Comparison between object based classfication and pixel based classfication technique to detect deforestation in Ulu Muda forest reserve, Kedah / Shahrul Nizam Jamaludin

Remote sensing is moving toward mapping the Earth surface using the highly
technology implement. The researcher has invented two types of classification that
can be integrated with the modern technology. Those categories of classifications are
pixel based classification and object based classification. Both methods purpose to
analyse forest cover and changes especially deforestation activity but, due to the early
stage of these methods, their abilities to classify land cover and monitor forest
dynamics have not fully evaluated and investigate. Here, the strength for both
methods was studied, to know which one is the best in detecting deforestation at Ulu
Muda Forest Reserve, Kedah. The forest cover at Ulu Muda will be classified, where
pixel based classification was done using the Erdas software while object based
classification completed using the eCognition software. Satellite imagery from SPOT
5 and 6 with size pixel of 12 metre and 7 metre were used in change detection
analysis. The accuracy assessment has been done to identify the overall accuracy of
for both classifications including the user and producer accuracy. The higher value of
that accuracy approaching to 100, the more accurate the classification can be said.
The possible best method of classification in detecting deforestation activity will be
determined and explained more its concept in this study.