Safety Assistant Driving System Based On Computer Vision

Safety Assistant Driving System is an important part of the construction of Intelligent Transport Systems and the Intelligent City Systems,which has significant meaning to reduce people injury,reduce traffic accidents and improve driver's driving experience.In recent years,Computer vision technology is widely used in Safety Assistant Driving System in the field of driverless vehicle and advanced commercial vehicle because of its advantages of low cost of hardware and similar to human vision.In order to provide more comprehensive,timely and reliable road environment information for Safety Assistant Driving System,many key technical issues,such as vehicle and pedestrian recognition,vehicle logo and license plate recognition,camera calibration and ultrasonic distance measurement based on servo control.Then some solutions based on target recognition and ultrasonic distance measurement are proposed.The main research contents and achievements are summarized as follows:(1)Vehicle and pedestrian recognition based on deep learning is deeply discussed.The model is trained on the dataset collected by myself,so that the trained model can be applied to this particular field of target recognition.The experimental results show a high performance both in accuracy and speed compared with other traditional target recognition methods.(2)Vehicle logo and license plate recognition based on cascaded convolutional neural network is presented.Both networks obtain candidate regions by selective search,and use prior knowledge to exclude most of the candidate regions,and then use multilayer convolutional neural network to determine whether the target is a vehicle logo or license plate.The experimental results show a high performance in accuracy when using cascaded convolutional neural network to recognize vehicle logo and license plate.(3)The ultrasonic distance measurement based on servo control is deeply discussed.Using the imaging principle of camera to establish a geometric model centered on a 2-DOF PTZ,according to the position of the target to be measured,the angle information that the PTZ needs to rotate is deduced.Then the distance-measuring system controls the servo to turn the PTZ to the target position,and finally activates the ultrasonic distance measuring device to obtain distance information.(4)The monocular distance measurement based on servo control is deeply discussed.Establishing a camera-centered geometric model using the imaging principle of camera,according to the position of the target to be measured,the angle information that the PTZ needs to rotate is deduced.According to the geometric model,the monocular distance-measuring formula is analyzed and deduced,camera calibration method is adopted to obtain internal parameter,which is then brought into the fomula for distance-measuring.Then the distance-measuring system controls the servo to turn the PTZ to the target position,and finally activates the monocular distance measurement method to obtain distance information.