Detection of Urban Areas Using Genetic Algorithms and Kohonen Maps on Multispectral Images

Abstract

In this article, the detection of urban areas on satellite multispectral Landsat images. The goal is to improve the visual interpretations of images from remote sensing experts who often remain subjective. Interpretations depend deeply on the quality of segmentation which itself depends on the quality of samples. A remote sensing expert must actually prepare these samples. To enhance the segmentation process, this article proposes to use genetic algorithms to evolve the initial population of samples picked manually and get the most optimal samples. These samples will be used to train the Kohonen maps for further classification of a multispectral satellite image. Results are obtained by injecting genetic algorithms in sampling phase and this paper proves the effectiveness of the proposed approach.

Genetic Algorithm for Artificial Neural Networks in Real-Time Strategy Games

“…Some previous studies have used this algorithm for several purposes. An example: academic scheduling [17], sentiment review analysis of fashion online companies [18], microgrid energy management [19], children activities model [20], operational planning of cement mills loading [21], corporal Portal Search Engines [22], the optimal combination of forest fire [23], heart sound segmentation [24], detection of urban areas [25], short-term solar power forecasting [26], breast cancer [27], mobile robot path [28], Bone Cancer Survivability Prognosis [29], Space-Based Telescopes [30], et cetera. This study aims to adjust the weight on the ANN using a genetic algorithm (GA) to determine the best crossover rate and mutation rate values for troops in RTS games.…”