Object-based crop pattern detection from IKONOS satellite images in agricultural areas


Creative Commons License

TAVUS B., Karatas K., TÜRKER M.

PAMUKKALE UNIVERSITY JOURNAL OF ENGINEERING SCIENCES-PAMUKKALE UNIVERSITESI MUHENDISLIK BILIMLERI DERGISI, cilt.25, sa.5, ss.603-614, 2019 (ESCI) identifier

Özet

Nowadays, with the development of remote sensing technologies and image processing methods, satellite images have become frequently preferred in studies to determine the crop pattern in agricultural areas. In this study, it is aimed to detection the crop pattern in agricultural areas with high accuracy by using object-based classification technique from high spatial resolution IKONOS satellite images. The study area is located on the South-west of the Karacabey district of the Bursa province in the Marmara Region and covers an area of nearly 9x9 km(2). Tomato, corn, pepper, wheat, rice and sugar beet are the main products grown in the region. In this study, the IKONOS satellite image is segmented using multi-resolution segmentation technique. The most appropriate value for the scale parameter, which is the most important parameter in the segmentation process, has been determined by ESP-2 (Estimation of Scale Parameter) software. Various combinations have been tried for shape and compactness parameters in order to find the optimal segmentation parameters. In order to increase classification accuracy, normalized difference vegetation index (NDVI) and GLCM texture measurement methods have been used, including homogeneity, contrast, dissimilarity, mean, variance, and entropy. Using the data set from consist 29 bands, the image classification process have been performed using the object-based nearest neighbor classification technique in the eCognition software. The obtained classification results have been tested on parcel basis using 2212 ground truth data. The overall accuracy of the classification has been calculated as 87.5%. The results show that the high spatial resolution IKONOS satellite image can be used to detection high accuracy with object-based classification of agricultural crop pattern.