HAL is a multi-disciplinary open access archive for the deposit and dissemination of scientific research documents, whether they are published or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers. L'archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d'enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.

Coastal areas host highly valuable ecosystems that are increasingly exposed to the threats of global and local changes. Monitoring their evolution at a high temporal and spatial scale is therefore crucial and mostly possible through remote sensing. This article demonstrates the relevance of topobathymetric lidar data for coastal and estuarine habitat mapping by classifying bispectral data to produce 3D maps of 21 land and sea covers at very high resolution. Green lidar full waveforms are processed to retrieve tailored features corresponding to the signature of those habitats. These features, along with infrared intensities and elevations, are used as predictors for random forest classifications, and their respective contribution to the accuracy of the results is assessed. We find that green waveform features, infrared intensities, and elevations are complimentary and yield the best classification results when used in combination. With this configuration, a classification accuracy of 90.5% is achieved for the segmentation of our dual-wavelength lidar dataset. Eventually, we produce an original mapping of a coastal site under the form of a point cloud, paving the way for 3D classification and management of land and sea covers.

Abstract. Mapping coastal habitats is essential to their preservation, but the presence of water hinders seamless data collection over land-water interfaces. Thanks to its dual-wavelength and optical properties, topo-bathymetric lidar can address this task efficiently. Topo-bathymetric lidar waveforms contain relevant information to classify land and water covers automatically but are rarely analysed for both infrared and green wavelengths. The present study introduces a point-based approach for the classification of coastal habitats using bispectral waveforms of topo-bathymetric lidar surveys and machine learning. Spectral features and differential elevations are fed to a random forest algorithm to produce three-dimensional classified point clouds of 17 land and sea covers. The resulting map reaches an overall accuracy of 86%, and 65% of the prediction probabilities are above 0.60. Using this prediction confidence, it is possible to map coastal habitats and eliminate the classification errors due to noise in the data, that generate a clear tendency of the algorithm to over-estimate some classes at the expense of some others. By filtering out points with a low prediction confidence (under 0.7), the classification can be highly improved and reach an overall accuracy of 97%.

No abstract

No abstract

No abstract

<p>Remote sensing techniques are now widely spread for the early detection of ground deformation, implementation of warning systems in case of imminent landslide triggering, and medium- and long-term slope instability monitoring. The large breadth of data available to the scientific community, associated with processing techniques improved as the data volume was increasing, has led to noticeable developments in the field of remote sensing data processing, using machine learning algorithms and more particularly deep neural networks.</p><p> </p><p>This arsenal of data and techniques is necessary for the present scientific challenges the community of researchers on landslides still have to meet. As landslides can be complex, for risk management and disaster mitigation strategies, it is necessary to have a precise idea of their location, shape, and size to be studied and monitored. The challenge aims to automate landslide detection and mapping, especially through learning methods. Machine learning methods based on Deep Neural Networks have recently been employed for landslide studies and provide promising efficient results for landslide detection [1].</p><p> </p><p>In this study, we propose an original neural network for landslide detection. More precisely, we exploit a fusion network [1] dealing with optical images on the one hand and Digital Elevation Models on the other hand. To improve the results, attention layers [3] (able to stabilize the training and more precise results) as well as mix up techniques [4] (able to generalize more efficiently) are exploited.</p><p>The model was trained and tested on the open Bijie landslide dataset.</p><p> </p><p>Keywords: Remote sensing for landslide monitoring and detection, landslide detection, deep neural networks, attention</p><p> </p><p>[1] Ji, S., Yu, D., Shen, C., Li, W., & Xu, Q. (2020). Landslide detection from an open satellite imagery and digital elevation model dataset using attention-boosted convolutional neural networks. <em>Landslides</em>, <em>17</em>(6), 1337-1352.</p><p>[2] Song, W., Li, S., Fang, L., & Lu, T. (2018). Hyperspectral image classification with deep feature fusion network. <em>IEEE Transactions on Geoscience and Remote Sensing</em>, <em>56</em>(6), 3173-3184.</p><p>[3] Niu, Z., Zhong, G., & Yu, H. (2021). A review on the attention mechanism of deep learning. <em>Neurocomputing</em>, <em>452</em>, 48-62.</p><p>[4] Thulasidasan, S., Chennupati, G., Bilmes, J. A., Bhattacharya, T., & Michalak, S. (2019). On mixup training: Improved calibration and predictive uncertainty for deep neural networks. <em>Advances in Neural Information Processing Systems</em>, <em>32</em>.</p>

No abstract