Le traitement de nuages de points 3D de lidars permet aux robots mobiles autonomes terrestres de construire des modèles sémantiques de l'environnement extérieur dans lequel ils évoluent. Ces modèles sont intéressants car ils représentent des informations qualitatives, et ainsi donnent à un robot la capacité de raisonner à un niveau plus élevé d'abstraction. Le coeur d'un système de modélisation sémantique est la capacité de classifier les observations venant du capteur. Nous proposons un système de classification centré sur l'apprentissage non-supervisé. La prémière couche, la couche intermédiaire, consiste en un modèle de mélange gaussien. Ce modèle est déterminé de manière non-supervisée lors d'une étape de training. Il definit un ensemble de classes intermédiaires qui correspond à une partition fine des classes présentes dans l'environnement. La deuxième couche, la couche finale, consiste en un regroupement des classes intermédiaires dans un ensemble de classes finales qui, elles, sont interprétables dans le contexte de la tâche ciblée. Le regroupement est déterminé par un expert lors de l'étape de training, de manière supervisée, mais guidée par les classes intermédiaires. L'évaluation est basée sur deux jeux de données acquis avec de différents lidars et possédant différentes caractéristiques. L'évaluation est quantitative pour l'un des jeux de données, et qualitative pour l'autre. La concéption du système utilise la procédure standard de l'apprentissage, basée sur les étapes de training, validation et test. L'opération suit la pipeline standard de classification. Le système est simple, et ne requiert aucun pré-traitement ou post-traitement. / The processing of 3D lidar point clouds enable terrestrial autonomous mobile robots to build semantic models of the outdoor environments in which they operate. Such models are interesting because they encode qualitative information, and thus provide to a robot the ability to reason at a higher level of abstraction. At the core of a semantic modelling system, lies the capacity to classify the sensor observations. We propose a two-layer classi- fication model which strongly relies on unsupervised learning. The first, intermediary layer consists of a Gaussian mixture model. This model is determined in a training step in an unsupervised manner, and defines a set of intermediary classes which is a fine-partitioned representation of the environment. The second, final layer consists of a grouping of the intermediary classes into final classes that are interpretable in a considered target task. This grouping is determined by an expert during the training step, in a process which is supervised, yet guided by the intermediary classes. The evaluation is done for two datasets acquired with different lidars and possessing different characteristics. It is done quantitatively using one of the datasets, and qualitatively using another. The system is designed following the standard learning procedure, based on a training, a validation and a test steps. The operation follows a standard classification pipeline. The system is simple, with no requirement of pre-processing or post-processing stages.

Discrimination of different tree crowns based on their 3D shapes is essential for a wide range of forestry applications, and, due to its complexity, is a significant challenge. This study presents a modified 3D shape descriptor for the perception of different tree crown shapes in discrete-return LiDAR point clouds. The proposed methodology comprises of five main components, including definition of a local coordinate system, learning salient points, generation of simulated LiDAR point clouds with geometrical shapes, shape signature generation (from simulated LiDAR points as reference shape signature and actual LiDAR point clouds as evaluated shape signature), and finally, similarity assessment of shape signatures in order to extract the shape of a real tree. The first component represents a proposed strategy to define a local coordinate system relating to each tree to normalize 3D point clouds. In the second component, a learning approach is used to categorize all 3D point clouds into two ranks to identify interesting or salient points on each tree. The third component discusses generation of simulated LiDAR point clouds for two geometrical shapes, including a hemisphere and a half-ellipsoid. Then, the operator extracts 3D LiDAR point clouds of actual trees, either deciduous or evergreen. In the fourth component, a longitude-latitude transformation is applied to simulated and actual LiDAR point clouds to generate 3D shape signatures of tree crowns. A critical step is transformation of LiDAR points from their exact positions to their longitude and latitude positions using the longitude-latitude transformation, which is different from the geographic longitude and latitude coordinates, and labeled by their pre-assigned ranks. Then, natural neighbor interpolation converts the point maps to raster datasets. The generated shape signatures from simulated and actual LiDAR points are called reference and evaluated shape signatures, respectively. Lastly, the fifth component determines the similarity between evaluated and reference shape signatures to extract the shape of each examined tree. The entire process is automated by ArcGIS toolboxes through Python programming for further evaluation using more tree crowns in different study areas. Results from LiDAR points captured for 43 trees in the City of Surrey, British Columbia (Canada) suggest that the modified shape descriptor is a promising method for separating different shapes of tree crowns using LiDAR point cloud data. Experimental results also indicate that the modified longitude-latitude shape descriptor fulfills all desired properties of a suitable shape descriptor proposed in computer science along with leaf-off, leaf-on invariance, which makes this process autonomous from the acquisition date of LiDAR data. In summary, the modified longitude-latitude shape descriptor is a promising method for discriminating different shapes of tree crowns using LiDAR point cloud data.