Research on image segmentation and feature extraction of freeway roadside landscape

Landscape feature extraction and quantification have been a bottleneck problem when studying the relationship between roadside landscape and driving psychology. According to the diversification and detailed texture of roadside landscape unit, three kind methods such as Marker-based watershed algorithm, Canny edge detection algorithm and Texture analysis algorithm were explored for the research of freeway roadside landscape segmentation. The above approaches were tested by using experimental data obtained in Kunyuan freeway in Yunnan province. A contrastive analysis was conducted to identify the applicability of each algorithm and finally the texture algorithm was selected. The RGB color histograms and HSV average values were extracted from the segmented image, correlation between the two above features and the composition and continuity of roadside landscape were analyzed respectively, which can provide a new method to quantify freeway roadside landscape features.