Optimization Techniques in Engineering

9Performance Comparison of Differential Evolutionary Algorithm-Based Contour Detection to Monocular Depth Estimation for Elevation Classification in 2D Drone-Based Imagery

Jacob Vishal, Somdeb Datta, Sudipta Mukhopadhyay, Pravar Kulbhushan, Rik Das^*, Saurabh Srivastava and Indrajit Kar

Siemens Advanta, Siemens Technology and Services Pvt. Ltd. Bengaluru, India

Abstract

Processing of drone-based land imagery is an open research area since it is meant to offer assorted real-time solutions. The images captured using drone cameras comprise of rich descriptor details to leverage content-based image classification. However, the land images captured using drone are two-dimensional in nature, which poses significant challenge in depth estimation for elevation classification of the image content. Moreover, absence of any prior scale (for example, Lidar data, etc.) for estimating the elevation makes the task even more nonconventional. There are state-of the-art techniques existing for understanding the depth of image content using monocular depth estimation of images captured from ground level. This chapter primarily attempts to apply the same technique to carry out sand elevation classification of sand deposits in desert area captured using a drone camera. Nevertheless, the results are not encouraging, which has led to proposition of an evolutionary algorithm-based contour identification method for efficient elevation classification. The structural similarity index (SSIM) to the ground ...