Design And Implementation Of Fatigue Detection System Based On FPGA+ARM

According to reports on road traffic safety accidents at home and abroad,fatigue driving is the leading cause of traffic safety accidents,and the concealment and severity of accidents are much higher than others.The existing fatigue detection system can only rely on the driver’s adjustment,and cannot realize remote monitoring and online reminding.No fatigue detection system is equipped on the low-end and middle-end vehicles,whether it is domestic or foreign.This subject researches and designs a system that can accurately detect the driver’s fatigue state,realizes HDMI and host computer multi-terminal real-time display,voice and SMS dual reminder,which is of great significance for reducing road traffic safety accidents.By comparing and analyzing the existing fatigue detection technologies at home and abroad,the method based on visual features is selected to extract driver fatigue characteristics.The Intel So C FPGA chip is used as the core of the system,and the FPGA+ARM architecture is adopted.Image acquisition,eye and mouth fatigue characteristic parameter extraction and fatigue determination,HDMI real-time display,Ethernet transmission upper computer real-time display,buzzer alarm,and SMS sending are integrated in an embedded system.With the Open CV computer vision library,through pre-processing steps such as image graying,histogram equalization,and Gaussian filtering,illumination compensation is performed to improve image contrast,and to filter out noise interference during image acquisition,transmission,and storage.The Ada Boost algorithm based on Haar-Like features is used to realize face detection,and the cascaded posture regression algorithm is used to realize the location of 68 feature points of the face,and the eyes and mouth regions are accurately located according to the position of the feature points of the face.It is proposed to use the eye aspect ratio EAR value combined with the cumulative number of black pixels in the binary image of adjacent frames to recognize the eye state,and the mouth aspect ratio MAR value to recognize the mouth state.Multi-feature fusion is used to establish a fatigue detection model,and the PERCLOS algorithm P80 standard combined with blink frequency and yawn frequency are used as evaluation criteria to determine the mental state of the driver.Combining the OV5640 image sensor and the SIM900 A module,the system architecture and related algorithms of this article are implemented on Terasic’s hardware development platform DE-10 Nano,the performance of the system is analyzed,and the system is tested in the laboratory and in the actual vehicle environment to prove the system.It can accurately realize the positioning of eyes and mouth regions,fatigue feature extraction and fatigue detection,and the constructed fatigue detection system is efficient,real-time and accurate.