An
Automated Selection Model of Ditch Based on Multi-objective
Optimization by Genetic
Algorithm
R. Zhai
Institute
of Surveying and Mapping, Information Engineering University, Zhengzhou, China
chxy_zrj@163.com
Abstract: It’s a key and
difficulty to consider the self-characteristic and keep the distributing characteristic of ditch
that is a structural selection, which is multi-objective optimization actually.
Combining the basic generalization principle of ditch, considering quality
characteristic of single ditch and selection results of ditch holistically, and
keeping the distributing characteristic, an automated selection model of ditch
based on genetic multi-objective optimization is developed.
As a network, ditch is denseness, which is described by route system model, entity of ditch is represented by a whole route, and
classified on modality, intersection types of which are concluded into: cross(╋), left intersection(┫), right intersection(┣), concept of ditch relation is proposed.
Knowledge rules of ditch selection are summarized. Evaluation of ditch about
importance includes structure and quality characteristic, when selecting,
importance of quality characteristic is represented by knowledge rules and
constraint in genetic algorithm, which is used to evaluate the reservation of
structure characteristic. The model adopts layered method in selection, which
combines the quality characteristic attribute of ditch first, selects ditches
according to knowledge rules initially, therefore, necessary and deleted
ditches are distinguished; remaining ditches is selected by genetic algorithm
structurally, according to distributing characteristic of ditch, combining
constraint code and M kinds of selection case created by local searching strategy,
objective function of ditch selection is designed, and proper adapted function
is developed, which is used to evaluate each selection case, iterative
calculation is done continually, genetic algorithm isn’t stopping until evolving
to satisfied the convergence condition, and optimized selection result is got
finally.
At last, data of dense ditches
(scale 1:250,000) is experimented by the model, each factor influencing the
selection of model is compared and analyzed, and three factors influencing
model selection are given, which are the selection sums of ditches, minimum space
between ditches, adding few “primary” ditches in initial selection. Experiments
show that the model is not only considering the ditches’ quality
characteristics, but also keeping graphical distribution of ditches. The model
is suitable for varieties of ditches, with certain intelligence.
Key
Words: automatic map generalization;
Genetic Algorithm; ditch network; ditch selection