1 INTRODUCTION
The extraction and quantification of freeway roadside landscape information are the bottlenecks in the research on its influence on driving psychology and behavior. Compared to the general image, the image of freeway roadside landscape is characteristic for the complexity of composing elements and variability with time and landform, so that the high-precision and high-robustness algorithm for image segmentation is needed. The freeway roadside landscape mainly consists of natural, cultural and ecological roadside landscape. The diversity of the composing elements lead to the following two marked characters: (a) The roadside landscape changes with regions, roadside slope and plant cover are large difference in different regions of China; (b) The roadside landscape changes with time, the color and plant cover are different in the same region in different times. In order to quantitatively research the influence of different roadside landscape on driving behavior, image processing technique is needed to automatically extract the characters of roadside landscape, and the elements of roadside landscape need to be expressed quantitatively, and then the relationship between driving behavior and characters of the roadside landscape can be established. The relationship among the above elements is shown as Fig. 1. Relationship among roadside landscape, image processing and driving behavior