Comparison of multiple filtering techniques on ALOS PALSAR image for detection of agricultural land abandonment

• Main Author: Md Yunos, Siti Hendon
• Format: Project Paper Report
• Language: English
• Published: 2015
• Online Access: http://psasir.upm.edu.my/id/eprint/90142/1/FP%202015%20147%20-%20IR.pdf; http://psasir.upm.edu.my/id/eprint/90142/

Land use is related to human activity or economic function in a specific piece of land such as for residential, recreational, industrialization and agriculture. Land use for agricultural activities are important to ensure sufficient food supply for human population that increases every day. Discrimination of agricultural land use abandonment essential to study the conversion of agricultural land to other land use activities. Besides, it is crucial to differentiate among the vegetation crops in agricultural land because each crop plays different economic and social values. In Malaysia, Department of Agriculture produces land use classification maps for every two years by using soil survey, optical satellite image interpretation, digitizing and ground verification through the optical satellite imagery. However, optical satellite images are usually suffered by cloud coverage that may decrease the accuracy of satellite image if atmospheric condition is not being eliminated from the image data. The cloud covers problem, however, can be reduced by using SAR image since it is not affected by atmospheric condition. SAR image can be more efficient after having filtering technique applied since it can enhance textural effect of various agricultural lands. Therefore, research objectives of this study are to determine which filtering technique and window size that results in the best discrimination of agricultural land use and abandonment from ALOS PALSAR images. The study area is located in the Sungai Siput, Perak Darul Ridzuan. Standard filtering techniques such as Bit Error Filter, Frost Filter, Gamma Filter, Kuan Filter, Lee Filter, and Local Sigma Filter were performed in ENVI5 software with different window sizes (3x3, 5x5 and 7x7) prior to discrimination of agricultural land use and abandonment. From the study, the best filtering techniques to identify paddy land use and abandonment was Bit Error (3x3, 5x5 and 7x7 window sizes). Meanwhile, rubber area identification would be more effective after the images being filtered with Lee and Local Sigma (5x5 and 7x7 window sizes). On the other hand, the best filtering techniques used for identification of oil palm land use and abandonment was Gamma (7x7 window size).