Abstract. BACKGROUND: Nowadays, more and more traffic accidents occur because of driver fatigue. OBJECTIVE: In order to reduce and prevent it, in this study, a calculation method using PERCLOS (percentage of eye closure time) parameter characteristics based on machine vision was developed. It determined whether a driver's eyes were in a fatigue state according to the PERCLOS value. METHODS: The overall workflow solutions included face detection and tracking, detection and location of the human eye, human eye tracking, eye state recognition, and driver fatigue testing. The key aspects of the detection system incorporated the detection and location of human eyes and driver fatigue testing. The simplified method of measuring the PERCLOS value of the driver was to calculate the ratio of the eyes being open and closed with the total number of frames for a given period. RESULTS: If the eyes were closed more than the set threshold in the total number of frames, the system would alert the driver. CONCLUSION: Through many experiments, it was shown that besides the simple detection algorithm, the rapid computing speed, and the high detection and recognition accuracies of the system, the system was demonstrated to be in accord with the real-time requirements of a driver fatigue detection system.

Abstract. In the AGV system, path planning is one of the key problems. In order to improve the efficiency of the task, the paper presents a method of AGV localization and path planning aiming at the existing algorithms. The system structure of the AGV is first designed, including hardware part and software part. The AGV is then localized based on non-contact RFID in real time. The location information, speed information, communication status, video information and other information are sent to the server and monitored remotely. An improved A* algorithm is proposed to achieve real-time path planning of multi-AGV based on the traditional A* algorithm. The simulation proves that the improved A* algorithm performs more efficiently than the traditional A* algorithm, and time-consuming of the improved A* algorithm is reduced dramatically by using the proposed algorithm. Finally, the paper establishes a whole software operation platform of the AGV system, including localization, wireless control and path planning.

This paper presents a method of mobile robot navigation based on the Kinect. The proposed method has advantages of flexibility and low cost compared with the traditional sensors, such as laser scanner. Using the Kinect depth information, the SLAM and path planning are realized and the mobile robot navigation is studied. Simulation is performed on two-wheeled mobile robot to verify the effectiveness of the proposed method.