Design And Implementation Of Fatigue Detection System Based On FPGA

With economy booming and social development,motor vehicles have become the most basic tools for people to travel,but the road traffic safety accidents caused by motor vehicles bring irreparable casualties and huge property losses to China and even the world.According to the statistical reports on related road traffic accidents at home and abroad,the road traffic accidents caused by fatigue driving account for 25% of the total,and there is a rising trend.Therefore,how to accurately and effectively detect the fatigue state of drivers has become one of the research hotspots of scholars in related fields at home and abroad.At present,most of the researches on fatigue driving detection are based on image processing and detection algorithm,and relatively few are based on hardware and system realization.Due to the factors such as environment,illumination and occlusion,the speed and accuracy of fatigue detection are seriously affected.At the same time,because the data cannot be processed in parallel,it is difficult to meet the real-time requirements of fatigue detection.In this thesis,based on the system design idea of hardware and software collaboration,a non-contact fatigue detection system architecture based on FPGA is constructed.Combining with the powerful parallel computing data capability of FPGA,image processing algorithm and fatigue detection algorithm based on embedded processor are implanted to realize fatigue detection.Finally,the system architecture and algorithm are proved to be real-time and efficient on PYNQ-Z2 development platform.The main research work of this thesis is as shown below:1.Study and build a video acquisition and display system based on FPGA hardware programmable logic design.2.In this thesis,the basic theories of face detection,eye detection and fatigue detection are studied,and the difficulties and key points of fatigue detection are analyzed and compared.3.Construct a EAR SVM classifier to judge eye state and blink detection by EAR3 d feature vector.In view of the secondary positioning and feature extraction of humaneyes,the feature points of human eyes are detected to accurately locate the human eyes and extract the aspect ratio of the eyes as the eye features,so as to avoid the interference of external light and glasses and other shielding objects in the extraction of human eyes by image processing technology.4.In this thesis,it is proposed that PERCLOS-P80 and blink frequency jointly evaluate fatigue,avoiding the use of a single fatigue evaluation index,and improving the accuracy and reliability of fatigue detection.5.In PYNQ-Z2 development,the system architecture and related algorithms in this thesis are fully implemented,and the real-time and validity of the constructed fatigue detection system is verified via the experimental and test results.