Assessment and Comparison of Classification Techniques for Forest Inventory Estimation: A Case Study using IRS-ID Imagery

要旨

Traditional satellite image classifications are mostly confined to supervised, unsupervised and hybrid methods. An alternative to these approaches, subpixel classification is gradually changing the concept of image classification. This technique is advantageous particularly for medium to low resolution satellite image, by removing the influence of associated features within a pixel. The Spectral Mixture Analysis (SMA), is a typical subpixel classifier, and is applied here in forest classification of Gurdaspur, Hosiarpur and Rupnagar districts of Punjab, India to explore its authenticity and accuracy. A special effort was made to accurately calculate the forest inventory in Punjab using the SMA, and the results were compared with the data obtained through traditional classification methods. This technique enables estimation of proportional forest type in a single pixel and may be used to estimate various aspects of forest vegetation important for different forest modeling (forest growth, forest yield etc.), carbon budgeting and decision making.