International audienceA path generator is proposed for a fixed-wing Unmaned Aereal Vehicle (UAV). Assuming that the vehicle maintain a constant altitude, and airspeed, and that the UAV is constrained by a turning rate. The Dubins paths serve as a strategy to find the shortest path for the non-holonomic model of the UAV. Dubins paths consist of three path segments which are based on straight lines or arcs of circle of a given radius. The Dubins path generation is combined with a nonlinear Lypaunov-based path-following control. Finally we present a complete simulation environment in which the path generator and path following strategy are validated. As an example of application we propose the scenario in which a missing person is located in some known area and we use the path generator along with this path-following strategy applied to the fixed wing UAV to search and find this person.Un générateur de trajectoire est proposé pour une aéronefs à voilure fixe véhicule aérien sans pilote (UAV). En supposant que le véhicule à maintenir une altitude constante, et la vitesse, et que l'UAV est contraint par un taux de rotation. Les chemins Dubins servent comme une stratégie pour trouver le chemin le plus court pour le modèle non-holonome de l'UAV. Chemins Dubins se composent de trois segments de chemin qui sont basés sur des lignes droites ou des arcs de cercle d'un rayon donné. La génération de trajectoire Dubins est combinée avec une commande de trajet suivant, basé sur Lyapunov non linéaire. Enfin, nous présentons un environnement de simulation complet dans lequel le générateur de trajet et stratégie de suivi sont validés. Comme exemple d'application, nous vous proposons le scénario dans lequel une personne disparue se trouve dans un domaine connu et nous utilisons le générateur du chemin avec cette stratégie suivi de trajectoire appliquée à l'UAV à ailes fixes pour rechercher et trouver cette personne

Crop and weed discrimination in natural field environments is still challenging for implementing automatic agricultural practices, such as weed control. Some weed control methods have been proposed. However, these methods are still restricted as they are implemented under controlled conditions. The development of a sound weed control system begins by recognizing the crop and the different weed plants presented in the field. In this work, a classification approach of Zea mays L. (Crop), narrow-leaf weeds (NLW), and broadleaf weeds (BLW) from multi-plant images are presented. Moreover, a large image dataset was generated. Images were captured in natural field conditions, in different locations, and growing stages of the plants. The extraction of regions of interest (ROI) is carried out employing connected component analysis (CCA), whereas the classification of ROIs is based on Convolutional Neural Networks (CNN) and compared with a shallow learning approach. To measure the classification performance of both methods, accuracy, precision, recall, and F1-score metrics were used. The best alternative for the weed classification task at early stages of growth and in natural corn field environments was the CNN-based approach, as indicated by the 97% accuracy value obtained.

This work presents the design and development of a multispectral imaging system to precision agriculture tasks. The imaging system features two small digital cameras controlled by a microcomputer embedded in a drone. One of the cameras has been modified to be sensitive to near-infrared radiation reflected by the vegetation, whereas the other one remains as a common RGB camera. In order to determine the health status of the crop, the Normalized Difference Vegetation Index (NDVI) is computed in a developed software. Once the aerial imagery is obtained by the drone, it is processed to eliminate image distortions and insert specific metadata needed for generating the orthomosaics with the health information of the plant or soil of interest. Finally, the vegetation index will be computed from the visible and near-infrared orthomosaics for a better interpretation of the user. Experiments are presented to show the effectiveness of the system.

The Visual Simultaneous Localization and Mapping (VSLAM) is a system based on the scene’s features to estimate a map and the system pose. Commonly, VSLAM algorithms are focused on a static environment; however, some dynamic objects are present in the vast majority of real-world applications. This work presents a feature-based SLAM system focused on dynamic environments using convolutional neural networks, optical flow, and depth maps to detect objects in the scene. The proposed system employs a stereo camera as the primary sensor to capture the scene. The neural network is responsible for object detection and segmentation to avoid erroneous maps and wrong system locations. Moreover, the proposed system’s processing time is fast and can run in real-time, running in outdoor and indoor environments. The proposed approach has been compared with state-of-the-art; besides, we present several experimental results outdoors that corroborate the approach’s effectiveness. Our code is available online.