A Road Segmentation and Road Type Identification Approach Based on New-Type Histogram Calculation

By analyzing the characteristics of off-road images, a new and computationally efficient algorithm of road segmentation and road type identification for Autonomous Land Vehicle (ALV) navigation system based on the proposed new-type histogram calculation was established. The new-type histogram is a new image, not a curve. It is also directional. It makes a statistic of the pixels under some special direction. With the histogram images, each pixel in the original image is given a threshold separately, thus the road images are correctly segmented. Moreover, the new-type histogram image could be used to identify road type. It avoids the deficiency that road images are wrongly segmented by reason that the preset road model does not fit in the real scene in the conventional methods. The proposed method does not need any extraction of the relevant information in the image (texture of the road, shadows, road edges, etc.). The method is evaluated on thousands of the cross-country road images under various lighting conditions. The experimental results demonstrate the method's accuracy, feasibility and robustness. It increases the segmentation accuracy to an extent. This would be a potentially significant contribution to the active area of road segmentation in the ALV navigation system.