The Fourth Industrial Revolution is based on the so called Digital Transformation. Technology has advanced by leaps and bounds over the past few years, and this has led to the creation of various smart devices. The push for digitalization comes mainly from the industrial sector; companies are called to compete in different ways, focusing not only on the quality of products, but also on services and innovation. The impact of digital processing on business models has been huge. The way in which both customers and suppliers interact is changing at a very fast pace thanks to the ever increasing implementation of articulated and automated process within the supply chain. Products and services are changing, renewing and repositioning themselves thanks to the inclusion of features that are both more functional and user-friendly which are made available by a plethora of digital applications. The competitive scenario and the way in which business have to respond is evolving constantly; companies owe their success predominantly to the efficiency of their supply chain. This brought about a profound change in demand in all major user sectors, from Finance to Industry, from Distribution to Utilities. Demand is now more focused than ever on the specific features and qualities of innovative components, such as Digital Enabler, which enable the user to interact in a better way and perform innovative operations in the field of digital transformation. In this pervasive world, smart devices such as smartphones, last-generation vehicles, smartwatches, or any Internet of Thing (IoT) devices are becoming ubiquitous and in the majority of cases involve communicating and storing information with huge database centers called cloud. With the proliferation of IoT devices and the advent of social medias, a huge amount of multimedia data is being generated and most of it is unstructured and multimodal. This sudden rise in the exchange of huge amounts of information has meant a sudden rise of demand for computation of multimedia data which has in turn created many opportunities for producers and special needs for the users who increasingly require larger storage spaces and accurate data processing. Alongside the proliferation of data, another challenging aspect in the realm of the 4.0 industry is the so called Computer Vision which in large part represents the future of the controversial Machine Learning (ML) and Deep Learning (DL) research - the basis of automatized learning that could allow in the near future machines to take their own decisions. This process will require various computationally efficient techniques to make use of this data in a meaningful manner. In the field of ML and DL there are Artificial Neural Networks, mathematical models that draw inspiration from the function that make a biological brain work. Most of the computer vision applications revolve around Convolutional Neural Network (CNN) architectures, a typology of artificial neural network based on the vision system of human beings. Computer Vision plays a fundamental role for Classification issues, which appear when the network is to be trained to recognize the various classes of data that are provided during the input process. A classification model is mainly based on three sets of data: the Training set, in which images are situated and used to train the network; the Validation set, which contains the images used to evaluate the model performance and finally the Test set where the images are located to test the model. The GO0DMAN project, a European project which regards the development of a smart portable laser scanner for Gap&Flash measurement (G3F), fits well into this context. The G3F device is intended to be used in an automotive production line for quality control...
La Quarta Rivoluzione Industriale si basa sulla cosiddetta Trasformazione Digitale. La tecnologia ha fatto passi da gigante negli ultimi anni e questo ha portato alla creazione di vari dispositivi intelligenti. La spinta alla digitalizzazione arriva principalmente dal settore industriale; le aziende sono chiamate a competere in modi diversi, puntando non solo sulla qualità del prodotto, ma anche sull'innovazione e sui servizi. L'impatto del processo digitale sui modelli di business è enorme. Il modo in cui sia i clienti che i fornitori interagiscono tra loro sta cambiando rapidamente grazie alla sempre maggior implementazione di un processo produttivo articolato e automatizzato all'interno della catena di fornitura. Prodotti e servizi stanno cambiando, rinnovandosi e riposizionandosi grazie all'inserimento di dispositivi al tempo stesso più funzionali e user-friendly messi a disposizione da una vasta gamma di applicazioni digitali. Lo scenario competitivo e il modo in cui le imprese devono rispondere è in continua evoluzione; le aziende devono il loro successo principalmente all'efficienza della loro supply chain. Ciò ha portato ad un profondo cambiamento nella domanda in tutti i principali settori di utenza, dalla finanza all'industria, dalla distribuzione ai servizi di pubblica utilità. Una domanda che ora è più che mai focalizzata sulle caratteristiche e sulle specifiche qualità di nuovi componenti, chiamati Digital Enabler, in quanto consentono all'utente di interagire e proiettarsi sempre più verso l'utilizzo di operazioni innovative nel campo della trasformazione digitale. In questo mondo pervasivo, dispositivi intelligenti come smartphone, veicoli di ultima generazione, smartwatch o un qualsiasi altro dispositivo Internet of Thing (IoT) stanno diventando onnipresenti e nella maggior parte dei casi riguardano la comunicazione e l'archiviazione di informazioni in specifici database chiamati cloud. Con la proliferazione dei dispositivi IoT e l'avvento dei social media, viene generata un'enorme quantità di dati e la maggior parte di essi è non strutturata e multimodale. Questo improvviso aumento di scambio di una notevole quantità di informazioni ha comportato la necessità di un'accurata analisi dei dati generati, che a sua volta ha creato molte opportunità per i produttori e specifiche esigenze per gli utenti, che richiedono sempre più ampi spazi di archiviazione e un'accurata elaborazione dei dati. Accanto alla proliferazione dei dati, un altro aspetto fondamentale e stimolante nel regno dell'Industry 4.0 è la cosiddetta Computer Vision, che in gran parte rappresenta il futuro della controversa ricerca nel campo del Machine Learning (ML) e del Deep Learning (DL) - la base dell'apprendimento automatico che nel prossimo futuro potrebbe consentire alle macchine di prendere le proprie decisioni. Questo processo richiede tecniche computazionalmente sempre più efficienti per utilizzare questi dati in modo significativo. Nel campo del ML e del DL, un ruolo chiave è giocato dalle Reti Neurali Artificiali, le Artificial Neural Networks, modelli matematici che traggono ispirazione dalla funzione del cervello biologico. La maggior parte delle applicazioni di visione artificiale ruota attorno alle architetture delle Reti Neurali Convoluzionali, le Convolutional Neural Network (CNN), una tipologia di rete neurale artificiale basata sul sistema di visione degli esseri umani. La visione artificiale gioca un ruolo fondamentale per i problemi di Classificazione, che compaiono quando la rete deve essere addestrata a riconoscere le varie classi di dati che vengono fornite in input. Un modello di classificazione si basa principalmente su tre set di dati: il Training set, in cui si trovano le immagini utilizzate per addestrare la rete; il Validation set, che contiene le immagini usate per valutare le prestazioni del modello e infine il Test set, con le immagini usate per testare il modello...
Sviluppo di un algoritmo di Deep-learning per il riconoscimento automatico dell'area inquadrata dalla telecamera di un dispositivo portatile per la rilevazione del Gap&Flush azionato da operatori umani
CALCABRINI, SAMUELE
2020/2021
Abstract
The Fourth Industrial Revolution is based on the so called Digital Transformation. Technology has advanced by leaps and bounds over the past few years, and this has led to the creation of various smart devices. The push for digitalization comes mainly from the industrial sector; companies are called to compete in different ways, focusing not only on the quality of products, but also on services and innovation. The impact of digital processing on business models has been huge. The way in which both customers and suppliers interact is changing at a very fast pace thanks to the ever increasing implementation of articulated and automated process within the supply chain. Products and services are changing, renewing and repositioning themselves thanks to the inclusion of features that are both more functional and user-friendly which are made available by a plethora of digital applications. The competitive scenario and the way in which business have to respond is evolving constantly; companies owe their success predominantly to the efficiency of their supply chain. This brought about a profound change in demand in all major user sectors, from Finance to Industry, from Distribution to Utilities. Demand is now more focused than ever on the specific features and qualities of innovative components, such as Digital Enabler, which enable the user to interact in a better way and perform innovative operations in the field of digital transformation. In this pervasive world, smart devices such as smartphones, last-generation vehicles, smartwatches, or any Internet of Thing (IoT) devices are becoming ubiquitous and in the majority of cases involve communicating and storing information with huge database centers called cloud. With the proliferation of IoT devices and the advent of social medias, a huge amount of multimedia data is being generated and most of it is unstructured and multimodal. This sudden rise in the exchange of huge amounts of information has meant a sudden rise of demand for computation of multimedia data which has in turn created many opportunities for producers and special needs for the users who increasingly require larger storage spaces and accurate data processing. Alongside the proliferation of data, another challenging aspect in the realm of the 4.0 industry is the so called Computer Vision which in large part represents the future of the controversial Machine Learning (ML) and Deep Learning (DL) research - the basis of automatized learning that could allow in the near future machines to take their own decisions. This process will require various computationally efficient techniques to make use of this data in a meaningful manner. In the field of ML and DL there are Artificial Neural Networks, mathematical models that draw inspiration from the function that make a biological brain work. Most of the computer vision applications revolve around Convolutional Neural Network (CNN) architectures, a typology of artificial neural network based on the vision system of human beings. Computer Vision plays a fundamental role for Classification issues, which appear when the network is to be trained to recognize the various classes of data that are provided during the input process. A classification model is mainly based on three sets of data: the Training set, in which images are situated and used to train the network; the Validation set, which contains the images used to evaluate the model performance and finally the Test set where the images are located to test the model. The GO0DMAN project, a European project which regards the development of a smart portable laser scanner for Gap&Flash measurement (G3F), fits well into this context. The G3F device is intended to be used in an automotive production line for quality control...File | Dimensione | Formato | |
---|---|---|---|
TESI.pdf
Open Access dal 21/07/2023
Descrizione: Documento di tesi
Dimensione
1.8 MB
Formato
Adobe PDF
|
1.8 MB | Adobe PDF | Visualizza/Apri |
I documenti in UNITESI sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/20.500.12075/94