The presented Thesis starts from the work done by the mOve research group regarding the application of wearable devices in driving conditions in order to enhance the driver's experience using the Augmented Reality technology. Up to now, it was proved that it is possible to realize an algorithm that could work with a general Head Mounted Display in order to project virtual features in a complex blend of a static (vehicle's cockpit) and a dynamic (the road) environments. After this successful proof of concept, we have analyzed if it was feasible to implement similar algorithms to a real device, such as Microsoft Hololens 1. To this purpose, we have performed a research about its state of the art in order to understand what are the limits of this device. After having verified that the Hololens built-in algorithms are not suited for our use-case, we have defined a computer vision approach based on tracking fiducial planar markers fixed to the vehicle's dashboard. Using the latter to define the relative position of holograms placed in the scene and Hololens, we have realized two on-line applications in order to evaluate the performance of the developed algorithms. The first app shows to the driver the braking points of a circuit in order to understand when a strong deceleration has to be performed. The second application projects a virtual instrumental panel fixed in the cockpit, in order to show the vehicle's information to the user while maintaining his gaze on the road. Due to the promising results obtained from these apps, we have started an on-line implementation of the proposed algorithms directly on Hololens.
Questa Tesi ha come punto di partenza il lavoro svolto dal gruppo di ricerca mOve riguardo il miglioramento dell'esperienza di guida del pilota, mediante l'utilizzo delle tecnologie indossabili a Realtà Aumentata. Fino ad ora è stata dimostrata la possibilità di realizzare un algoritmo per un generico dispositivo di visualizzazione "Head Mounted", capace di proiettare elementi virtuali in una complessa combinazione di ambienti statici (come l'abitacolo del veicolo) e dinamici (ad esempio la strada). Dopo il successo di questa prova di concetto, abbiamo analizzato se fosse possibile implementare simili algoritmi su un dispositivo reale, come Hololens 1 di Microsoft. A questo scopo, abbiamo svolto una ricerca riguardante il suo stato dell'arte per capire quali fossero i limiti di questo dispositivo. Dopo aver verificato che gli algoritmi di base integrati in Hololens non sono adatti al nostro caso di utilizzo, abbiamo definito un approccio di computer vision basato sulla localizzazione di marker planari solidali alla plancia del veicolo. Usando questi ultimi per definire la posizione relativa degli ologrammi e gli Hololens, abbiamo realizzato due applicazioni off-line per valutare le performance degli algoritmi proposti in questa Tesi. La prima app mostra al pilota i punti di staccata di un circuito per fargli capire quando frenare con decisione. La seconda applicazione proietta un cruscotto virtuale fisso nell'abitacolo, usato per mostrare lo stato del veicolo al pilota senza dover distogliere lo sguardo dalla strada. Avendo ottenuto dei risultati promettenti, abbiamo iniziato l'implementazione on-line di queste applicazioni direttamente su Hololens.
Development of an augmented reality driver assistance system based on hololens
RIPOLI, GIORGIO
2019/2020
Abstract
The presented Thesis starts from the work done by the mOve research group regarding the application of wearable devices in driving conditions in order to enhance the driver's experience using the Augmented Reality technology. Up to now, it was proved that it is possible to realize an algorithm that could work with a general Head Mounted Display in order to project virtual features in a complex blend of a static (vehicle's cockpit) and a dynamic (the road) environments. After this successful proof of concept, we have analyzed if it was feasible to implement similar algorithms to a real device, such as Microsoft Hololens 1. To this purpose, we have performed a research about its state of the art in order to understand what are the limits of this device. After having verified that the Hololens built-in algorithms are not suited for our use-case, we have defined a computer vision approach based on tracking fiducial planar markers fixed to the vehicle's dashboard. Using the latter to define the relative position of holograms placed in the scene and Hololens, we have realized two on-line applications in order to evaluate the performance of the developed algorithms. The first app shows to the driver the braking points of a circuit in order to understand when a strong deceleration has to be performed. The second application projects a virtual instrumental panel fixed in the cockpit, in order to show the vehicle's information to the user while maintaining his gaze on the road. Due to the promising results obtained from these apps, we have started an on-line implementation of the proposed algorithms directly on Hololens.| File | Dimensione | Formato | |
|---|---|---|---|
|
Thesis_final_revised.pdf
non accessibile
Dimensione
33.11 MB
Formato
Adobe PDF
|
33.11 MB | Adobe PDF | Visualizza/Apri |
I documenti in POLITesi sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
https://hdl.handle.net/10589/165466