The Rural Road Extraction Method Research Based On High Resolution Images

New-type urbanization is a new sustainable development strategy for the current problems of the evolution of the rural to the urban, in which the development of rural road traffic plays a major role in decision-making for subsequent development. Currently updating road database mainly relies on two methods, one is to mark roads manually on remote sensing images, and the other is to use hand-held GPS to get field road data directly, the two of which are a waste of human resources and time. Therefore, it is necessary to give an efficient road extraction algorithm to update road database.This paper proposes a new object-oriented region-growing method and uses the other six different methods for four different road conditions to do research based on high-resolution images of UAV 0.2m in experimental areas of Changsha City, Hunan Province.In this paper, the object-oriented region-growing method, based on collection of homogeneous cells, firstly uses spectral and spatial information to select seed objects, and then sets growing conditions to extract information in unit of object. Compared to traditional information extraction methods resulting in break and incompleteness, the method uses the concept of object to eliminate the influence of salt-and-pepper noise and get a complete extraction result, while providing a new spatial context condition for the object-oriented approach.The object-oriented region-growing method includes two steps, one of which is to use object-oriented method to complete image segmentation in order to obtain road seed objects, and the other is regional growth based on the objects. In segmentation section, use multiresolution segmentation method for image segmentation, set extraction conditions of road seed objects based on spectral and special features, and make basic region-growing image based on mean brightness values of objects. In the part of regional growth, firstly separate road seed objects gotten from object-oriented approach in order to get independent objects by using edge detection and mathematical morphology algorithm like expansion and corrosion. Secondly, based on basic region-growing image, calculate the brightness value difference between every object in basic region-growing image and road seed objects. Treating the difference as region-growing conditions, if difference is less than the setting threshold, the object is road, while if it is greater than the setting threshold, the object is background. Finally, after complete the second step getting a binary image, use mathematical morphology reconstruction method to obtain the final result of extraction roads. Experiment shows that compared with the traditional methods, the object-oriented region-growing method can extract more complete roads.Additionally, this paper makes a comparative analysis of qualitative and quantitative in terms of supervised classification, unsupervised classification, regional separation and merger, edge detection, object-oriented method, regional growth, and the object-oriented region-growing method.Due to complex road conditions, this paper divides the road experimental environment into four types. One includes roads and vegetation, two includes roads with shadow and vegetation, three includes roads, houses and vegetation, and the last includes roads with shadow, houses and vegetation. The research results that the seven kinds of methods have their own advantages, but the new object-oriented region-growing method proposed in this paper can give better road extraction results in different conditions.