Applying Fully Convolutional Architectures for Semantic Segmentation of a Single Tree Species in Urban Environment on High Resolution UAV Optical Imagery

• Published in: Sensors (Basel, Switzerland) Vol. 20; no. 2; p. 563
• Main Authors: Lobo Torres, Daliana, Queiroz Feitosa, Raul, Nigri Happ, Patrick, Elena Cué La Rosa, Laura, Marcato Junior, José, Martins, José, Olã Bressan, Patrik, Gonçalves, Wesley Nunes, Liesenberg, Veraldo
• Format: Journal Article
• Language: English
• Published: Switzerland MDPI AG 20.01.2020; MDPI
• Subjects: Accuracy; Classification; Climate change; Conditional random fields; Deep learning; Endangered & extinct species; Endangered species; Forests; fully convolution neural networks; Neural networks; Performance enhancement; Post-production processing; Remote sensing; Semantic segmentation; Semantics; Teaching methods; Trees; Trends; unmanned aerial vehicle (uav); Unmanned aerial vehicles; Urban areas; Urban environments; deep learning; unmanned aerial vehicle (UAV); semantic segmentation; fully convolution neural networks
• ISSN: 1424-8220; 1424-8220
• DOI: 10.3390/s20020563

This study proposes and evaluates five deep fully convolutional networks (FCNs) for the semantic segmentation of a single tree species: SegNet, U-Net, FC-DenseNet, and two DeepLabv3+ variants. The performance of the FCN designs is evaluated experimentally in terms of classification accuracy and computational load. We also verify the benefits of fully connected conditional random fields (CRFs) as a post-processing step to improve the segmentation maps. The analysis is conducted on a set of images captured by an RGB camera aboard a UAV flying over an urban area. The dataset also contains a mask that indicates the occurrence of an endangered species called Vogel, also known as cumbaru, taken as the species to be identified. The experimental analysis shows the effectiveness of each design and reports average overall accuracy ranging from 88.9% to 96.7%, an F1-score between 87.0% and 96.1%, and IoU from 77.1% to 92.5%. We also realize that CRF consistently improves the performance, but at a high computational cost.