Abstract | ||
---|---|---|
Most image segmentations require control parameters setting that depends on the variability of processed images characteristics. This paper introduces a meta-learning system using stacked generalization to adjust segmentation parameters within an object-based analysis of very high resolution urban satellite images. The starting point of our system is the construction of the knowledge database from the concatenation of images characterization and their correct segmentation parameters. Meta-knowledge database is then built from the integration of base-learners performance evaluated by cross-validation. It will allow knowledge transfer to second-level learning and the generation of the meta-classifier that will predict new image segmentation parameters. An experimental study on a satellite image covering the urban area of Strasbourg region enabled us to evaluate the effectiveness of the adopted approach. |
Year | DOI | Venue |
---|---|---|
2014 | 10.1007/978-3-319-11758-4_21 | IMAGE ANALYSIS AND RECOGNITION, ICIAR 2014, PT I |
Keywords | DocType | Volume |
Object-based analysis, Segmentation, Very high resolution satellite image, Meta-learning, Stacked generalization | Conference | 8814 |
ISSN | Citations | PageRank |
0302-9743 | 0 | 0.34 |
References | Authors | |
11 | 3 |
Name | Order | Citations | PageRank |
---|---|---|---|
Aymen Sellaouti | 1 | 0 | 0.34 |
Yasmina Jaâfra | 2 | 0 | 0.34 |
Atef Hamouda | 3 | 40 | 12.57 |