Detection, Based On The Distance Of Vehicles And Vehicles Of The Monocular Vision At Night

In recent years, Intelligent Vehicle System has become a new field whichattracts more attention over the world, where vehicle detection and trackingbased on monocular vision is one of the important research about driverassistance system. Present day, vision-based vehicles detecting algorithmdeveloped rapidly and has been basically put into practical use because oflarge amount of information and more features can be obtained from images.Night driving, inadequate lighting, low visibility, Detection of the day onthe brink of Gray common mutation characteristics and the symmetrical featuresof the vehicle's tail will greatly diminish, or even disappear. So thealgorithm at night phase for the day is still in the exploration,So theresearch of this paper is a good exploration to the algorithm at night.The objective of this research is to develop a vision-based driverassistance system to enhance the driver's safety in the nighttime. Theproposed system performs both preceding vehicle recognition and evaluatingdistance between the own car and the leading vehicle based on monocularcamera. This paper presents a new detecting method based on mathematicalmorphology and color information. Vehicle recognition in the nighttime isachieved by using an evident feature which are extracted through foursteps: taillight standing-out process, adaptive thresholding, centroiddetection with color information and taillight pairing algorithm. Then, themethod of leading vehicle distance evaluating uses only the intrinsicparameters of the camera, off-line extrinsic parameters of the cameracalibration and projectctive geometry, and the rate error of detecting causedby is also ff-line extrinsic parameters of the camera calibration is alsodiscussed in the paper.