Research Of Algorithm And Design Of Hardware System For Driver Fatigue Detection

As the transportation industry develops, the traffic accidents have caused a great loss in the property and damage to the society. So how to identify the state of fatigue for drivers fast and accurately is significant to reduce the frequency of traffic accident. Driver’s state of fatigue can be reflected in driver’s facial features; therefore, it is an efficient means that the technique of detection for driver’s facial features determines the state of driver’s fatigue. At present, some organizations have done some research about the detection of driver fatigue. There are also some detection systems under development. While, the accuracy of the detection can not meet the practical needs and indeed more accurate and effective methods are still being developed.Driving behavior is an all-weather behavior. The images from the driver’s video include the images of day and night. Based on the analysis of predecessor’s work, this paper focuses on the driver fatigue detection method for video mainly in night. For the infrared images are grayscale in night, we use the approach to the thesis based on gray image. Above all, gray images are detected through the use Prewitt Edge Detection. In order to solve the problem that a single column pixel has a smaller accumulation, this article uses the window gray integral projection method. We confirm the edge of the face according to the window integral projection of the face edge is relatively large.We propose eye location method for gray image, and target the human eyes preliminary by median filtering and complexity of calculation methods. We employ horizontal and vertical compensate by the difference between two complexity-maximum. According to the prior knowledge of the distance from eye to face midpoint, the vertical boundary of the eyes is ascertained. In the thesis, it is ascertained that the eyes position is by binary horizontal integral projection in the region of the human eyes and the eyes’vertical boundary is by vertical searching. Similarly, we can also use this method to achieve mouth localization accurately.In connection with eye status and mouth status detection method, this thesis applies the principles of blink frequency, yawn frequency and PERCLOS to determine if the driver is in fatigue or if we should give alarm.According to the driver fatigue detection algorithm requirements, we build a driver fatigue detection system by choosing DM642as CPU. We focus on the design of the video capture circuit making the analog signals transformed into digital signals, besides making the video output circuit transformed into digital signals to realize analog signals and the Live View images in real time. Meanwhile, another major module as an audio input and output circuits is applied to receive and send audio data for prompting the warming message. In addition, we design FPGA circuit used to achieve the enhanced video capabilities, generate EMIF buffer control signals, and page flash and configuration information, programs and important data stored in FLASH. Finally, we use PROTEL DXP software for completing the design of PCB board.In conclusion, this paper focus on the night driver fatigue detection algorithms, where a new fatigue detection device is designed for setting the firm foundation to achieve further.