A Study On The Rearview Monitor System Based On Fisheye Lens

With people’s living standards improve, the car retains the quantity is increasing year byyear. Each year due to traffic accidents caused casualties and property losses are greatlyincreased. The safety performance of automobile more and more attention, the trafficsafety problems are more and more attention. Accordingly, also emerge in an endless stream of newtechnology on traffic safety. The vehicle monitoring system is an important part of automobilesafety problem.The fish eye lens has the advantages of large visual angle, very suitable for vehicle visualsurveillance camera. If you don’t use a fisheye lens, need at least two more ordinary lens canachieve the view without dead angle coverage. Rear traditionalmonitoring systemsare passive monitoring, is only a image acquisition, cannot take the initiative to find theabnormal target and alarm. In order to meet the needs of intelligent vehicle rear-view monitoringrequirements, this topic using fish eye lens,with the technology of image processing, computervision based, on the vehicle rearview systemconducted in-depth research, realized the detection ofmoving targets in the process of reversing.In this paper, the main work and achievements are as follows:1. We analyzedthe characteristicsof the fisheye lens and the vehicle videomonitoring development to illustrate that the fisheye lens usein car rearview system are rational andvalid.2. We described the imaging principle and imaging model of fisheye lens, the realization of theordinaryfisheye image and180degree fisheye image correction to be normal images. Then, putforward a kind of chessboard corner detection algorithm which adapting to large distortion image.This algorithm is also automated. What’s more, it significantly improves the accuracy of corners aswell as the effect of fisheye image correction.3. We established background subtraction algorithm model, aiming at the shortcomings ofViBe algorithm, we also proposed an improved ViBe algorithm with which combinedforeground pixel statistics in order to revise the defect of foreground pixels cannot updatebackground model. The improved ViBe algorithm, combined with morphological computing andContour finding, can achieve the moving target detection fairly well in complex background,especially in the moving background. We also transplant this algorithm to Android platform wherethe performance is quite low, and find out it works fairly well.4. We also discussed the image stitching, finally realized the fisheye image mosaic with Harris corner and SURF feature points.The experimental results show that the fisheye lens can be corrected well enough to be appliedinto the vehicle rearview monitoring system. In moving object detection, with a very good detectioneffect, and low performance requirements, We meet the requirements of embedded systems, and hastheoreticalbasis and practical significance.