TY  - JOUR
TI  - Comparison of Fully Convolution Network (FCN) and Received Accepted U-Net for Road Segmentation from High Resolution Imageries
AB  - Segmentation is one of the most popular classification techniques which still have semantic labels. In this context, the segmentationof different objects such as cars, airplanes, ships, and buildings that are independent of background and objects such as land use andvegetation classes, which are difficult to discriminate from the background is considered. However, in image segmentation studies,various difficulties such as shadow, image blockage, a disorder of background, lighting, shading that cause fundamentalmodifications in the appearance of features are often encountered. With the development of technology, obtaining high spatialresolution satellite imageries and aerial photographs contain detailed texture information have been facilitated easily. Parallel to theseimprovements, deep learning architectures have widely been used to solved several computer vision tasks with an increasing level ofdifficulty. Thus, the regional characteristics, artificial and natural objects, can be perceived and interpreted precisely. In this study,two different subset data that were produced from a great open-source labeled image sets were used to segmentation of roads. Theused labeled data set consists of 150 satellite images of size 1500 x 1500 pixels at a 1.2 m resolution, which was not efficient fortraining. In order to avoid any problem, the imageries were divided into smaller dimensions. Selected images from the data setdivided into small patches of 256 x 256 pixels and 512 x 512 pixels to train the system, and comparisons between them were carriedout. To train the system using these datasets, two different artificial neural network architectures U-Net and Fully ConvolutionalNetworks (FCN), which are used for object segmentation on high-resolution images, were selected. When the test data with the samesize as the training data set were analyzed, approximately 97% extraction accuracy was obtained from high-resolution imageriestrained by FCN in 512 x 512 dimensions.
AU  - Seker, Dursun Zafer
AU  - Sariturk, Batuhan
AU  - Öztürk, Ozan
DO  - 10.30897/ijegeo.737993
PY  - 2020
JO  - International Journal of Environment and Geoinformatics
VL  - 7
IS  - 3
SN  - 2148-9173
SP  - 272
EP  - 279
DB  - TRDizin
UR  - http://search/yayin/detay/429692
ER  -