Research On Vehicle Driving Assistance Technology Based On Panoramic Vision

With the improvement of the level of automobile intellectualization,the problem of driving safety is becoming more and more noticeable.Therefore,advanced driver assistance systems appeared.Visual-based driving assistant technology is mainly based on vision and perceives the surrounding environment of vehicles through visual sensors installed around the body.For example,360-degree panoramic vision system uses the ultra-wide-angle fisheye lens installed around the car to obtain the images around the car body and stitchs them by a series of image processing algorithms.Finally,a panoramic view image is formed to eliminate the driving blind area,which also reduces the frequency of traffic accidents.Since 360-degree panoramic view system of the car can obtain a large amount of environmental information around the body,the image information obtained by the 360-degree panoramic view system of the car is fully utilized in this paper.And then we study the lane departure warning and parking assistance technology on the basis of panoramic vision.And the algorithms involved are implemented in detail.The main research contents are as follows:(1)In this paper,the images are acquired through the fish-eye cameras,and the spherical perspective projection algorithm is used for distortion correction.Through the study of the inverse perspective mapping algorithm,the algorithm of inverse perspective mapping based on the calibration point is improved,which reduces the distortion of the image edge after the inverse perspective transformation.(2)Through the research on the algorithm of image stitching,an image mosaic algorithm that based on the idea of clipping-stitching is proposed,which depends on vehicle driving assistance technology based on panoramic vision.And the weighted average algorithm is further used to fuse the stitching areas generated by stitching,which can eliminate obvious gaps.Finally,a clear 360。 panoramic image could be formed.(3)The lane departure warning technology is fused on the basis of panoramic images in this paper,and the lane departure warning module is analyzed and studied.An lane departure warning strategy based on fusion of fuzzy Kalman filter and fuzzy TLC(time to lane crossing)are proposed,which can effectively and timely generate multi-level warning when the car is about to deviate from the lane lines.(4)On the basis of the panoramic image,parking assistant technology is further integrated.And then,the algorithm of parking space recognition is studied detailly,which is divided into two steps: rough extraction and fine matching.And rotation angle matching is increased.In addition,the types of parking spaces are identified by the corner points of parking spaces by the improved traditional template matching algorithm.Finally,the parking entrance position and parking direction could be determined.Experiments show that the driving assistance technology based on panoramic vision proposed in this paper can make full use of the information obtained from panoramic images and integrate lane departure warning and parking assistance technologies friendly.At the same time,the algorithm improvement and optimization proposed in this paper can effectively improve the computational accuracy and operational efficiency of the algorithm,which could make good experiment effects.