SEGMENTATION OF ELECTRICAL SUBSTATIONS USING DEEP CONVOLUTIONAL NEURAL NETWORK

Abstract. The location of electrical substations is one the factors affecting improvement energy distribution, as well management and control this source. Less cost manpower will be spent through automating process detection segmentation these features with help deep neural networks potential existing high spatial resolution satellite images. In study, a encoder-decoder network was used. This most updated learning methods in image processing segmentation. has been trained three RGB bands high-resolution images (∼1m) eventually segmented areas related to relatively accuracy. As results convolutional network, IOU Precision parameters were obtained, their values 88.2 93.7%, respectively, indicating efficiency proposed method