Research On Intelligent Car Trajectory Generation System Based On Monocular Vision

The smart car integrates a series of advanced technologies such as image processing and intelligent control,and is widely used in the field of unmanned driving,rescue and rescue,area search,industrial production,family life and other fields.Since the 19 th century,artificial intelligence has continuously and rapidly developed,and the research on smart cars has also been paid attention to by more and more researchers,and has gradually become an important research direction.For the research of smart car,its trajectory is the most basic and important function of smart car.Trajectory extraction is divided into moving target detection and moving target positioning.Aiming at the realization of moving target detection function,this paper adopts the real-time image processing and returning method;in the realization of moving target positioning function,this paper uses the gopro camera to collect motion data in real time,and the monocular vision plane positioning method is used for positioning processing.The main research content and work focus of this paper are as follows:1)Aiming at the real-time performance of motion information extraction and the accuracy of motion target extraction in monocular vision motion target detection,a real-time motion target detection algorithm combining the inter-frame difference method and the Vibe algorithm is proposed.The algorithm uses the characteristics of the inter-frame difference method to quickly update the background model,ensuring that the smart car target can be separated from the background even when the first frame of the video image is not a background image,and the car image information can be obtained from the moving video image.The comparative analysis of experimental results shows that the algorithm overcomes the shortcomings of traditional target detection algorithms(interframe difference method,background modeling method,etc.),and can quickly extract the motion information of the smart car at different resolutions and different video durations.,Has good realtime performance and image integrity.2)Aiming at the global accuracy of monocular vision moving target positioning,a monocular vision plane similarity positioning algorithm based on function external parameter compensation is proposed.The algorithm uses the rotation matrix and translation matrix of the camera's external parameters to deal with the image distortion problem so as to improve the positioning accuracy when the image edge distortion is large.The comparison of the actual results of the camera height,angle,and the position of the smart car shows that the algorithm can reduce the distance error of the smart car at the edge of the image by 1-2 cm through the compensation of the external parameter function,so that it can be used to clearly capture the smart car's trajectory and quickly update The demand for the three-dimensional location information of the smart car.3)The construction and verification of the intelligent car trajectory system for the moving target detection algorithm and moving target positioning under monocular vision.Usingg c++ and Open CV3.6.0 vision library to build an experimental platform,and establishes a smart car target motion detection mobile positioning system,which mainly includes a smart car real-time motion interface display module,a moving target detection module,a positioning module,and a trajectory reproduction module.The improved target motion detection algorithm and the improved monocular vision positioning algorithm can accurately detect the smart car and be able to locate it fully automatically,which has a good prospect.