To 3D visualization of terrain data with large ranges, we often use the strategy as follows: terrain model is firstly divided into regular patches and then index mechanism is set up on the basis of data patches. Next, we load batches of data patches around the viewpoint dynamically. However, when many road networks are nested into terrain models, it is difficult for terrain models to divide into regular patches in an original way. It is well known that the data is along the road in a linear way and the roads connect to each other. According to the above feature, a hybrid index mechanism combining connectivity with dynamic segmentation is proposed in this paper, which can realize the process of query in two steps. The first step is to search roughly by means of connectivity index. Based on the results of the first step, the second step can finish the process of exact query satisfactorily using dynamic segmentation index. This index mechanism with high query speed guarantees speed of loading data dynamically, and meets the demand of visualization of road network. Finally, we implement the algorithm by means of some experiments, and give a comparison in space efficiency and time efficiency with R-tree index.