This paper proposes an approach referred as: iterative refinement of possibility distributions by learning (IRPDL) for pixel-based image classification. The IRPDL approach is based on the use of possibilistic reasoning concepts exploiting expert knowledge sources as well as ground possibilistic seeds learning. The set of seeds is constructed by incrementally updating and refining the possibility distributions. Synthetic images as well as real images from the RIDER Breast MRI database are being used to evaluate the IRPDL performance. Its performance is compared with three relevant reference methods: region growing, semi-supervised fuzzy pattern matching, and Markov random fields. The IRDPL performance (in terms of recognition rate, 87.3%) is close to the Markovian method (88.8%) that is considered to be the reference in pixel-based image classification. IRPDL outperforms the other two methods, respectively, at the recognition rates of 83.9% and 84.7%. In addition, the proposed IRPDL requires fewer parameters for the mathematical representation and presents a reduced computational complexity.

Convolutional Neural Networks (CNN) have high performance in the fields of object recognition and classification. The strength of CNNs comes from the fact that they are able to extract information from raw-pixel content and learn features automatically. Feature extraction and classification algorithms can be either hand-crafted or Deep Learning (DL) based. DL detection approaches can be either two stages (region proposal approaches) detector or a single stage (non-region proposal approach) detector. Region proposal-based techniques include R-CNN, Fast RCNN, and Faster RCNN. Non-region proposal-based techniques include Single Shot Detector (SSD) and You Only Look Once (YOLO). We are going to compare the speed and accuracy of Faster RCNN, YOLO, and SSD for effective drone detection in various environments. We have found that both Faster RCNN and YOLO have high recognition ability compared to SSD; on the other hand, SSD has good detection ability.

International audienceThe growth of marine renewable energy and marine protected areas in France leads to a growing need for animal population knowledge at sea. Offshore energy generator projects (wind turbines for example) must obey these regulations and show their harmlessness to the environment, particularly to the wildlife and to protected species, which are vulnerable and threatened. This paper presents a supervised learning method of object detection and classification using numerical HD photography: birds, marine mammals, turtles and sharks (near the surface), ships, other mankind objects (waste). The proposed method is based on HOG (Histogram of Oriented Gradients) features extraction and an SVM (Support Vector Machine) classification process. An evaluation of this method on real data is discussed

International audienceThis paper proposes an approach for pixel unmixing based on possibilistic similarity. The approach exploits possibilistic concepts to provide flexibility in the integration of both contextual information and a priori knowledge. Possibility distributions are first obtained using a priori knowledge given in the form of learning areas delimitated by an expert. These areas serve for the estimation of the probability density functions of different thematic classes also called endmembers. The resulting probability density functions are then transformed into possibility distributions using Dubois-Prade's probability-possibility transformation. The pixel unmixing is then performed based on the possibilistic similarity between a local possibility distribution estimated around the considered pixel and the obtained possibility distributions representing the predefined endmembers in the analyzed image. Several possibilistic similarity measures have been tested to improve the discrimination between endmembers. Results show that the proposed approach represents an efficient estimator of the proportion of each endmember present in the pixel (abundances) and achieves higher classification accuracy. Performance analysis has been conducted using synthetic and real images

No abstract