Automatic Lake and River Detection from Satillite Images Using MobileNet Scheme

Computer algorithm plays a vital role in image examination tasks and hence a number of methods are developed to process grey- and RGB-scaled images. This work considered the examination of the satellite images using MobileNet (MN) based classification system. This work considered the benchmark satellite images having the lake and the satellite images and the obtained images are examined using the following procedures; image collection and resizing, image improvement using the contrast enhancement technique, implementing the MN-approach to extract the necessary features from the image, and binary classification to detect the lake and the river from the chosen images. In this work, lightweight deep-learning technique is implemented by considering the MN and its variants and the obtained result is confirmed and verified. The experimental outcome confirms that the MN - V2 provides a better detection result (accuracy >96%) compared to other MN-schemes of this study.