| Summary: | 碩士 === 明道大學 === 管理研究所 === 95 === The purpose of this research is to study the application of Artificial Neural Networks for auto-steering vehicles equipped with digital image system. In this research, a multi-layer perceptron was used for the network architecture and the break-propagation algorithm was applied to learn the relationship between the images in front of the vehicles and the control action of the auto-steering system. After being instructed properly, unmanned vehicles are capable of making a turn successfully on the basis of the image in the road ahead As a result, we can achieve the goal-- automatic driving according to road conditions.
In this study, there is a digital video recorder placed on the vehicle to capture the images of the road ahead. First and foremost, the images are changed from color ones to black and white ones and undergone Sobel Edge Detection operator. After we get the images of borders on the sides of the road, their size is transformed as the binary of 8x6 pixels and is outputted. Then, these images are used as the input of Back-propagation Network. Via input, Back-propagation Network learns to make a left turn, to make a right turn, or to go straight. The results of this study indicate that the unmanned vehicle is capable of automatically moving on the real street as long as we connect the smart, visual recognition system with steering control equipped in Ming Dao unmanned vehicle. In our findings of current testing, the fact shows that the unmanned vehicle is able to move straight approximately 45 meters and to make a turn approximately 30 meters.
Chapter 3 and Chapter elaborate how to develop an automatic monitoring system by the technique, digital image process. Chapter 3 deals with the application of the automatic detection system of tire width on the manufacturing process of the tires. Chapter 4 talks about a digital hydrogen gauge of Ming Dao hydrogen-based vehicle.
|
|---|
