Semantic segmentation of vegetation images acquired by unmanned aerial vehicles using an ensemble of ConvNets

Abstract

Vegetation segmentation in high resolution images acquired by unmanned aerial vehicles (UAVs) is a challenging task that requires methods capable of learning high-level features while dealing with fine-grained data. In this paper, we propose a combination of different methods of semantic segmentation based on Convolutional Networks (ConvNets) to obtain highly accurate segmentation of individuals of different vegetation species. The objective is not only to learn specific and adaptable features depending on the data, but also to learn and combine appropriate classifiers. We conducted a systematic evaluation using a high-resolution UAV-based image dataset related to a campo rupestre vegetation in the Brazilian Cerrado biome. Experimental results show that the ensemble technique overcomes all segmentation strategies.