Resource information
Vegetation mapping provides basic information for forest management and planning. In remote sensing research, the process of creating an accurate vegetation map is an important subject. Recently, there has been growing research interest in the object-oriented image classification techniques. The object-oriented image classification consists of multi-dimensional features including object features and thus requires multi-dimensional image classification approaches. For example, a linear model such as the maximum likelihood method of pixel-based classification cannot characterize the patterns or relations of multi-dimensional data. In multi-dimensional image classification, data mining and ensemble learning have been shown to increase accuracy and flexibility. This study examined the use of the object-oriented image classification by Random Forest classification for vegetation mapping. Vegetation maps were also created using the Nearest Neighbor method and the Classification and Regression Tree method for comparison of classification accuracy. The study area was Sado Island in Niigata Prefecture, Japan. SPOT/HRG imagery (June 2007) was used and classified into the following seven classes: broad-leaved deciduous forest, Japanese cedar, Japanese red pine, bamboo forest, paddy field, urban/road, and bare land. We employed eCognition software for the object-oriented image classification. We selected 18 object features: the mean, standard deviation, ratio, shape, length, and compactness for each band and normalized difference vegetation index value. High accuracy was found for the vegetation maps produced using the Random Forest and Nearest Neighbor methods. The accuracies of these two methods were significantly different from that of the Classification and Regression Tree method, as shown by kappa analysis. Among the three techniques, the Random Forest method showed the highest classification accuracies when class accuracies such as user's accuracy and producer's accuracy were considered. This study demonstrates that the Random Forest classification is effective for vegetation mapping by multi-dimensional image classification.
