The Lane And Obstacle Detective Algorithm For Vision Based Intelligent Vehicle

The visual information is one of the most important information that can be obtained from the traffic environment. The main content of this paper is approaching to road boundary information extraction and recognition by using the machine vision method. Besides, there is a mothed to detection vechiles based on the Haar-like features. By doing this can provide the necessary basis to intelligent vehicles.In this paper, a model of camers has been workd out. Preprocessing the image of lands by the color spaces transform, histogram equlization and other methos.Denosing images using the neighborhood average method. Using the Canny operator to get binary images. In case that there will be much more lines by Hough transform on binary images, so we use the method that groups seeking to find the required maximum to got the boundarys of road. fter the initial lane boundary has been detected, the ROI can be obtained based on the initial lane boundary. Then the Hough transform on the ROI has been taken, the lane boundary can be found out. In this way, the algorithm can speed up computation and improving the accuracy of the lane boundary identification. At the end of this paper the accuracy of the algorithm is analyzed.The road obstacle detection mostly be the detection of other vehicles on the road. There are two methods had been worked out, one of them is based on the features of the front view and back view of a vehicle, the other one is based on training. In the first method, three features of vehicle have been used. The first one is the shadow below the vehicle, and the shadow has been detected using threshold on the gray space. The second feature has been used is the symmetry of a vehicle, a symmetry measure function is used to determined is there a vehicle in the image. The last feature is the taillight of a vehicle, and it can be detective easily in the HSI color space. The second way to detective a vehicle is some kind of training that base on Haar-like features. To get a cascade classifier that has been compensative by a number of weak classifier, the training used a huge number of samples. In the method the image that to be processing is transformed to integral map. After that a series of different detective windows scanning over the whole image, and get loss of regions that containing a vehicle with different size, by the end those small regions have been combined to a big region using a specific algorithm. And the different of those two methods has been analysis.After that, the intelligent vehicle lane detection and obstacle detection system has been worked out preliminary.