Real-time vehicle detection in highway based on improved Adaboost and image segmentation

With the development of road transportation and the automotive market, the frequent traffic accidents have been paid more attention. Therefore, the safety driving assistant system has gradually become one hot field in recent years. For the collision-warning problem, this paper adapts the multi-dimensional Haar-like features, as well as the Adaboost algorithm, to implement training of the cascade classifier, which will achieve the reliable vehicle detection. On-board monocular camera is used to obtain the images. At the same time, the color space division is achieved by using the self-adaptive sky segmentation algorithm. Then, the space of lane is separated with other areas by using the lane-edge detection algorithm. As a result, the detection area can be reduced. Therefore, the rate of vehicle detection can be further improved. Finally multi-scale sub-window will be used to scan the image parallelly. This can improve the detection efficiency greatly. Experimental result shows that, compared with traditional Adaboost method, the proposed algorithms can effectively improve the accuracy and efficiency of the vehicle detection.