Assessment of Performance Improvement in Hyperspectral Image Classification Based on Adaptive Expansion of Training Samples
محورهای موضوعی : Image Processing
1 - Tarbiat Modares
کلید واژه: Classification, Hyperspectral Image, Limited Training Data, Pseudo-Training Samples,
چکیده مقاله :
High dimensional images in remote sensing applications allow us to analysis the surface of the earth with more details. A relevant problem for supervised classification of hyperspectral image is the limited availability of labeled training samples, since their collection is generally expensive, difficult and time consuming. In this paper, we propose an adaptive method for improving the classification of hyperspectral images through expansion of training samples size. The represented approach utilizes high-confidence labeled pixels as training samples to re-estimate classifier parameters. Semi-labeled samples are samples whose class labels are determined by GML classifier. Samples whose discriminator function values are large enough are selected in an adaptive process and considered as semi-labeled (pseudo-training) samples added to the training samples to train the classifier sequentially. The results of experiments show that proposed method can solve the limitation of training samples in hyperspectral images and improve the classification performance.
High dimensional images in remote sensing applications allow us to analysis the surface of the earth with more details. A relevant problem for supervised classification of hyperspectral image is the limited availability of labeled training samples, since their collection is generally expensive, difficult and time consuming. In this paper, we propose an adaptive method for improving the classification of hyperspectral images through expansion of training samples size. The represented approach utilizes high-confidence labeled pixels as training samples to re-estimate classifier parameters. Semi-labeled samples are samples whose class labels are determined by GML classifier. Samples whose discriminator function values are large enough are selected in an adaptive process and considered as semi-labeled (pseudo-training) samples added to the training samples to train the classifier sequentially. The results of experiments show that proposed method can solve the limitation of training samples in hyperspectral images and improve the classification performance.
