Zhu Zhongxiang

Research on Information Fusion on Evaluation of Driver Fatigue

Multi-source information fusion was introduced for the evaluation on driver fatigue which is divided into sub-systematic and systematic evaluation by integrating information from visual cues in common use and steering wheel behavior and vehicle¿s trajectory information. Neural network is combined with Dempster-Shafer evidence theory to finish character fusion and decision fusion. In the phrase of fusion, analytic hierarchy process (AHP) is applied to determine the basic weight, at the same time, considering the impact of data reliability on weight, the nonstatic weight is adapted the system to the current situation by the combination data reliability with basic weight. According to the simulation experiment on the drive simulator, compared with the single index, the adaptation of the evaluation method to monitor driver fatigue was more accurate, reliable and robust.

Research on Recognition Methods for Traffic Signs

In order to insure safety of driving vehicle in the driver assistance system and effective navigation in the automatic driving system, a new recognition method based on the neural network and the invariant moments for traffic signs was proposed in this paper. Firstly, the area of the traffic sign was located from the complicated image background. Secondly, the features of each traffic sign were obtained by Hu invariant moments. Finally, the traffic signs were recognized by the BP neural network. Experimental results proved that the recognition method can be used for fast and efficient recognition of the traffic signs with high accuracy. This research can provide technological supports to the autonomous vehicles and the vehicle safety driving assistance.

Real-Time Dynamic Visualization of Wheel Ruts Based on Vega Prime

For now, the research and implementation of dynamic terrain are all based on OpenGL. In order to realize virtual field experiments of agro-machinery in the Vega Prime virtual reality platform, a real-time visualization method of wheel ruts based on class vrGeometry in Vega Prime was proposed. Firstly, the small three-dimensional terrain is established in Multi gen Creator, the 3d terrain models of variable points are redrawn and generated and the terrain models are generated and loaded quickly by using multi threading technology. Then the collision points between wheels and the ground are obtained by collision detection, and Multi-Thread technology is used to search the vertices on the terrain near collision points synchronously and quickly, then the real-time dynamic wheel ruts are produced by modifying coordinates of the vertices and updating the terrain data real-time.

Study on virtual experiment methods of agricultural equipment based on Vega Prime

The design and development process of virtual experiment was analyzed in this paper. Then a virtual experiment system for agricultural equipment was founded by using MultiGen Creator, Vega Prime software and VC++ programming language. The system running process was introduced, together with tractor 3D modeling, operation data acquisition, dynamic modeling and drive, and collision detection. Furthermore, visual drive, synchronous display, tractor dynamics solution in network and other key techniques related to virtual experiment were discussed as well. The system provides a new method to tractor performance experiment, and the experimental results show stable operation of the system with the reliability and validity.

Experimental Research on Driver Fatigue Based on SOFM

An evaluation model of driver fatigue state was constructed based on the self-organizing feature map (SOFM), in which reaction time, steering angle and vehicle trajectory were considered. 10 males in the same age phrase were recruited to drive under monotonous conditions on self-made driving simulator. The statistic analysis of the test data showed that the three indexes mentioned above were correlated with the subjective evaluation which was performed by the experts according to the recorded videos distinctly. From the experimental results, it was found that the fatigue levels classified by SOFM evaluation model were in accordance with subjective evaluation, and SOFM evaluation model could recognize the levels of driver fatigue state automatically and accurately.