|Title||A comparison of three feature selection methods for object-based classification of sub-decimeter resolution UltraCam-L imagery|
|Publication Type||Journal Article|
|Year of Publication||2012|
|Authors||Laliberte, Andrea S., Browning D.M., Rango A.|
|Journal||International Journal of Applied Earth Observation and Geoinformation|
|ARIS Log Number||258448|
|Keywords||aerial imagery, classification, feature selection, object-based image analysis, very high resolution|
The availability of numerous spectral, spatial, and contextual features with object-based image analysis (OBIA) renders the selection of optimal features a time consuming and subjective process. While several feature election methods have been used in conjunction with OBIA, a robust comparison of the utility and efficiency of approaches would facilitate broader and more effective implementation. In this study, we evaluated three feature selection methods, 1) Jeffreys-Matusita distance (JM), 2) classification tree analysis (CTA), and 3) feature space optimization (FSO) for object-based vegetation classifications with sub-decimeter digital aerial imagery in arid rangelands of the southwestern U.S. We assessed strengths, weaknesses, and best uses for each method using the criteria of ease of use, ability to rank and/or reduce input features, and classification accuracies. For the five sites tested, JM resulted in the highest overall classification accuracies for three sites, while CTA yielded highest accuracies for two sites. FSO resulted in the lowest accuracies. CTA offered ease of use and ability to rank and reduce features, while JM had the advantage of assessing class separation distances. FSO allowed for determining features relatively quickly, because it operates within the OBIA software used in this analysis (eCognition). However, the feature ranking in FSO is not transparent and accuracies were relatively low. While all methods offered an objective approach for determining suitable features for classifications of sub-decimeter resolution aerial imagery, we concluded that CTA was best suited for this particular application. We explore the limitations, assumptions, and appropriate uses for this and other datasets.