Feature selection by genetic algorithms in object-based classification of IKONOS imagery for forest mapping in Flanders, Belgium

• Published in: Remote sensing of environment Vol. 110; no. 4; pp. 476 - 487
• Main Authors: Van Coillie, Frieke M.B., Verbeke, Lieven P.C., De Wulf, Robert R.
• Format: Journal Article Conference Proceeding
• Language: English
• Published: New York, NY Elsevier Inc 30.10.2007; Elsevier Science
• Subjects: Animal, plant and microbial ecology; Applied geophysics; Areal geology. Maps; Biological and medical sciences; Classification; Earth sciences; Earth, ocean, space; Exact sciences and technology; Feature selection; Forest mapping; Fundamental and applied biological sciences. Psychology; General aspects. Techniques; Genetic algorithms; Geologic maps, cartography; IKONOS; Internal geophysics; Neural networks; Segmentation; Teledetection and vegetation maps; Belgium; Western Europe; Belgium, Flanders; Feature selection; IKONOS; Segmentation; Neural networks; Classification; Forest mapping; Genetic algorithms; Integration; Europe; Space remote sensing; neural networks; Average; Forest management; accuracy; cartography; classification; Image; forests; Genetic algorithm; Forest map; imagery; segmentation; Plant cover; Added value; strategy; policy
• ISSN: 0034-4257; 1879-0704
• DOI: 10.1016/j.rse.2007.03.020

Obtaining detailed information about the amount of forest cover is an important issue for governmental policy and forest management. This paper presents a new approach to update the Flemish Forest Map using IKONOS imagery. The proposed method is a three-step object-oriented classification routine that involves the integration of 1) image segmentation, 2) feature selection by Genetic Algorithms (GAs) and 3) joint Neural Network (NN) based object-classification. The added value of feature selection and neural network combination is investigated. Results show that, with GA-feature selection, the mean classification accuracy (in terms of Kappa Index of Agreement) is significantly higher ( p < 0.01) than without feature selection. On average, the summed output of 50 networks provided a significantly higher ( p < 0.01) classification accuracy than the mean output of 50 individual networks. Finally, the proposed classification routine yields a significantly higher ( p < 0.01) classification accuracy as compared with a strategy without feature selection and joint network output. In addition, the proposed method showed its potential when few training data were available.