With the development of remote sensing and aero photography, we can quickly get all kinds of inexpensive image with
high resolution. To efficiently manage these increasing high-volume data, the spatial database management system is the
best solution. In this paper many problems and key techniques are analyzed and discussed on establishing remotely
sensed image database, including image dividing, image encoding, image indexing, establishment of image pyramid, etc.
The paper describes an extended arc data structure, it is especially suitable in some vector-based operations. The existing
vectorial arc data structure can store and organize coordinate information effectively, but some vector-based calculations
are inefficient. Enlightened by the idea of spatial index, the paper proposed to divide arc into subsections, obtain spatial
limit information of each segment and build up to arc data structure. The paper also gives realizing method for this modified structure, analyses time efficiency and space efficiency in some frequently used procedures. Taking the computation of the distance between point and arc for example, this paper analyses the relationship between computational efficiency and sub-arc node number, gives practical results using artificial and real arcs from a GIS environment. Tests results indicated that it would improve the efficiency of vector operations, and accelerate some vector-based algorithm practical indirectly with the extended data structure.
Delaunay triangulation is always used to construct TIN, and is also widely applied in manifold fields, for it can avoid
long and skinny triangles resulting in a nice looking map. A wide variety of algorithms have been proposed to construct
delaunay triangulation, such as divide-and-conquer, incremental insertion, trangulation growth, and so on. The
compound algorithm is also researched to construct delaunay triangulation, and prevalently it is mainly based on divide-and-conquer and incremental insertion algorithms. This paper simply reviews and assesses sweepline and divide-and-conquer
algorithms, based on which a new compound algorithm is provided after studying the sweepline algorithm
seriously. To start with, this new compound algorithm divides a set of points into several grid tiles with different
dividing methods by divide-and-conquer algorithm, and then constructs subnet in each grid tile by sweepline algorithm.
Finally these subnets are recursively merged into a whole delaunay triangulation with a simplified efficient LOP
algorithm. Because topological structure is important to temporal and spatial efficiency of this algorithm, we only store
data about vertex and triangle, thus edge is impliedly expressed by two adjacent triangles. In order to fit two subnets
merging better, we optimize some data structure of sweepline algorithm. For instance, frontline and baseline of
triangulation are integrated into one line, and four pointers point to where maximum and minimum of x axis and y axis
are in this outline. The test shows that this new compound algorithm has better efficiency, stability and robustness than
divide-and-conquer and sweepline algorithms. Especially if we find the right dividing method reply to different
circumstance,its superiority is remarkable.