This article aims at proposing the concept of thick fuzzy sets (TFSs). A TFS is based on the joint use of thick sets (TSs) and α-cuts concepts. A TFS is represented by a family of nested TSs. A TS is an uncertain set, which is represented by a pair of crisp sets (CSs). These CSs characterize the upper and lower bounds of the TS. Therefore, a TS can be regarded as an interval of CSs. In this framework, as a type-1 fuzzy set (T1FS) is regarded as a family of nested CSs, a TFS can be represented by a family of nested TSs. Furthermore, according to the vertical dimension α, a TFS can be regarded as an interval with T1FS bounds. The potentialities of the TFS concept have been validated using application examples where a real-world application for modeling the zone explored by an underwater robot is given.

In the framework of belief functions, basic belief assignments building is an important step that should be made carefully since it can greatly influence the performances of a system. In the context of tree species recognition through a leaf and a bark, we analyze the impact of Bayesian as well as consonant basic belief assignments in the case of fusion of uncertain and not equally reliable sources of information. We show that, despite the imprecision they bring, consonant belief functions prove their performances in managing and representing uncertain information. In this application, photos of leaves and barks are the sources of information allowing tree species recognition.

In the context of tree species recognition, botanists knowledge was used in different works specially when recognising tree species through leaves. In this paper, two sub-classification strategies for tree species recognition are proposed. For each sub-classification strategy, Basic belief assignment (Bba) was determined and obtained data were fused thanks to a totally adaptive fusion system implemented in the general framework of belief functions.

Nowadays, robotics technologies act more and more important roles in our industrial life. However, developing a robot with intelligent behaviors that follow human perception and reasoning is really a challenge. This paper introduces an artificial intelligence technique that helps a NAO robot intuitively recognize the color of a required object. Firstly, fuzzy logic is used to infer a linguistic color from pixel values. After that, evidence theory is employed to fuse fuzzy results from multiple cameras to make better decision. These methodologies obtain a good recognition quality through real time experimentations.

Nowadays, the interaction between humans and systems is much better, which allows humans to be able to accomplish a huge range of tasks. Focusing on intelligent communication and human understanding of systems, in this work we develop an application, which allows a humanoid robot to receive oral commands from humans to find a desired colored object. Fuzzy logic is used for color recognition following human perception and reasoning, and Evidence theory is employed to fuse information from multiple cameras in order to improve recognition quality. Experimentation results show that the combination of these methods help the robot to recognize linguistic colors efficiently.

Recognition of human gesture leads to a dynamic field that produces many various methods. The main way to improve the recognition process is to perform data fusion based on a qualification of each recognition method. The advance in data fusion gives also several solutions and the choice of a fusion method is a crucial point. The goal of this paper is to present an approach were the choice of the fusion method is driven by the constraints of the final user of data. Such constraints are expressed as bounds on indicators like confidence or accuracy. A practical experimentation illustrating this approach is presented in this paper. From a depth camera sensor, human gesture is interpreted and, fusion method is selected in accordance with the constraints of the final information consumer. This approach is illustrated in a robotic control by human gesture: outputs perform the action to be done by a NAO robot.

Within the framework of a smartphone-based application, helping people to identify plant species in the wild, a sub-classifier strategy has been introduced. It aims at recognizing the botanical properties of a leaf, relatively to various global and local shape criteria used in flora books. A decision function is applied on these classified shape categories to produce a final decision on the species of the leaf. In this paper, the fusion strategy and its corresponding Random-Forest-based sub-classifiers are described. The results of these algorithms for botanical leaf shape recognition demonstrate that our classification algorithm can achieve good performance on leaf species identification while providing the user with relevant information for educational purposes.

In this paper, we propose a specific method for the identification of compound-leaved tree species, with the aim of integrating it in an educational smartphone application. Our work is based on dedicated shape models for compound leaves, designed to estimate the number and shape of leaflets. A deformable template approach is used to fit these models and produce a high-level interpretation of the image content. The resulting models are later used for the segmentation of leaves in both plain and natural background images, by the use of multiple region-based active contours. Combined with other botany-inspired descriptors accounting for the morphological properties of the leaves, we propose a classification method that makes a semantic interpretation possible. Results are presented over a set of more than 1000 images from 17 European tree species, and an integration in the existing mobile application Folia1is considered.

The use of cooperative fusion systems for image interpretation according to expert knowledge has known a wide growth and they now need reliable ways to evaluate their performance. Local evaluation is an interesting way to better characterize the system behaviour and consequently to have information on which subpart needs to be adjusted. A previously proposed local evaluation based on separability indexes, generates many numerical indicators. In this paper we proposes a symbolic representation of these separability indexes easier to understand by the experts. To overcome the application of a threshold, the uncertainty of the numeric measures is represented thanks to possibility distributions.

This paper deals with a cooperative fusion system parameter adjustment based on an evaluation quality of the fusion system output. The concerned application is an electro- technical part quality evaluation in which geophysicists need time and experience to analyze the 3D tomography of the parts. To help them in their task, a fusion system was designed to aggregate some measurements extracted from the image. The Choquet integral was used to achieve the fusion. The adjustment of the fusion system is difficult to do by the end-user. This paper presents a procedure to guide experts in the use of the system. First of all, classified images obtained by the fusion system are evaluated. The problematic of image evaluation is addressed and an evaluation based on both quantitative and qualitative criteria is proposed. Then, some actions are defined and presented as guidelines to the expert in order to adjust the parameters in a structured way. The approach is illustrated on the classification of the 3D tomographic bloc.

In this paper an improved cut detection algorithm, adapted to the segmentation of animation movies, is proposed. As color is a major feature of animation movies (each movie has its own particular color distribution) the proposed algorithm applies second order derivatives on Euclidean distances between color histograms of frames quadrants in order to improve the cut detection. For the frame classification, an automatic threshold estimation is proposed. Also, in order to reduce false detections, we propose an algorithm to detect an effect specific to animation movies, named "short color change" (i.e. thunders, lightening). The resulting method achieved better results compared to the classical histogram-based and motion-discontinuity based approaches, as shown by tests conducted on several animation movies

This paper presents a method for the automatic generation of animation movie trailers. First, the movie is divided into shots, by detecting the video transitions (cuts, fades and dissolves) and an animation movie specific color effect named short color change or SCC (i.e. explosions, thunders). The movie action content is further highlighted by analyzing the movie at two different granularity levels. First, an inter-shot analysis is performed by measuring the video transition temporal distribution. As repetitive shot changes are related to action, we define as an action shot, a movie segment containing a high shot change ratio. On the other hand, an inter-frame analysis is performed: for each shot within an action shot, an histogram of cumulative inter-frame distances is computed to serve as a measure of the frame spatial activity. Repetitive color changes are also related to action. Since movie trailers only show some of the most attractive movie scenes, the proposed trailer is a moving-image abstract computed on the retained action shots. It provides the user with a compact and efficient representation of the movie action content. The proposed approach was tested on several animation movies.

In this paper, a 3D nonstationary local distance operator is introduced. This kind of operator is adapted to the measurement of distances between objects sampled on nonuniform grids. This operator can be regarded as a nonstationary version of local distance operators adapted to parallelepipedic (noncubic) grids. The principle of coefficient computation and spatial adaptation are given. An application to dissimilarity measurement between color images is presented. We compare the nonstationary local distance operator with two stationary versions: 1) images are sampled with the lowest sampling step; and 2) the local operator is designed according to the average voxel size.

Many skeletonization methods based on distance transformation have been proposed. They are generally designed for square grids. However most industrial vision systems digitize images on rectangular grids. These images are then resampled to obtain square pixels. In this paper we propose a skeletonization algorithm directly adapted to rectangular grids. We use the efficient skeletonization method of Arcelli and Sanniti di Baja (1993), based on the Euclidean distance transform, and we adapt it to chamfer distances computed on rectangular grids. We propose some modifications to this algorithm. Some results are given and discussed.

Proposes a procedure for automatic classification by successive stages in order to construct models. The method proposed is based on hierarchical classification. In a way, it is close to a sequential decision method and it demands prior knowledge of the different categories of individuals to be modeled; which is generally the case in many fields of application. The method was compared to the classic method of hierarchical classification and gave excellent results with a reduced number of parameters. The method applied to zooplankton classification.>