| Summary: | 碩士 === 國立彰化師範大學 === 機電工程學系 === 106 === Currently most of the dashboard camera only record and save the image of driving process. A few of them have the traffic safety early warning functions which include detecting the distance of the leading car, the road offset and so on. In view of the fact that it is easy to cause traffic accidents such as crashing, scratching and other traffic accidents, resulting in irreparable results. The image recognition combined with depth learning to establish a smart vehicle safety image recognition system is developed in this paper.
The purpose of this paper is to develop an intelligent safety driving assistance system using real-time images from the dashboard camera to achieve early warning due to dangerous driving of the leading car driver. Firstly, the image is transformed to grayscale type and set the detection area to reduce the computation time. Then, the lane line is marked using the canny method and the Hough transform algorithm.
The vehicle save driving is identified by using the pixel amount and the YOLO (You Only Look Once, YOLO) neural network to compare the speed and accuracy of the two methods. Testing the effect of lane and vehicle identification in different environments, and trigger the warning according to the conditions of the set vehicle distance and lane offset. Finally, we use the Raspberry Pi and the neural joystick to test the detection speed of the neural network and build an embedded system. The initial structure.
|
|---|
