Using the U‐net convolutional network to map forest types and disturbance in the Atlantic rainforest with very high resolution images

Abstract

Mapping forest types and tree species at regional scales to provide information for ecologists and forest managers is a new challenge for the remote sensing community. Here, we assess the potential of a U‐net convolutional network, a recent deep learning algorithm, to identify and segment (1) natural forests and eucalyptus plantations, and (2) an indicator of forest disturbance, the tree species Cecropia hololeuca, in very high resolution images (0.3 m) from the WorldView‐3 satellite in the Brazilian Atlantic rainforest region. The networks for forest types and Cecropia trees were trained with 7611 and 1568 red‐green‐blue (RGB) images, respectively, and their dense labeled masks. Eighty per cent of the images were used for training and 20% for validation. The U‐net network segmented forest types with an overall accuracy >95% and an intersection over union (IoU) of 0.96. For C. hololeuca, the overall accuracy was 97% and the IoU was 0.86. The predictions were produced over a 1600 km2 region using WorldView‐3 RGB bands pan‐sharpened at 0.3 m. Natural and eucalyptus forests compose 79 and 21% of the region's total forest cover (82 250 ha). Cecropia crowns covered 1% of the natural forest canopy. An index to describe the level of disturbance of the natural forest fragments based on the spatial distribution of Cecropia trees was developed. Our work demonstrates how a deep learning algorithm can support applications such as vegetation, tree species distributions and disturbance mapping on a regional scale.

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

“…In recent years, a few studies have already evaluated the potential of the FCN architectures, specifically U-Net, for forest mapping from optical images [39,40]. In [39], the authors used a U-Net to identify instances of a given tree species from WorldView-3 images. Similarly, in [40], the U-Net was trained with the RGB bands and the digital elevation models (DEM) from high resolution UAV imagery.…”

Mapping Atlantic rainforest degradation and regeneration history with indicator species using convolutional network

“…The natural regeneration, which occurs mainly in abandoned pasture lands, increases the provision of ecosystem services and habitat availability [11]. However, regional scale indicators to assess recovery stage, diversity or disturbance levels of the current natural forests are still underdeveloped, thereby adding uncertainty to the estimation of the ecosystem services and their value for conservation [12][13][14]. In this context, remote sensing is a key tool to monitor biodiversity, resources, and ecosystem services, as well as the human impact on natural ecosystems at a regional/biome scale [15][16][17][18].…”

“…Recently, a deep learning method called U-net was used with very high resolution images (WorldView-3, 30 cm spatial resolution, Digital Globe) to map natural and planted forests and to produce the first regional map of all individual adults of a natural tree species, Cecropia hololeuca, in a region of 1600 km 2 of the Atlantic forest in the state of São Paulo [14]. The classification accuracies above 95% for vegetation type and over 97% for C. hololeuca show that Unet outperforms other image segmentation methods and could support species mapping at regional scale [14]. Since C. hololeuca is a pioneer species, the spatial distribution of the individuals was then used to estimate anthropogenic disturbances inside the natural forest fragments, which are not accessible by others means.…”

Forest Damage Assessment Using Deep Learning on High Resolution Remote Sensing Data

“…In remote sensing related to forestry, the use of deep learning is a promising field of research. To our knowledge, there exist just a few publications on the automatic detection and classification of trees [23,24] based on a U-net architecture [19] and to map forest types and disturbance in the Atlantic rainforest [25]. In addition, there is a conference paper on forest species recognition based on CNNs [26], but in general, there is still a knowledge gap between computer science and remote sensing in general where deep learning is widely used and in other fields of research such as forestry.…”