Gatha Cognition®
Perception, Learning and Reasoning

Article Title :

A Multi-Scale Feature Extraction Approach to Improve Land Use / Land Cover Classification Accuracy using IRS LISS-IV Imagery

Remote Sensing of Land

1

1

3-17

Segmentation , Onscreen Knowledge , Multi-scale Feature Extraction , Object-based classification , Land use/Land cover , IRS LISS-IV

Gatha Cognition Free Publication
Cross Referance

The study presents an approach to map Land Use / Land Cover Change (LULCC) at large scale and processing techniques that permit higher accuracy. IRS RESOURCESAT-2 LISS-IV images of Nellore district of Andhra Pradesh were used to apply the classification technique. In multi-scale feature extraction approach LULCC takes two forms i.e. conversion from one category of LULCC to another and modification of condition within a category. Thus, major LULCC classes were extracted using object based approach and uncertain classes were identified using onscreen knowledge based method. The results showed in 2009, the accuracy of cropland, water body and built-up segments were 99.3%, 94.79% and 89.72%, respectively, whereas, in 2013 the accuracies were 94.31%, 88.26% and 81.20%, respectively. Hence, this classification approach can be useful in different landscape structure over the time, which can be quantified and assessed to achieve a better understanding of the land cover.

National and Global scale LULC mapping demands functional, organized, economic and operational approaches.

IRS LISS-IV data supports object based classification and segmentation at several scales.

The combination of object based and onscreen techniques is promising to improve classification of remotely sensed data.

About 43% of TGA in the study area classified as cropland, scrubland, aquaculture and industrial zone is dynamic in nature.

Cropland, industrial zone and built up area considerably increases whereas fallow land, aquaculture and scrubland decreases.

Recommend this Article