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

Torres, Daliana Lobo; Feitosa, Raul Queiroz; Happ, Patrick Nigri; Rosa, Laura Elena Cué La; Marcato, José; Martins, José Augusto Legatheaux; Bressan, Patrik Olã; Gonçalves, Wesley Nunes; Liesenberg, Veraldo

doi:10.3390/s20020563

Cited by 100 publications

(68 citation statements)

References 52 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…This layer increases the resolution by replicating the values of the neighbors. To refine the spatial precision, FCN fuses the prediction layer with shallower layers of the network by summing predictions and applying a softmax function at the end (as shown in Figure 2a) (Long et al., 2015; Torres et al., 2020).…”

Section: Methodsmentioning

confidence: 99%

Using Deep Learning for Automatic Water Stage Measurements

Eltner

Bressan

Akiyama

et al. 2021

Water Resources Research

Self Cite

View full text Add to dashboard Cite

Spatiotemporally dense hydrological observation networks are required to provide a suitable database for modeling and planning of water resources. One important hydrological observation is the water stage, which can be retrieved using several methods. Floating and pressure gauges are very common (Morgenschweis, 2010); however, they require the installation in the water and therefore entail the risk of losing them during severe flooding. Thus, remote observation techniques, such as those based on radar or ultrasonic devices (Herschy, 2008), might be preferred in such situations because they have the advantage that they can be installed remotely. Another remote option is the application of image-based methods. In general, they can be low-cost in terms of the device compared to conventional water stage retrieval approaches if, for instance, basic camera circuit boards are used in combination with single board computers or microcontrollers. These setups also have the advantage that data processing can be performed on the spot, and only small data amounts have to be transmitted if needed, for example, via IoT infrastructures (Da Xu et al., 2014). Furthermore, camera gauges provide the benefit of potentially measuring flow velocities simultaneously by capturing short videos and tracking particles at the water surface (e.g., Eltner et al., 2020), which allows quantifying discharge with the same device eventually and thereby avoiding the installation of additional equipment. The application of camera gauges is suitable at rivers with periodic heavy debris loads or low angle river banks with large river width fluctuations leading to failure of other gauging techniques. Observing the river reach with an optical device enables capturing qualitative and quantitative information beyond the water stage or flow velocities. For instance, ephemeral rivers and extended flooded areas become observable, and snow coverage and ice growth can be assessed. Furthermore, the vegetation evolution might be monitored, or the river cross-section changes determined to ensure a continuous and reliable water stage-discharge curve. Thus, camera observations can practically support hydrological monitoring.

show abstract

Section: Methodsmentioning

confidence: 99%

Using Deep Learning for Automatic Water Stage Measurements

Eltner

Bressan

Akiyama

et al. 2021

Water Resources Research

Self Cite

View full text Add to dashboard Cite

show abstract

“…The results confirmed that the original COVID-19 data could be well preprocessed to have comparable intensity distribution to that of well-curated normal data. F. Segmentation network analysis 1) Comparison with U-Net: Recall that we chose FC-DenseNet103 as a backbone segmentation network architecture thanks to its higher segmentation performance with smaller number of parameters (9.4 M) [31]. To demonstrate the effectiveness of CXR segmentation by the FC-Densenet103, we trained U-Net [32] under identical training conditions and compared the results.…”

Section: E Cross-database Generalization Capabilitymentioning

confidence: 99%

Deep Learning COVID-19 Features on CXR Using Limited Training Data Sets

Park

2020

IEEE Trans. Med. Imaging

810

677

View full text Add to dashboard Cite

Under the global pandemic of COVID-19, the use of artificial intelligence to analyze chest X-ray (CXR) image for COVID-19 diagnosis and patient triage is becoming important. Unfortunately, due to the emergent nature of the COVID-19 pandemic, a systematic collection of CXR data set for deep neural network training is difficult. To address this problem, here we propose a patch-based convolutional neural network approach with a relatively small number of trainable parameters for COVID-19 diagnosis. The proposed method is inspired by our statistical analysis of the potential imaging biomarkers of the CXR radiographs. Experimental results show that our method achieves state-of-the-art performance and provides clinically interpretable saliency maps, which are useful for COVID-19 diagnosis and patient triage.

show abstract

“…In [ 8 ] the authors showed that the fusion of Mask-Fast RCNN and OBIA methods increases by 25% the overall accuracy of the segmentation of scattered shrubs in UAV, airborne and GoogleEarth imagery. In [ 36 ] the authors evaluated the performance of five CNN-based methods for the semantic segmentation of a single endangered tree species, called Dipteryx alata Vogel, in UAV images. In particular, they evaluated SegNet, U-Net, FC-DenseNet, and two DeepLabv3+ variants and found that FC-DensNet overcomes all the previous methods with an overall accuracy of 96.7%.…”

Section: Related Workmentioning

confidence: 99%

Olive Tree Biovolume from UAV Multi-Resolution Image Segmentation with Mask R-CNN

Safonova

Guirado

Maglinets

et al. 2021

Sensors

View full text Add to dashboard Cite

Olive tree growing is an important economic activity in many countries, mostly in the Mediterranean Basin, Argentina, Chile, Australia, and California. Although recent intensification techniques organize olive groves in hedgerows, most olive groves are rainfed and the trees are scattered (as in Spain and Italy, which account for 50% of the world’s olive oil production). Accurate measurement of trees biovolume is a first step to monitor their performance in olive production and health. In this work, we use one of the most accurate deep learning instance segmentation methods (Mask R-CNN) and unmanned aerial vehicles (UAV) images for olive tree crown and shadow segmentation (OTCS) to further estimate the biovolume of individual trees. We evaluated our approach on images with different spectral bands (red, green, blue, and near infrared) and vegetation indices (normalized difference vegetation index—NDVI—and green normalized difference vegetation index—GNDVI). The performance of red-green-blue (RGB) images were assessed at two spatial resolutions 3 cm/pixel and 13 cm/pixel, while NDVI and GNDV images were only at 13 cm/pixel. All trained Mask R-CNN-based models showed high performance in the tree crown segmentation, particularly when using the fusion of all dataset in GNDVI and NDVI (F1-measure from 95% to 98%). The comparison in a subset of trees of our estimated biovolume with ground truth measurements showed an average accuracy of 82%. Our results support the use of NDVI and GNDVI spectral indices for the accurate estimation of the biovolume of scattered trees, such as olive trees, in UAV images.

show abstract

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

Cited by 100 publications

References 52 publications

Using Deep Learning for Automatic Water Stage Measurements

Using Deep Learning for Automatic Water Stage Measurements

Deep Learning COVID-19 Features on CXR Using Limited Training Data Sets

Olive Tree Biovolume from UAV Multi-Resolution Image Segmentation with Mask R-CNN

Contact Info

Product

Resources

About