| Summary: | 碩士 === 國立成功大學 === 測量及空間資訊學系碩博士班 === 98 === Label placement is a fundamental and important process and also a hot topic in the field of cartography. The goal of label placement is to clearly represent the labeled objects and thus, must avoid overcrowded labeling and label-label overlap. Thus, the label placement problem has a very high computational complexity. In order to reduce the complexity of this problem, some of researches reduce and fix the search space of labeling and give a prior priority for each position to solve the problem of label overlapping. Although this strategy can conspicuously reduce the computational complexity, the problem of label-label overlap may still occur when the number of labels is very large. In this thesis, a novel label placement approach is introduced to place the selected image labels according to a significance map which combines the importance of all visual elements. The image label placement is regarded as an optimization problem and solved by a genetic algorithm with an objective function. In the proposed genetic algorithm, a label position is encoded as a gene and a non-random approach is adopted to select the initial population for the purpose of speeding up convergence of the iterative process. By utilizing the genetic operators, crossover and mutation, a near optimal result can be obtained. The experimental results show that the proposed approach can reduce label-label overlap when the number of label increases largely, and can generate exceptional label placement results.
|
|---|
