This dissertation studies the diversity multiplexing tradeoff and the capacity of wireless multiple-relay network. In part 1, we study the setup of the parallel Multi-Input Multi-Output (MIMO) relay network. An amplify-and-forward relaying scheme, Incremental Cooperative Beamforming, is introduced and shown to achieve the capacity of the network in the asymptotic case of either the number of relays or the power of each relay goes to infinity. In part 2, we study the general setup of multi-antenna multi-hop multiple- relay network. We propose a new scheme, which we call random sequential (RS), based on the amplify-and-forward relaying. Furthermore, we derive diversity- multiplexing tradeoff (DMT) of the proposed RS scheme for general single-antenna multiple-relay networks. It is shown that for single-antenna two-hop multiple- access multiple-relay (K > 1) networks (without direct link between the source(s) and the destination), the proposed RS scheme achieves the optimum DMT. In part 3, we characterize the maximum achievable diversity gain of the multi- antenna multi-hop relay network and we show that the proposed RS scheme achieves the maximum diversity gain. In part 4, RS scheme is utilized to investigate DMT of the general multi-antenna multiple-relay networks. First, we study the case of a multi-antenna full-duplex single-relay two-hop network, for which we show that the RS achieves the optimum DMT. Applying this result, we derive a new achievable DMT for the case of multi-antenna half-duplex parallel relay network. Interestingly, it turns out that the DMT of the RS scheme is optimum for the case of multi-antenna two parallel non-interfering half-duplex relays. Furthermore, we show that random unitary matrix multiplication also improves the DMT of the Non-Orthogonal AF relaying scheme in the case of a multi-antenna single relay channel. Finally, we study the general case of multi-antenna full-duplex relay networks and derive a new lower-bound on its DMT using the RS scheme. Finally, in part 5, we study the multiplexing gain of the general multi-antenna multiple-relay networks. We prove that the traditional amplify-forward relaying achieves the maximum multiplexing gain of the network. Furthermore, we show that the maximum multiplexing gain of the network is equal to the minimum vertex cut-set of the underlying graph of the network, which can be computed in polynomial time in terms of the number of network nodes. Finally, the argument is extended to the multicast and multi-access scenarios.

Duomenų „surinkimas“ – tai procesai, skirti tam tikros informacijos, apie dominančią objektinę sritį, kaupimui. Tokių duomenų „apdorojimo“ sąvoka apima jų filtravimą, priežiūrą, saugojimą, analizę ir pan. Kriterijai pagal kuriuos duomenys yra renkami, taip pat metodika ir filtravimas juos padaro specifiniais arba „paskirstytais“. Paskirstytų duomenų surinkimo ir apdorojimo sąvoka yra gana plati ir apima ne tik elementariausius Web klausimynus ar testus, bet ir dideles sistemas, skirtas analizuoti ir apdoroti struktūrizuoto (pvz.: XML) duomenų srauto informaciją, prireikus ją konvertuoti, saugoti ir adaptuoti. / Term WIKI defines ideas and basis for data collection and management, necessary and mandatory methods required to ensure quality of collectible data. The purpose of this master work is analysis of existing solutions of data collecting and analysis systems in order to create a new unique WIKI community based system for price and other information (related to products on the market) analysis. System target – build up user manageable knowledge base for recognition of changes in the market, fast, easy and cheapest products information withdraw. It gives an overview of the best data collection approaches structure, techniques and logic, data analysis and collection methods (based on the system target), also proposes created model and defines means and tools, for model creation and realization.

Dinaminiai kelio paieškos algoritmai apjungia euristinės ir plečiamos (angl. incremental) paieškos metodus, sprendžiant eiles panašių paieškos uždavinių tiek žinant visą informaciją apie aplinką, tiek neturint jokios informacijos. Yra trys plečiamą paiešką naudojančių algoritmų klasės. Šiame darbe pateikiama trumpa dinaminių kelio paieškos algoritmų, naudojančių plečiamos paieškos metodus analizė. Pagrindinis darbo tikslas – visų trijų plečiamos paieškos klasių algoritmų, kuriuos naudojant gaunamas optimalus kelio paieškos sprendinys aplinkose, kur perėjimų tarp viršūnių svoriai gali didėti ir mažėti, palyginimas. Algoritmai lyginami trijose skirtingose situacijose: stacionarioje, judėjimo link tikslo (angl. goal – directed) bei judančio taikinio (angl. moving – target). Tyrimo rezultatai parodė, jog A* ir FSA* nepasiekiamų viršūnių kiekiui esant ~16000 yra ~23,6% našesni už GAA* ir trečios plečiamos klasės algoritmus, o pasiekiamumą keičiančių viršūnių kiekiui esant ~8000 – 42,3%. Nepasiekiamų viršūnių kiekiui kintant nuo 1000 iki 16000 trečios plečiamos klasės algoritmai yra vidutiniškai ~58,7% našesni už GAA* ir ~54% našesni už A* ir FSA*. Pasiekiamumą keičiančių viršūnių kiekiui kintant nuo 500 iki 8000 trečios plečiamos klasės algoritmai yra vidutiniškai ~69,3% našesni už GAA* ir ~47,8% našesni už A* ir FSA*. / Dynamic path finding algorithms combine heuristic and incremental search methods to solve a series of similar search tasks in both known and unknown environments. There are three classes of incremental search algorithms. In this document we provide a brief summary of dynamic path finding algorithms, that uses incremental search methods, but its main focus is on comparing main algorithms of all three incremental classes, that are guaranteed to give optimal solution in environment where action costs can increase and decrease over time, and showing their strong and weak sides. The algorithms are compared in three different situations: stationary, goal – directed and moving – target. At the end of the document conclusions are given based on performed work. In this paper, research showed that A* and FSA* are ~23,6% more efficient than GAA* and third incremental class algorithms when the amount of untraversable cells is ~16000 and ~42,3% more efficient when the amount of traversability changing cells is ~8000. When the amount of untraversable cells is between 1000 and 16000, the third incremental class algorithms are ~58,7% more efficient than GAA* and ~54% – than A* and FSA*. When the amount of traversability changing cells is between 500 and 8000, the third incremental class algorithms are ~69,3% more efficient than GAA* and ~47,8% – than A* and FSA*.

No description available.

Robots moving about in complex environments must be capable of determining and performing difficult motion sequences to accomplish tasks. As the tasks become more complicated, robots with greater dexterity are required. An increase in the number of degrees of freedom and a desire for autonomy in uncertain environments with real-time requirements leaves much room for improvement in the current popular robot motion planning algorithms. In this thesis, state of the art robot motion planning techniques are surveyed. A solution to the general movers problem in the context of motion planning for robots is presented. The proposed robot motion planner solves the general movers problem using a sample-based tree planner combined with an incremental simulator. The robot motion planner is demonstrated both in simulation and the real world. Experiments are conducted and the results analyzed. Based on the results, methods for tuning the robot motion planner to improve the performance are proposed.

Thesis (MComm (Industrial Psychology))--University of Stellenbosch, 2009. / ENGLISH ABSTRACT: The last two decades have witnessed an upsurge in the research and use of psychometric tests to aid in the prediction of managerial performance. Currently the most prevailing predictor constructs of managerial performance are cognitive ability, personality, and experience. However, researchers and practitioners are still looking for ways in which to maximise the prediction of managerial performance. In recent years, Situational Judgement Tests (SJTs) have become an increasingly popular selection tool. SJTs are multidimensional psychometric instruments designed to assess an individual’s judgement concerning work-related situations. Evidence to date indicates that SJTs are valid predictors of performance, especially for managerial positions in which interpersonal interactions are important. The main objective of this study was to examine whether SJTs significantly add to the prediction of managerial performance over other measures used for managerial selection, such as measures of cognitive ability and personality. Measures of specific cognitive abilities, personality and a SJT were administered to branch managers in a South African retail bank (N = 124) to investigate the ability of the measures to predict managerial performance. Managerial performance was measured using three measures; Performance Ranking, a Behavioural Observation Scale (BOS) and an Overall Performance Rating. Hierarchical multiple regression was used to investigate the relationship between the predictor composites and the managerial performance measures. Findings reveal different prediction patterns for the three criteria. A multiple correlation coefficient of .442 (p > .05) was obtained when predicting Performance Ranking measures, .308 (p < .05) was obtained for predicting the Behavioural Observation Scale (BOS) measure, and .318 (p > .05) was obtained when predicting the Overall Performance Rating measure. Therefore, only when predicting the BOS measure, the SJT provided incremental validity over cognitive ability and personality measures. Consequently, the average of the scores of the three criterion measures, i.e., the Managerial Performance Composite, was used to evaluate the a priori hypotheses. A multiple correlation of .366 (p > .05) was obtained for predicting the Managerial Performance Composite criterion. Results therefore indicate that the SJT did not exhibit meaningful or statistically significant incremental prediction over cognitive ability and personality to predict the composite managerial performance measure. / AFRIKAANSE OPSOMMING: Die laaste twee dekades het ‘n toename in die gebruik van psigometriese toetse in die voorspelling van bestuurdersprestasie waargeneem. Tans is kognitiewe vermoë, persoonlikheid en ervaring die mees algemene voorspellingskonstrukte vir bestuurdersprestasie. Navorsers en praktisyns is egter op soek na maniere om die voorspelling van bestuurdersprestasie te verbeter. ‘n Onlangse verwikkeling is dat “Situational Judgement Tests” (SJTs) toeneem in gewildheid as seleksie-metode. SJTs is multi-dimensionele psigometriese toetse wat ontwerp is om ‘n individu se oordeelsvermoë ten opsigte van werksverwante situasies te assesseer. Navorsing tot op hede wys dat SJTs geldige voorspellers van prestasie is, veral vir bestuursposisies waarin interpersoonlike interaksies belangrik is. Die hoofdoel van hierdie studie was om te ondersoek of SJTs betekenisvolle waarde toevoeg tot die voorspelling van bestuurdersprestasie bo die gebruik van ander meetinstrumente wat vir bestuurskeuring gebruik word, soos metings van kognitiewe vermoë en persoonlikheid. Vir hierdie doel, is takbestuurders in ‘n Suid Afrikaanse bank (N = 124) se kognitiewe vermoëns, persoonlikheid en situasionele beoordelingsvermoë getoets om die vermoë van die meetinstrumente om bestuurdersprestasie te voorspel, te ondersoek. Bestuurdersprestasie was deur middel van drie meetinstrumente bepaal; prestasie-rangordening (“Performance Ranking”), ‘n gedragsobservasieskaal (“Behavioural Observation Scale”) en ‘n algehele prestasiebeoordelingsmeting (“Overall Performance Rating”). Hiërargiese meervoudige regressie-ontleding was gebruik om die verhouding tussen die voorspellers en die bestuurdersprestasiemetings te ondersoek. Verskillende voorspellingspatrone is vir die drie meetinstrumente gevind. ‘n Meervoudige korrelasie koeffisiënt van .308 (p < .05) is vir die voorspelling van die BOS meting verkry, terwyl .442 (p > .05) en .308 (p < .05) onderskeidelik vir die voorspelling van die prestasie-rangordening en algehele prestasiebeoordelingsmeting verkry is. Gevolglik het slegs die BOS meting inkrementele geldigheid getoon. Die gemiddeld van hierdie drie metings se tellings is gebruik om ‘n bestuurdersprestasie-kombinasietelling “Managerial Performance Composite” te skep wat gebruik is om die finale besluit rakende die a priori hipoteses te maak. ‘n Meervoudige korrelasie van .366 (p >.05) was gevind ten einde die bestuurdersprestasie-kombinasietelling te voorspel aan die hand van die voorspellers.

Thesis (MScEng)--Stellenbosch University, 2013. / ENGLISH ABSTRACT: Osteoarthritis (OA), also known as degenerative joint disease is a progressive disorder of the joints caused by gradual loss of cartilage and resulting in the development of bony spurs and cysts at the margins of the joints. The degradation of the musculoskeletal system, which is mainly caused by joint injury, obesity (leading to musculoskeletal fatigue) and aging can also lead to osteoarthritis. The hands, feet, spine, and large weight-bearing joints, such as the hips and knees are commonly affected. The only medical solution to severe cases of osteoarthritis is the surgical reconstruction or replacement of a malformed or degenerated joint, better known as arthroplasty. Arthroplasty makes use of biomedical implants and replacements to restore functionality of the joints. Biomedical engineering in arthroplasty is an ever increasing field of interest as a result of its innovative improvements to surgical quality. Certain cases of partial osteoarthritis require less surgical action. Partial knee replacement surgery, also known as unicondylar (or unicompartmental) knee arthroplasty involves a covering which is placed over the affected area to resurface the affected bone and protect the patient from further degeneration. Advantages of partial replacement include faster recovery time and less post-operative pain. The biomedical implants used for these operations consist of a standardized implant that is fit onto the bone by modifying (cutting away) the outer structure of the bone. The result is known to cause post-operative discomfort among some patients. The problem with these standard designs includes the requirement of the removal of unaffected (healthy) bone matter, leading to induced trauma and pain for patients during the recovery phase of the operations. A preferred alternative to the standard design would be to create a custom implant for every patient, reducing the need to remove parts of unaffected bone matter. The implementation of this proposed method tends toward Minimally Invasive Surgery (MIS). MIS is normally preferred as it reduces the risk of various negative consequences of normal arthroplasty such as nerve or tendon damage during surgery. It could be argued that the proposed method may cause less damage to the fragile tendon, bloodflow, and nerve networks of the knee. Increasing material costs of metal products introduce great interest in more cost efficient forming processes to reduce the loss of redundant blank material. Incremental Sheet Forming (ISF), a relatively new class of forming process, has the potential to meet the need for this more efficient forming process. The ISF process is highly flexible, can be developed in normal milling machines, and can reduce production cost by up to 90% in comparison to processes such as stamping. The ISF process is a non-patented process, as the existing patents are referred to the designed machines and not the process. The availability of the ISF process contributes greatly to its attractiveness. ISF can be implemented in any facility that has access to a three- or more-axis CNC machine. The advantage of ISF implemented in CNC machines is that CNC technology has already reached a mature stage in development, contributing to the accuracy and methodology (such as feed rate or angular velocity of the tool) of the ISF process. The forming of valuable lightweight materials is well covered by ISF processes. A variety of studies contribute to research on the forming of titanium and titanium based alloys as part of ISF of lightweight materials. The ISF process utilizes the functionality of commercial CNC machines, improving the process availability of many manufacturing companies. The ISF process offers fast setup times and flexibility of the forming process. The purpose of this project is to define a process chain for creating a customized biomedical implant as well as determining the validity of the process chain by applying each step. The design and development procedure of a titanium based biomedical arthroplasty implant using innovative Incremental Sheet Forming (ISF) techniques will be documented, as well as an investigation of the financial cost and potential gain that this implant can offer. / AFRIKAANSE OPSOMMING: Osteoartritis is 'n gewrig siekte wat degeneratiewe newe-effekte behels in die gewrigte. Hierdie siekte lei to die geleidelike verlies van kraakbeen en lei tot die onreelmatige ontwikkeling van abnormale beengroei. Osteoartritis kan ook deur beserings in die gewrig veroorsaak word. Die hande, voete, ruggraat, en enige groter gewigdraende gewrigte, soos die heupe en knieë kan geaffekteer word. Die enigste mediese oplossing tot ernstige gevalle van die siekte is chirurgiese rekonstruksie of vervanging van die gewrig, meer bekend as artroplastie. Artroplastie maak gebruik van biomediese implantate om funksionaliteit van die gewrig te herstel. Biomediese ingenieurswese in artroplastie is 'n toenemende navorsingsveld as gevolg van die innoverende aspekte om chirurgiese kwaliteit te verhoog. Sekere gevalle van gedeeltelike osteoartritis vereis veel minder chirurgiese behandeling. Gedeeltelike knie vervanging chirurgie, meer bekend as unikompartementele knie artroplastie, behels 'n bedekking wat slegs die geaffekteerde been bedek, om die pasiënt van verdere degenerasie te beskerm. Voordele van gedeeltelike vervanging sluit vinniger herstel tyd en minder pyn in. Die biomediese implantate wat gebruik word vir hieride operasies bestaan uit standaard ontwerpe wat aan die been gepas word deur die wysiging (of wegsny) van die buitenste beenstruktuur. Die nagevolg van hierdie chirurgie is lang herstel periodes en kan ongemaklikheid in die knie veroorsaak. Die probleem met die bogenoemde standaard is dat die prosedure die verweidering van selfs ongeaffekteerde (of gesonde) been in sluit, wat lei tot verdere kniepyn en ongemak vir pasiënte lei tydens die herstelperiode. 'n Verkiesde alternatief tot die standaard ontwerpe is om 'n persoonlikke implantaat vir elke pasiënt te skep, en so kan die behoefte om dele van ongeaffekteerde been te behou moontlik wees. Die toepassing van die voorgestelde metode neig na Minimale Skade Chirurgie (MSC). MSC word gewoonlik verkies om die risiko van verskeie negatiewe nagevolge te verminder, en skade aan die tendon, bloed- en senunetwerke van die knie te beperk. Die toenemende materiaalkoste vand metal produkte lei tot 'n groot belangstelling in meer koste besparing vormings prosesse, om sodoende die verlies van oortollige materiaalverlies te verminder. Inkrementele Plaat Vervorming (IPV), 'n relatiewe nuwe klas van vervorming, is 'n waardige kanidaat om hierdie doel te bereik. Die IPV proses is baie toepaslik, en kan deur die gebruik van Rekenaar Numeriese Kontrole (RNK) masjienerie toegepas word. Verder sal dit vervaardigingskoste kan verlaag met soveel as 90% in vergelyking met ander prosese soos die stempel metode. Die beskikbaarheid van die IPV proses dra grootliks by tot die proses se aantreklikheid in die industrie. IPV kan geimplementeer word in enige fasiliteit wat toegang tot 'n drie-as RNK masjien het. Die voordeel van dit is die feit dat RNK masjienerie klaar ontwikkel en volwasse is, wat kan bydra tot goeie akkuraatheid in die vormingsproses. Die vervaardiging van laegewig materiale soos titaan of aluminium is gedokumenteer. 'n Verskeidenheid van studies dra waarde tot navorsing van die vormingsproses van titaan as deel hiervan. Die IPV proses bied vinnige opstel tye en goeie buigsaamheid met die vormingsproses, veral met behulp van 'n vyf-as masjien. Die doel van hierdie projek is om 'n proses ketting te ontwerp. Die proses ketting, wat uit vele stappe bestaan, sal die ontwerp en vervaardigingsproses van 'n persoonlike biomediese knie implantaat bevestig deur middel van die IPV vormings tegniek. Validasie van die proses ketting sal dus plaasvind deur die stappe van die voorgestelde proses ketting uit te voer. 'n Finale ondersoek sal die finansiele en regalutoriese aspekte van die projek aanspreek.

Orientador: Jozué Vieira Filho / Banca: Carlos Antonio Alves / Banca: Tony Inácio da Silva / Resumo: A automação de esteiras rolantes é algo comum e importante em sistemas industriais, mas problemas práticos ainda representam desafios. Um dos desses desafios é manter a precisão em sistemas que exigem paradas sistemáticas, pois folgas mecânicas tendem a provocar variações nas posições de paradas ao longo do tempo. A aplicação de motores de indução têm se tornado comum e soluções eficientes e de baixo custo têm sido pesquisadas. Neste trabalho foi desenvolvido e implementado um sistema de controle de posição e velocidade aplicado em esteiras transportadoras utilizando inversor de frequência, microcontrolador, encoder óptico incremental e sensor indutivo. O movimento da esteira transportadora é efetuado por um motor de indução trifásico, que é acionado pelo conjunto microcontrolador - inversor de frequência. Este conjunto impõe uma frequência no estator do motor através de uma troca de mensagens entre microcontrolador e inversor de frequência (Sistema Mestre-Escravo). Para o envio e recebimento das mensagens, utilizou-se o protocolo de comunicação serial USS® (Universal Serial Interface Protocol) através do padrão RS-485. Os controles de posição e velocidade de rotação do eixo do motor fundamentam-se no sinal gerado pelo encoder óptico incremental, responsável por informar a posição do eixo do motor ao longo da trajetória, e no sensor indutivo que determina uma referência externa importante para a esteira transportadora. Para o funcionamento automático da esteira, elaborou-se um software em linguagem de programação C. Como resultado obteve-se um sistema de controle de posição e velocidade do eixo do motor de indução trifásico que apresenta bons resultados / Abstract: Automated conveyors system have been largely used in industrial applications. However, there are still practical issues to be overcome. One of them is due to the system mechanical limitation which can lead to low accuracy for applications based on "stop-and-go" movements. Induction motors have been largely used in such applications and low costs solutions have been searched. In this work it was developed and implemented a system of positioning and velocity control applied to conveyors which is based on frequency inverter, microcontroller, optical incremental encoder and inductive sensor. The conveyor's movement is made by means of a three-phase induction motor, which is driven by the couple microcontroller-frequency inverter. There are messages exchange between the microcontroller and the frequency inverter (Master - Slave configuration) which is based on the communication serial protocol USS through the RS-485 standard. The position and velocity of the motor spindle are controlled using an optical incremental encoder, which is responsible to provide the position of the trajectory, and an inductive sensor which determines the initial reference to the conveyor. The software used to control the system was developed in C language. The results show a low cost system with good results / Mestre

This thesis investigates the role of linguistically-motivated generative models of syntax and semantic structure in natural language processing (NLP). Syntactic well-formedness is crucial in language generation, but most statistical models do not account for the hierarchical structure of sentences. Many applications exhibiting natural language understanding rely on structured semantic representations to enable querying, inference and reasoning. Yet most semantic parsers produce domain-specific or inadequately expressive representations. We propose a series of generative transition-based models for dependency syntax which can be applied as both parsers and language models while being amenable to supervised or unsupervised learning. Two models are based on Markov assumptions commonly made in NLP: The first is a Bayesian model with hierarchical smoothing, the second is parameterised by feed-forward neural networks. The Bayesian model enables careful analysis of the structure of the conditioning contexts required for generative parsers, but the neural network is more accurate. As a language model the syntactic neural model outperforms both the Bayesian model and n-gram neural networks, pointing to the complementary nature of distributed and structured representations for syntactic prediction. We propose approximate inference methods based on particle filtering. The third model is parameterised by recurrent neural networks (RNNs), dropping the Markov assumptions. Exact inference with dynamic programming is made tractable here by simplifying the structure of the conditioning contexts. We then shift the focus to semantics and propose models for parsing sentences to labelled semantic graphs. We introduce a transition-based parser which incrementally predicts graph nodes (predicates) and edges (arguments). This approach is contrasted against predicting top-down graph traversals. RNNs and pointer networks are key components in approaching graph parsing as an incremental prediction problem. The RNN architecture is augmented to condition the model explicitly on the transition system configuration. We develop a robust parser for Minimal Recursion Semantics, a linguistically-expressive framework for compositional semantics which has previously been parsed only with grammar-based approaches. Our parser is much faster than the grammar-based model, while the same approach improves the accuracy of neural Abstract Meaning Representation parsing.

Cette thèse s’intéresse à la problématique de la classification dynamique en environnements stationnaires et non stationnaires, tolérante aux variations de quantités des données d’apprentissage et capable d’ajuster ses modèles selon la variabilité des données entrantes. Pour cela, nous proposons une solution faisant cohabiter des classificateurs one-class SVM indépendants ayant chacun leur propre procédure d’apprentissage incrémentale et par conséquent, ne subissant pas d’influences croisées pouvant émaner de la configuration des modèles des autres classificateurs. L’originalité de notre proposition repose sur l’exploitation des anciennes connaissances conservées dans les modèles de SVM (historique propre à chaque SVM représenté par l’ensemble des vecteurs supports trouvés) et leur combinaison avec les connaissances apportées par les nouvelles données au moment de leur arrivée. Le modèle de classification proposé (mOC-iSVM) sera exploité à travers trois variations exploitant chacune différemment l’historique des modèles. Notre contribution s’inscrit dans un état de l’art ne proposant pas à ce jour de solutions permettant de traiter à la fois la dérive de concepts, l’ajout ou la suppression de concepts, la fusion ou division de concepts, tout en offrant un cadre privilégié d’interactions avec l’utilisateur. Dans le cadre du projet ANR DIGIDOC, notre approche a été appliquée sur plusieurs scénarios de classification de flux d’images pouvant survenir dans des cas réels lors de campagnes de numérisation. Ces scénarios ont permis de valider une exploitation interactive de notre solution de classification incrémentale pour classifier des images arrivant en flux afin d’améliorer la qualité des images numérisées. / This research contributes to the field of dynamic learning and classification in case of stationary and non-stationary environments. The goal of this PhD is to define a new classification framework able to deal with very small learning dataset at the beginning of the process and with abilities to adjust itself according to the variability of the incoming data inside a stream. For that purpose, we propose a solution based on a combination of independent one-class SVM classifiers having each one their own incremental learning procedure. Consequently, each classifier is not sensitive to crossed influences which can emanate from the configuration of the models of the other classifiers. The originality of our proposal comes from the use of the former knowledge kept in the SVM models (represented by all the found support vectors) and its combination with the new data coming incrementally from the stream. The proposed classification model (mOC-iSVM) is exploited through three variations in the way of using the existing models at each step of time. Our contribution states in a state of the art where no solution is proposed today to handle at the same time, the concept drift, the addition or the deletion of concepts, the fusion or division of concepts while offering a privileged solution for interaction with the user. Inside the DIGIDOC project, our approach was applied to several scenarios of classification of images streams which can correspond to real cases in digitalization projects. These different scenarios allow validating an interactive exploitation of our solution of incremental classification to classify images coming in a stream in order to improve the quality of the digitized images.