Study of the Image Segmentation Process Using the Optimized U-Net Model for Drone-Captured Images

Abstract

Aerial views of the scenes captured by UAV or drone have become very familiar as they easily cover the wide view of the scene with different terrain types and landscapes. The detection of the scene images captured by drone and their subparts have been done on the basis of simple image processing approach involving the pixel intensity information. Many computer vision-based algorithms have successfully performed the tasks of segmentation. The manual approach of such segmentation has become time consuming, resource intensive, and laborious. Moreover, the perfection of segmentation on the irregular and noisy images captured by the drones have been lowered to greater extents with application of machine learning algorithms. The machine learning-based UNet model has successfully performed the task of segmentation, and the performance has been enhanced due to optimization. This chapter highlights the different variations of the model and its optimization towards the betterment of accuracy.