Research on image processing algorithm of intelligent car with visual navigation

Abstract

This paper, with the background of Freescale's intelligent car using a camera as a path detection sensor to scan the path in advance, process the image got from the camera loaded on the intelligent car. First, when the image contrast is not obvious because of lack of light etc all reasons, this paper adopts sub-linear transformation to process the image. For median filter's treatment failure in case of points located on edge of image, this paper proposed an median filtering algorithm with T-shaped window, form our results obtained form experiments, we conclude that this filtering algorithm can not only filter out the noise of edge, but also more conductive to filter the whole image making the edge of image more continuous and smooth. For image segmentation, reference to Sobel operator's high computation speed and Roberts operator's high accuracy, this paper proposed an improved edge detection fusion algorithm between these two operators, this fusion algorithm is better suited to the image division. Test results show that, the amount of computation of this detection algorithm is much less than the Sobel operator, meanwhile, isolated noise can be removed more effectively using this algorithm.