Combination of Image Analysis Techniques for Rice Area EStimation

Abstract

Most of the image analysis techniques have both strong and weak aspects. Classifying the image with clustering algorithm, an unsupervised classification method for example, is a time-consuming process. On the other hand converting into NDVI (Normalized Difference Vegetative Index) and assigning pixels to different features is also a vague process. Using a combination of NDVI and unsupervised method to estimate area under rice can reduce processing time while, at the same time, reinforcing the performances of both classifiers. This paper deals with estimating area under rice in Nalgonda district of Andhra Pradesh. IRS-ID digital data from NISS-III (path 100/60 16 October 2004 and 101/60 14 November 2004) was selected and ERDAS image8.6 package was used for image processing. The NDVI map is derived first and refined to set the threshold level for vegetation. Next, a query was developed to set zero to the non-vegetative features in IRS-ID image. This image was further classified with unsupervised classification using ISODATA for clustering. This study shows that the combination method of NDVI and unsupervised classification appears to be better choice for estimating area under rice.