminate a considerable portion of the image generally The concept of kernel probability map is introduced to detect face relatedwindows of the image Combining the kernel probability map concept with the newintroduced texture LHP has increased the performance of the face detector For hardwareimplementation a new low cost and high speed platform based on FPGA and MATLAB isproposed which can be used for the implementation of various signal image processingalgorithms The system is utilizable for other purposes including on chip verification using computer as an external memory for the FPGA developing hardware software co designs and modeling analog digital systems This setup is compromised of two connectedhardware and software cores implemented on FPGA and a computer respectively In thesoftware part a management core is designed in order to either send or receive the raw orprocessed data In an image processing application the image is derived from a camerathat is connected to the PC Data will be transferred to the FPGA using Ethernettechnology and based on UDP protocol The architecture of the FPGA design is divided tothree parts For establishing the connection a relatively optimized UDP IP stack isdesigned using pipeline structure Due to the need of a high speed dense memory in manyapplications a 1Gb s DDR3 SDRAM is employed thus the second part of hardwaredesign is related to the architecture which allows the dynamic data transmission betweenthe FPGA and SDRAM The third part of the hardware design is related to the directimplementation of an algorithm based on constraints and limitations that the system andthe circuit imposes Real time processing is achieved through this system for skindetection and motion detection algorithm face detection is also possible based on theobtained information Both post place route simulation and hardware implementation results confirm the validity of the system Keywords 1 Skin detection 2 Face Detection 3 Hardware implementation 4 FPGA
