An Artificial Neural Network Combined to Object Oriented Method for Land Cover Classification of High Resolution RGB Remote Sensing Images

Autor:	Sohaib Baroud, Zineb Hidila, Soumia Chokri, Mohammed Mestari, Safa Belhaous
Rok vydání:	2020
Předmět:	Set (abstract data type) Similarity (geometry) Artificial neural network business.industry Computer science Perspective (graphical) RGB color model Segmentation Pattern recognition Artificial intelligence Land cover Object (computer science) business
Zdroj:	Communications in Computer and Information Science ISBN: 9783030451820 SADASC
DOI:	10.1007/978-3-030-45183-7_17
Popis:	Land cover is the observed physical cover on the earth’s surface. The complexity and similarity of the elements that occupy the earth’s surface make the task of interpreting aerial and satellite images very tedious, especially when it comes to very high-resolution RGB images. A neural network can be used to learn from a large amount of imagery data and is useful at classifying semantic features of imagery data. Coupled with the object-oriented approach that enriches the characteristics of the images beyond the spectral data, the abilities of ANN can be significantly improved. This paper aims to evaluate the application of ANN combined to Object-Oriented Method for the classification of a high resolution (20 cm) RGB aerial images, we took the Moulay Bouselham Region in Morocco as example. First a quad-tree then multi-resolution algorithms are used to generate homogeneous objects. Second, a set of object’s features is generated on the basis of segmentation results. Third, an Artificial neural network is tuned to rich the maximum classification performance. Finally, we discuss the results and elucidate perspective of this study.
Databáze:	OpenAIRE
Externí odkaz:	https://explore.openaire.eu/search/publication?articleId=doi_________::cd75684611128738fff5683aa6e2cc78 https://doi.org/10.1007/978-3-030-45183-7_17 Zobrazit plný text záznamu