4 Change in parameter values of energy function
By changing the parameters of the energy equation of SNAKE, more appropriate values or range of values are searched for so that SNAKE can fit the feature better.

For the cases where the fitting failed by the method of Chapter 2 and Chapter 3, the values of parameter ?1 and ?2 of the energy equation (equation 1) are changed. By monitoring the relation of these parameter and the movement of SNAKE in changing the parameter values, a suitable value is examined for the fitting. Moreover, the value of the parameter obtained for this research is inputted to the energy equation again to see the behavior of Snake for the building cases in Chapter 2. Detailed results will be reported in the conference.

5 Conclusions
In this research, it is examined through numerical experiments what conditions are preferable for the success and can cause the failure of the fitting taking the building model and using the high-resolution satellite image as examples.

The conditions are summarized as follows;
·Accurate initial values (map data).
·Clear edge image.
Especially, it is understood that the result of the fitting changes greatly depending on the initial values. The presence of a strong peripheral edge is important for the success of the fitting, too. However, there is room for conducting more quantitative analysis on the influences of the initial values. These results suggest the following procedure of applying SNAKE for revising spatial database using aerial or space images.

First of all, the model fitting is done using the out-of-dated digital map; The initial value is inputted for the one not having succeeded, and the fitting of the model can be tried again; For the one failed again, detailed vector data is made manually; It is thought that an automation and efficiency improvement can be achieved by depending on a series of work like these for the extraction of the building outline.

If an appropriate edge map etc. are made, updating the existing residential maps data will become easier, though it is still problematic to find the building which should be renewed.

References

• MICHAEL KASS, ANDREW WITKIN, and DEMETRI TERZOPOULOS, 1988. Snakes (Active Contour Models). International Journal of Computer Vision, pp321-331.