Individual Palm Tree Detection Using Deep Learning on RGB Imagery to Support Tree Inventory

Phoenix palms cover more than 1.3 million hectares in the Mediterranean, Middle East, and North Africa regions and they represent highly valued assets for economic, environmental, and cultural purposes. Despite their importance, information on the number of palm trees and the palm distribution across different scenes is difficult to obtain and, therefore, limited. In this work, we present the first region-wide spatial inventory of Phoenix dactylifera (date palm) and Phoenix canariensis (canary palm) trees, based on remote imagery from the Alicante province in Spain. A deep learning architecture that was based on convolutional neural networks (CNN) was implemented to generate a detection model able to locate and classify individual palms trees from aerial high-resolution RGB images. When considering that creating large labeled image datasets is a constraint in object detection applied to remote sensing data, as a strategy for pre-training detection models on a similar task, imagery and palm maps from the autonomous community of the Canary Islands were used. Subsequently, these models were transferred for re-training with imagery from Alicante. The best performing model was capable of mapping Phoenix palms in different scenes, with a changeable appearance, and with varied ages, achieving a mean average precision (mAP) value of 0.861. In total, 511,095 Phoenix palms with a probability score above 0.5 were detected over an area of 5816 km2. The detection model, which was obtained from an out-of-the-box object detector, RetinaNet, provides a fast and straightforward method to map isolated and densely distributed date and canary palms—and other Phoenix palms. The inventory of palm trees established here provides quantitative information on Phoenix palms distribution, which could be used as a baseline for long-term monitoring of palms’ conditions. In addition to boosting palm tree inventory across multiple landscapes at a large scale, the detection model demonstrates how image processing techniques that are based on deep learning leverage image understanding from remote sensing data.

Download Full-text

Taking the pulse of the Amazon rainforest by fusing multitemporal Sentinel 1 and 2 data for advanced deep-learning

10.5194/egusphere-egu21-3749 ◽

2021 ◽

Author(s):

Melanie Brandmeier ◽

Eya Cherif

Keyword(s):

Remote Sensing ◽

Deep Learning ◽

Land Cover ◽

Supervised Learning ◽

Large Scale ◽

Remote Sensing Data ◽

Ground Truth ◽

Brazilian Amazon ◽

Sensing Data ◽

Weakly Supervised

Degradation of large forest areas such as the Brazilian Amazon due to logging and fires can increase the human footprint way beyond deforestation. Monitoring and quantifying such changes on a large scale has been addressed by several research groups (e.g. Souza et al. 2013) by making use of freely available remote sensing data such as the Landsat archive. However, fully automatic large-scale land cover/land use mapping is still one of the great challenges in remote sensing. One problem is the availability of reliable &#8220;ground truth&#8221; labels for training supervised learning algorithms. For the Amazon area, several landcover maps with 22 classes are available from the MapBiomas project that were derived by semi-automatic classification and verified by extensive fieldwork (Project MapBiomas). These labels cannot be considered real ground-truth as they were derived from Landsat data themselves but can still be used for weakly supervised training of deep-learning models that have a potential to improve predictions on higher resolution data nowadays available. The term weakly supervised learning was originally coined by (Zhou 2017) and refers to the attempt of constructing predictive models from incomplete, inexact and/or inaccurate labels as is often the case in remote sensing. To this end, we investigate advanced deep-learning strategies on Sentinel-1 timeseries and Sentinel-2 optical data to improve large-scale automatic mapping and monitoring of landcover changes in the Amazon area. Sentinel-1 data has the advantage to be resistant to cloud cover that often hinders optical remote sensing in the tropics.We propose new architectures that are adapted to the particularities of remote sensing data (S1 timeseries and multispectral S2 data) and compare the performance to state-of-the-art models.&#160; Results using only spectral data were very promising with overall test accuracies of 77.9% for Unet and 74.7% for a DeepLab implementation with ResNet50 backbone and F1 measures of 43.2% and 44.2% respectively.&#160; On the other hand, preliminary results for new architectures leveraging the multi-temporal aspect of &#160;SAR data have improved the quality of mapping, particularly for agricultural classes. For instance, our new designed network AtrousDeepForestM2 has a similar quantitative performances as DeepLab &#160;(F1 of 58.1% vs 62.1%), however it produces better qualitative land cover maps.To make our approach scalable and feasible for others, we integrate the trained models in a geoprocessing tool in ArcGIS that can also be deployed in a cloud environment and offers a variety of post-processing options to the user.Souza, J., Carlos M., et al. (2013). "Ten-Year Landsat Classification of Deforestation and Forest Degradation in the Brazilian Amazon." Remote Sensing 5(11): 5493-5513.&#160;&#160;&#160;Zhou, Z.-H. (2017). "A brief introduction to weakly supervised learning." National Science Review 5(1): 44-53."Project MapBiomas - Collection&#160; 4.1 of Brazilian Land Cover & Use Map Series, accessed on January 2020 through the link: https://mapbiomas.org/colecoes-mapbiomas?cama_set_language=en"

Download Full-text

Deep learning approach for large scale land cover mapping based on remote sensing data fusion

2016 IEEE International Geoscience and Remote Sensing Symposium (IGARSS) ◽

10.1109/igarss.2016.7729043 ◽

2016 ◽

Cited By ~ 14

Author(s):

Nataliia Kussul ◽

Andrii Shelestov ◽

Mykola Lavreniuk ◽

Igor Butko ◽

Sergii Skakun

Keyword(s):

Remote Sensing ◽

Deep Learning ◽

Land Cover ◽

Data Fusion ◽

Large Scale ◽

Remote Sensing Data ◽

Land Cover Mapping ◽

Learning Approach ◽

Sensing Data ◽

Large Scale Land

Download Full-text

Deep Learning for Monitoring Agricultural Drought in South Asia Using Remote Sensing Data

Remote Sensing ◽

10.3390/rs13091715 ◽

2021 ◽

Vol 13 (9) ◽

pp. 1715

Author(s):

Foyez Ahmed Prodhan ◽

Jiahua Zhang ◽

Fengmei Yao ◽

Lamei Shi ◽

Til Prasad Pangali Sharma ◽

...

Keyword(s):

Remote Sensing ◽

Deep Learning ◽

South Asia ◽

Model Performance ◽

Remote Sensing Data ◽

Agricultural Drought ◽

Gradient Boosting ◽

Complex Nature ◽

Sensing Data

Drought, a climate-related disaster impacting a variety of sectors, poses challenges for millions of people in South Asia. Accurate and complete drought information with a proper monitoring system is very important in revealing the complex nature of drought and its associated factors. In this regard, deep learning is a very promising approach for delineating the non-linear characteristics of drought factors. Therefore, this study aims to monitor drought by employing a deep learning approach with remote sensing data over South Asia from 2001–2016. We considered the precipitation, vegetation, and soil factors for the deep forwarded neural network (DFNN) as model input parameters. The study evaluated agricultural drought using the soil moisture deficit index (SMDI) as a response variable during three crop phenology stages. For a better comparison of deep learning model performance, we adopted two machine learning models, distributed random forest (DRF) and gradient boosting machine (GBM). Results show that the DFNN model outperformed the other two models for SMDI prediction. Furthermore, the results indicated that DFNN captured the drought pattern with high spatial variability across three penology stages. Additionally, the DFNN model showed good stability with its cross-validated data in the training phase, and the estimated SMDI had high correlation coefficient R2 ranges from 0.57~0.90, 0.52~0.94, and 0.49~0.82 during the start of the season (SOS), length of the season (LOS), and end of the season (EOS) respectively. The comparison between inter-annual variability of estimated SMDI and in-situ SPEI (standardized precipitation evapotranspiration index) showed that the estimated SMDI was almost similar to in-situ SPEI. The DFNN model provides comprehensive drought information by producing a consistent spatial distribution of SMDI which establishes the applicability of the DFNN model for drought monitoring.

Download Full-text