Research On Vehicle Distance Detection System Based On Active Labelled Vision

The development of automobile technology is always accompanied by people's demand for driving safety,at present,the technology of vehicle assisted driving based on photographic camera technology is becoming an important research direction in the field of active safety and intelligent driving.One of the tasks of visual-assisted driving is to extract valid information from the acquired digital images and to give feedback to the driver in real time,therefore the driving operation carried out by the driver can be assisted reasonably.The process of reducing the position relation between other traffic participants and the vehi cle is inseparable from the application of parallax method,and the key of the parallax method is to match the image acquired from different angles.In this paper,the defects in the previous image feature matching are optimized,and the related software and hardware design are studied,and the main research contents are as follows:1)The complexity of the current feature matching algorithm in the embedded system is analyzed combined with the mathematical model of the binocular vision reduction.The method of laser assisted with narrow band filter for digital image acquisition and processing is proposed thus the current feature matching process is simplified as threshold segmentation of digital image,and the design,simulation and experiment of SPL_PL90₃ semiconductor laser drive circuit are completed.2)According to the functional requirements of visual assistant driving,the overall design plan of binocular vision system is developed based on STM32 F series microcontroller and OV2640 camera module.The hardware circuit design of each sub module of the image acquisition system is completed,and the control algorithm and corresponding driving code are designed for each module.3)The common noise model in the digital image and the common noise reduction principle are analyzed.According to the special application of the laser point and the filter in the image acquisition and the requirement of the image processing,the image denoising algorithm is designed,and the rationality of the noise reduction algorithm used in the built visual system is verified by the experiment.4)The key factors affecting the gray distribution of the image are analyzed by the imaging principle of the image sensor.The threshold estimation algorithm based on natural light is creatively proposed and the calibration theory of the light intensity to gray mapping relation is established.The feasibility of the contrast experiment is designed to get the threshold value out of the statistical calculation of the image itself.5)After calibrating the basic parameters of the camera on the Matlab calibration toolbox,the dynamic distance detection simulation experiment is designed to verify the practicability of the designed detection system,and the performance evaluation of the system is completed by the data analysis of the experimental results.The experimental results show that the vehicle distance detection system based on active label vision developed in this paper has practical application value and provides a reference for further research and development of visual technology on low cost embedded platform.