Scene Mapping and Understanding by Robotic Vision

The first mechanical Automaton concept was found in a Chinese text written in the 3rd century BC, while Computer Vision was born in the late 1960s. Therefore, visual perception applied to machines (i.e. the Machine Vision) is a young and exciting alliance. When robots came in, the new field of Robotic Vision was born, and these terms began to be erroneously interchanged. In short, we can say that Machine Vision is an engineering domain, which concern the industrial use of Vision. The Robotic Vision, instead, is a research field that tries to incorporate robotics aspects in computer vision algorithms. Visual Servoing, for example, is one of the problems that cannot be solved by computer vision only. Accordingly, a large part of this work deals with boosting popular Computer Vision techniques by exploiting robotics: e.g. the use of kinematics to localize a vision sensor, mounted as the robot end-effector. The remainder of this work is dedicated to the counterparty, i.e. the use of computer vision to solve real robotic problems like grasping objects or navigate avoiding obstacles. Will be presented a brief survey about mapping data structures most widely used in robotics along with SkiMap, a novel sparse data structure created both for robotic mapping and as a general purpose 3D spatial index. Thus, several approaches to implement Object Detection and Manipulation, by exploiting the aforementioned mapping strategies, will be proposed, along with a completely new Machine Teaching facility in order to simply the training procedure of modern Deep Learning networks.

Analytics and Methods for Logistic System Design and Operations Management

Analytics is the technology working with the manipulation of data to produce information able to change the world we live every day. Analytics have been largely used within the last decade to cluster people’s behaviour to predict their preferences of items to buy, music to listen, movies to watch and even electoral preference. The most advanced companies succeded in controlling people’s behaviour using analytics. Despite the evidence of the super-power of analytics, they are rarely applied to the big data collected within supply chain systems (i.e. distribution network, storage systems and production plants). This PhD thesis explores the fourth research paradigm (i.e. the generation of knowledge from data) applied to supply chain system design and operations management. An ontology defining the entities and the metrics of supply chain systems is used to design data structures for data collection in supply chain systems. The consistency of this data is provided by mathematical demonstrations inspired by the factory physics theory. The availability, quantity and quality of the data within these data structures define different decision patterns. Ten decision patterns are identified, and validated on-field, to address ten different class of design and control problems in the field of supply chain systems research.

Kernel Methods for Tree Structured Data

Machine learning comprises a series of techniques for automatic extraction of meaningful information from large collections of noisy data. In many real world applications, data is naturally represented in structured form. Since traditional methods in machine learning deal with vectorial information, they require an a priori form of preprocessing. Among all the learning techniques for dealing with structured data, kernel methods are recognized to have a strong theoretical background and to be effective approaches. They do not require an explicit vectorial representation of the data in terms of features, but rely on a measure of similarity between any pair of objects of a domain, the kernel function. Designing fast and good kernel functions is a challenging problem. In the case of tree structured data two issues become relevant: kernel for trees should not be sparse and should be fast to compute. The sparsity problem arises when, given a dataset and a kernel function, most structures of the dataset are completely dissimilar to one another. In those cases the classifier has too few information for making correct predictions on unseen data. In fact, it tends to produce a discriminating function behaving as the nearest neighbour rule. Sparsity is likely to arise for some standard tree kernel functions, such as the subtree and subset tree kernel, when they are applied to datasets with node labels belonging to a large domain. A second drawback of using tree kernels is the time complexity required both in learning and classification phases. Such a complexity can sometimes prevents the kernel application in scenarios involving large amount of data. This thesis proposes three contributions for resolving the above issues of kernel for trees. A first contribution aims at creating kernel functions which adapt to the statistical properties of the dataset, thus reducing its sparsity with respect to traditional tree kernel functions. Specifically, we propose to encode the input trees by an algorithm able to project the data onto a lower dimensional space with the property that similar structures are mapped similarly. By building kernel functions on the lower dimensional representation, we are able to perform inexact matchings between different inputs in the original space. A second contribution is the proposal of a novel kernel function based on the convolution kernel framework. Convolution kernel measures the similarity of two objects in terms of the similarities of their subparts. Most convolution kernels are based on counting the number of shared substructures, partially discarding information about their position in the original structure. The kernel function we propose is, instead, especially focused on this aspect. A third contribution is devoted at reducing the computational burden related to the calculation of a kernel function between a tree and a forest of trees, which is a typical operation in the classification phase and, for some algorithms, also in the learning phase. We propose a general methodology applicable to convolution kernels. Moreover, we show an instantiation of our technique when kernels such as the subtree and subset tree kernels are employed. In those cases, Direct Acyclic Graphs can be used to compactly represent shared substructures in different trees, thus reducing the computational burden and storage requirements.

Higher-order co-occurrences for exploratory point pattern analysis and decision tree clustering on spatial data

Analyzing geographical patterns by collocating events, objects or their attributes has a long history in surveillance and monitoring, and is particularly applied in environmental contexts, such as ecology or epidemiology. The identification of patterns or structures at some scales can be addressed using spatial statistics, particularly marked point processes methodologies. Classification and regression trees are also related to this goal of finding "patterns" by deducing the hierarchy of influence of variables on a dependent outcome. Such variable selection methods have been applied to spatial data, but, often without explicitly acknowledging the spatial dependence. Many methods routinely used in exploratory point pattern analysis are2nd-order statistics, used in a univariate context, though there is also a wide literature on modelling methods for multivariate point pattern processes. This paper proposes an exploratory approach for multivariate spatial data using higher-order statistics built from co-occurrences of events or marks given by the point processes. A spatial entropy measure, derived from these multinomial distributions of co-occurrences at a given order, constitutes the basis of the proposed exploratory methods. © 2010 Elsevier Ltd.

Higher-order co-occurrences for exploratory point pattern analysis and decision tree clustering on spatial data

Analyzing geographical patterns by collocating events, objects or their attributes has a long history in surveillance and monitoring, and is particularly applied in environmental contexts, such as ecology or epidemiology. The identification of patterns or structures at some scales can be addressed using spatial statistics, particularly marked point processes methodologies. Classification and regression trees are also related to this goal of finding "patterns" by deducing the hierarchy of influence of variables on a dependent outcome. Such variable selection methods have been applied to spatial data, but, often without explicitly acknowledging the spatial dependence. Many methods routinely used in exploratory point pattern analysis are2nd-order statistics, used in a univariate context, though there is also a wide literature on modelling methods for multivariate point pattern processes. This paper proposes an exploratory approach for multivariate spatial data using higher-order statistics built from co-occurrences of events or marks given by the point processes. A spatial entropy measure, derived from these multinomial distributions of co-occurrences at a given order, constitutes the basis of the proposed exploratory methods. © 2010 Elsevier Ltd.

Fast ray-tracing algorithm for circumstellar structures (FRACS) II. Disc parameters of the B[e] supergiant CPD-57 degrees 2874 from VLTI/MIDI data

Context. B[e] supergiants are luminous, massive post-main sequence stars exhibiting non-spherical winds, forbidden lines, and hot dust in a disc-like structure. The physical properties of their rich and complex circumstellar environment (CSE) are not well understood, partly because these CSE cannot be easily resolved at the large distances found for B[e] supergiants (typically greater than or similar to 1 kpc). Aims. From mid-IR spectro-interferometric observations obtained with VLTI/MIDI we seek to resolve and study the CSE of the Galactic B[e] supergiant CPD-57 degrees 2874. Methods. For a physical interpretation of the observables (visibilities and spectrum) we use our ray-tracing radiative transfer code (FRACS), which is optimised for thermal spectro-interferometric observations. Results. Thanks to the short computing time required by FRACS (<10 s per monochromatic model), best-fit parameters and uncertainties for several physical quantities of CPD-57 degrees 2874 were obtained, such as inner dust radius, relative flux contribution of the central source and of the dusty CSE, dust temperature profile, and disc inclination. Conclusions. The analysis of VLTI/MIDI data with FRACS allowed one of the first direct determinations of physical parameters of the dusty CSE of a B[e] supergiant based on interferometric data and using a full model-fitting approach. In a larger context, the study of B[e] supergiants is important for a deeper understanding of the complex structure and evolution of hot, massive stars.

Different Approaches for Modeling Grouped Survival Data: A Mango Tree Study

Interval-censored survival data, in which the event of interest is not observed exactly but is only known to occur within some time interval, occur very frequently. In some situations, event times might be censored into different, possibly overlapping intervals of variable widths; however, in other situations, information is available for all units at the same observed visit time. In the latter cases, interval-censored data are termed grouped survival data. Here we present alternative approaches for analyzing interval-censored data. We illustrate these techniques using a survival data set involving mango tree lifetimes. This study is an example of grouped survival data.

GA topology optimization using random keys for tree encoding of structures

Topology optimization consists in finding the spatial distribution of a given total volume of material for the resulting structure to have some optimal property, for instance, maximization of structural stiffness or maximization of the fundamental eigenfrequency. In this paper a Genetic Algorithm (GA) employing a representation method based on trees is developed to generate initial feasible individuals that remain feasible upon crossover and mutation and as such do not require any repairing operator to ensure feasibility. Several application examples are studied involving the topology optimization of structures where the objective functions is the maximization of the stiffness and the maximization of the first and the second eigenfrequencies of a plate, all cases having a prescribed material volume constraint.

Energy efficient scheduling for cluster-tree wireless sensor networks with time-bounded data flows: application to IEEE 802.15.4/ZigBee

Cluster scheduling and collision avoidance are crucial issues in large-scale cluster-tree Wireless Sensor Networks (WSNs). The paper presents a methodology that provides a Time Division Cluster Scheduling (TDCS) mechanism based on the cyclic extension of RCPS/TC (Resource Constrained Project Scheduling with Temporal Constraints) problem for a cluster-tree WSN, assuming bounded communication errors. The objective is to meet all end-to-end deadlines of a predefined set of time-bounded data flows while minimizing the energy consumption of the nodes by setting the TDCS period as long as possible. Sinceeach cluster is active only once during the period, the end-to-end delay of a given flow may span over several periods when there are the flows with opposite direction. The scheduling tool enables system designers to efficiently configure all required parameters of the IEEE 802.15.4/ZigBee beaconenabled cluster-tree WSNs in the network design time. The performance evaluation of thescheduling tool shows that the problems with dozens of nodes can be solved while using optimal solvers.

Understanding the low-temperature limitations to forest growth through calibration of a forest dynamics model with tree-ring data

The sensitivity of altitudinal and latitudinal tree-line ecotones to climate change, particularly that of temperature, has received much attention. To improve our understanding of the factors affecting tree-line position, we used the spatially explicit dynamic forest model TreeMig. Although well-suited because of its landscape dynamics functions, TreeMig features a parabolic temperature growth response curve, which has recently been questioned. and the species parameters are not specifically calibrated for cold temperatures. Our main goals were to improve the theoretical basis of the temperature growth response curve in the model and develop a method for deriving that curve's parameters from tree-ring data. We replaced the parabola with an asymptotic curve, calibrated for the main species at the subalpine (Swiss Alps: Pinus cembra, Larix decidua, Picea abies) and boreal (Fennoscandia: Pinus sylvestris, Betula pubescens, P. abies) tree-lines. After fitting new parameters, the growth curve matched observed tree-ring widths better. For the subalpine species, the minimum degree-day sum allowing, growth (kDDMin) was lowered by around 100 degree-days; in the case of Larix, the maximum potential ring-width was increased to 5.19 mm. At the boreal tree-line, the kDDMin for P. sylvestris was lowered by 210 degree-days and its maximum ring-width increased to 2.943 mm; for Betula (new in the model) kDDMin was set to 325 degree-days and the maximum ring-width to 2.51 mm; the values from the only boreal sample site for Picea were similar to the subalpine ones, so the same parameters were used. However, adjusting the growth response alone did not improve the model's output concerning species' distributions and their relative importance at tree-line. Minimum winter temperature (MinWiT, mean of the coldest winter month), which controls seedling establishment in TreeMig, proved more important for determining distribution. Picea, P. sylvestris and Betula did not previously have minimum winter temperature limits, so these values were set to the 95th percentile of each species' coldest MinWiT site (respectively -7, -11, -13). In a case study for the Alps, the original and newly calibrated versions of TreeMig were compared with biomass data from the National Forest Inventor), (NFI). Both models gave similar, reasonably realistic results. In conclusion, this method of deriving temperature responses from tree-rings works well. However, regeneration and its underlying factors seem more important for controlling species' distributions than previously thought. More research on regeneration ecology, especially at the upper limit of forests. is needed to improve predictions of tree-line responses to climate change further.

Schistosomiasis risk mapping in the state of Minas Gerais, Brazil, using a decision tree approach, remote sensing data and sociological indicators

Schistosomiasis mansoni is not just a physical disease, but is related to social and behavioural factors as well. Snails of the Biomphalaria genus are an intermediate host for Schistosoma mansoni and infect humans through water. The objective of this study is to classify the risk of schistosomiasis in the state of Minas Gerais (MG). We focus on socioeconomic and demographic features, basic sanitation features, the presence of accumulated water bodies, dense vegetation in the summer and winter seasons and related terrain characteristics. We draw on the decision tree approach to infection risk modelling and mapping. The model robustness was properly verified. The main variables that were selected by the procedure included the terrain's water accumulation capacity, temperature extremes and the Human Development Index. In addition, the model was used to generate two maps, one that included risk classification for the entire of MG and another that included classification errors. The resulting map was 62.9% accurate.

Tree-based message diffusion for managing replicated data in unreliable and resource-constrained peer-to-peer environment

Abstract This thesis proposes a set of adaptive broadcast solutions and an adaptive data replication solution to support the deployment of P2P applications. P2P applications are an emerging type of distributed applications that are running on top of P2P networks. Typical P2P applications are video streaming, file sharing, etc. While interesting because they are fully distributed, P2P applications suffer from several deployment problems, due to the nature of the environment on which they perform. Indeed, defining an application on top of a P2P network often means defining an application where peers contribute resources in exchange for their ability to use the P2P application. For example, in P2P file sharing application, while the user is downloading some file, the P2P application is in parallel serving that file to other users. Such peers could have limited hardware resources, e.g., CPU, bandwidth and memory or the end-user could decide to limit the resources it dedicates to the P2P application a priori. In addition, a P2P network is typically emerged into an unreliable environment, where communication links and processes are subject to message losses and crashes, respectively. To support P2P applications, this thesis proposes a set of services that address some underlying constraints related to the nature of P2P networks. The proposed services include a set of adaptive broadcast solutions and an adaptive data replication solution that can be used as the basis of several P2P applications. Our data replication solution permits to increase availability and to reduce the communication overhead. The broadcast solutions aim, at providing a communication substrate encapsulating one of the key communication paradigms used by P2P applications: broadcast. Our broadcast solutions typically aim at offering reliability and scalability to some upper layer, be it an end-to-end P2P application or another system-level layer, such as a data replication layer. Our contributions are organized in a protocol stack made of three layers. In each layer, we propose a set of adaptive protocols that address specific constraints imposed by the environment. Each protocol is evaluated through a set of simulations. The adaptiveness aspect of our solutions relies on the fact that they take into account the constraints of the underlying system in a proactive manner. To model these constraints, we define an environment approximation algorithm allowing us to obtain an approximated view about the system or part of it. This approximated view includes the topology and the components reliability expressed in probabilistic terms. To adapt to the underlying system constraints, the proposed broadcast solutions route messages through tree overlays permitting to maximize the broadcast reliability. Here, the broadcast reliability is expressed as a function of the selected paths reliability and of the use of available resources. These resources are modeled in terms of quotas of messages translating the receiving and sending capacities at each node. To allow a deployment in a large-scale system, we take into account the available memory at processes by limiting the view they have to maintain about the system. Using this partial view, we propose three scalable broadcast algorithms, which are based on a propagation overlay that tends to the global tree overlay and adapts to some constraints of the underlying system. At a higher level, this thesis also proposes a data replication solution that is adaptive both in terms of replica placement and in terms of request routing. At the routing level, this solution takes the unreliability of the environment into account, in order to maximize reliable delivery of requests. At the replica placement level, the dynamically changing origin and frequency of read/write requests are analyzed, in order to define a set of replica that minimizes communication cost.

Intégration de méthodes de data mining dans le renseignement criminel : analyse par des structures issues de la théorie des graphes dans le profilage des stupéfiants

Data mining can be defined as the extraction of previously unknown and potentially useful information from large datasets. The main principle is to devise computer programs that run through databases and automatically seek deterministic patterns. It is applied in different fields of application, e.g., remote sensing, biometry, speech recognition, but has seldom been applied to forensic case data. The intrinsic difficulty related to the use of such data lies in its heterogeneity, which comes from the many different sources of information. The aim of this study is to highlight potential uses of pattern recognition that would provide relevant results from a criminal intelligence point of view. The role of data mining within a global crime analysis methodology is to detect all types of structures in a dataset. Once filtered and interpreted, those structures can point to previously unseen criminal activities. The interpretation of patterns for intelligence purposes is the final stage of the process. It allows the researcher to validate the whole methodology and to refine each step if necessary. An application to cutting agents found in illicit drug seizures was performed. A combinatorial approach was done, using the presence and the absence of products. Methods coming from the graph theory field were used to extract patterns in data constituted by links between products and place and date of seizure. A data mining process completed using graphing techniques is called ``graph mining''. Patterns were detected that had to be interpreted and compared with preliminary knowledge to establish their relevancy. The illicit drug profiling process is actually an intelligence process that uses preliminary illicit drug classes to classify new samples. Methods proposed in this study could be used \textit{a priori} to compare structures from preliminary and post-detection patterns. This new knowledge of a repeated structure may provide valuable complementary information to profiling and become a source of intelligence.

BIO-INSPIRED COMPUTATIONAL TECHNIQUES APPLIED TO THE CLUSTERING AND VISUALIZATION OF SPATIO-TEMPORAL GEOSPATIAL DATA

The coverage and volume of geo-referenced datasets are extensive and incessantly¦growing. The systematic capture of geo-referenced information generates large volumes¦of spatio-temporal data to be analyzed. Clustering and visualization play a key¦role in the exploratory data analysis and the extraction of knowledge embedded in¦these data. However, new challenges in visualization and clustering are posed when¦dealing with the special characteristics of this data. For instance, its complex structures,¦large quantity of samples, variables involved in a temporal context, high dimensionality¦and large variability in cluster shapes.¦The central aim of my thesis is to propose new algorithms and methodologies for¦clustering and visualization, in order to assist the knowledge extraction from spatiotemporal¦geo-referenced data, thus improving making decision processes.¦I present two original algorithms, one for clustering: the Fuzzy Growing Hierarchical¦Self-Organizing Networks (FGHSON), and the second for exploratory visual data analysis:¦the Tree-structured Self-organizing Maps Component Planes. In addition, I present¦methodologies that combined with FGHSON and the Tree-structured SOM Component¦Planes allow the integration of space and time seamlessly and simultaneously in¦order to extract knowledge embedded in a temporal context.¦The originality of the FGHSON lies in its capability to reflect the underlying structure¦of a dataset in a hierarchical fuzzy way. A hierarchical fuzzy representation of¦clusters is crucial when data include complex structures with large variability of cluster¦shapes, variances, densities and number of clusters. The most important characteristics¦of the FGHSON include: (1) It does not require an a-priori setup of the number¦of clusters. (2) The algorithm executes several self-organizing processes in parallel.¦Hence, when dealing with large datasets the processes can be distributed reducing the¦computational cost. (3) Only three parameters are necessary to set up the algorithm.¦In the case of the Tree-structured SOM Component Planes, the novelty of this algorithm¦lies in its ability to create a structure that allows the visual exploratory data analysis¦of large high-dimensional datasets. This algorithm creates a hierarchical structure¦of Self-Organizing Map Component Planes, arranging similar variables' projections in¦the same branches of the tree. Hence, similarities on variables' behavior can be easily¦detected (e.g. local correlations, maximal and minimal values and outliers).¦Both FGHSON and the Tree-structured SOM Component Planes were applied in¦several agroecological problems proving to be very efficient in the exploratory analysis¦and clustering of spatio-temporal datasets.¦In this thesis I also tested three soft competitive learning algorithms. Two of them¦well-known non supervised soft competitive algorithms, namely the Self-Organizing¦Maps (SOMs) and the Growing Hierarchical Self-Organizing Maps (GHSOMs); and the¦third was our original contribution, the FGHSON. Although the algorithms presented¦here have been used in several areas, to my knowledge there is not any work applying¦and comparing the performance of those techniques when dealing with spatiotemporal¦geospatial data, as it is presented in this thesis.¦I propose original methodologies to explore spatio-temporal geo-referenced datasets¦through time. Our approach uses time windows to capture temporal similarities and¦variations by using the FGHSON clustering algorithm. The developed methodologies¦are used in two case studies. In the first, the objective was to find similar agroecozones¦through time and in the second one it was to find similar environmental patterns¦shifted in time.¦Several results presented in this thesis have led to new contributions to agroecological¦knowledge, for instance, in sugar cane, and blackberry production.¦Finally, in the framework of this thesis we developed several software tools: (1)¦a Matlab toolbox that implements the FGHSON algorithm, and (2) a program called¦BIS (Bio-inspired Identification of Similar agroecozones) an interactive graphical user¦interface tool which integrates the FGHSON algorithm with Google Earth in order to¦show zones with similar agroecological characteristics.

Contrasting tree-ring data with fire record in a pine-dominated landscape in the Komi Republic (Eastern European Russia) : recovering a common climate signal

Abstract