Publicação
Autostereoscopic Head-Up Display: picture generating unit’s performance increase
| Resumo: | Nowadays three-dimensional (3D) displays are starting to be used in many areas such as: cinema, Head-Up Displays (HUD) for automotive and avionics industry, advertising, video-game industry and many other areas. This master’s dissertation will be focused in an Augmented Reality (AR) HUD for automotive industry in partnership with Bosch Car Multimedia (CM) Portugal [1], with the objective of improving the Picture Generation Unit (PGU) performance off the AR-HUD prototype. At the end, the AR-HUD should be capable to allow the driver to observe 3D objects projected in a car windshield with several depths distances with a rendering time above 60 milliseconds (ms) or 16.7 frames per second (FPS). Based on the above goals, it is necessary to study binocular vison [2] in order to understand how human being perceives 3D vision and to know the several optical parameters associated with it, such as: the Interocular distance, Convergence, Binocular Disparity, Accommodation and others. These parameters will be needed to produce a correct stereo image, which is composed by two images, one corresponding to the view of the right eye and the other corresponding to the view of the left eye. To perceive 3D object in real world, which defines augmented reality, without the use of a special glasses or the need to wear other devices, which defines an Autostereoscopic (AS) Displays that are capable to make such job. Many types of technologies for these displays were developed with different optical goals, such as: Parallax Barrier, Lenticular Lenses or Light Field displays [3]. For this project it was required from Bosch the use of a Lenticular Lenses Display. With the use of an array of a lenticular lens on top of a Liquid Crystal Display (LCD) it is possible to “send” a set of pixels columns for each eye separately, displayed in flat LCD based on the light refraction [4]. To produce the right Pixel Assignment Matrix (PAM), which has the responsibility to guide the stereo image generation to be showed in the 3D Display, several parameters must be in count, which are required by the lenticular lenses configuration, such as: the number of Lens Per Inch (LPI), the lens’s slanted angle, viewing distance, Field Of View (FOV) and others. The PGU receives. as an input, the lenticular lens’s parameters, the optical parameters previously mentioned and the chosen scene, generating a 3D image, which was executed by the Central Unit Processing (CPU), which generates a huge rendering time, around 1.3 seconds which is less than 1 frame per second. The solution, to reduce this rendering time, will pass to migrate the processing from CPU to the Graphical Processing Unit (GPU), which is present in the graphics card. This solution also requires a pre-processing of the PAM which will releases a lot of processing during the executing of the pixel assignment on sub-pixel level. To implement this solution on sub-pixel level, through the GPU programming, it will be use OpenGL libraries [5] and Python [6] 2.7 as programming language. For the rendering process of the stereo image, it will be applied the rasterization method. The lenticular lenses can be used to provide two-view (one point of view) or multi-view (several points of views). For these AR-HUD, it be used for two-view, with an eye-tracking system as future feature system, which will allow the inclination and the adjustment of the 3D display, based on the location of the driver’s eyes. |
|---|---|
| Autores principais: | Raimundo, Guilherme Nuno Henriques de Oliveira Brochado |
| Assunto: | Engenharia e Tecnologia::Engenharia Eletrotécnica, Eletrónica e Informática |
| Ano: | 2018 |
| País: | Portugal |
| Tipo de documento: | dissertação de mestrado |
| Tipo de acesso: | acesso restrito |
| Instituição associada: | Universidade do Minho |
| Idioma: | inglês |
| Origem: | RepositóriUM - Universidade do Minho |
| Resumo: | Nowadays three-dimensional (3D) displays are starting to be used in many areas such as: cinema, Head-Up Displays (HUD) for automotive and avionics industry, advertising, video-game industry and many other areas. This master’s dissertation will be focused in an Augmented Reality (AR) HUD for automotive industry in partnership with Bosch Car Multimedia (CM) Portugal [1], with the objective of improving the Picture Generation Unit (PGU) performance off the AR-HUD prototype. At the end, the AR-HUD should be capable to allow the driver to observe 3D objects projected in a car windshield with several depths distances with a rendering time above 60 milliseconds (ms) or 16.7 frames per second (FPS). Based on the above goals, it is necessary to study binocular vison [2] in order to understand how human being perceives 3D vision and to know the several optical parameters associated with it, such as: the Interocular distance, Convergence, Binocular Disparity, Accommodation and others. These parameters will be needed to produce a correct stereo image, which is composed by two images, one corresponding to the view of the right eye and the other corresponding to the view of the left eye. To perceive 3D object in real world, which defines augmented reality, without the use of a special glasses or the need to wear other devices, which defines an Autostereoscopic (AS) Displays that are capable to make such job. Many types of technologies for these displays were developed with different optical goals, such as: Parallax Barrier, Lenticular Lenses or Light Field displays [3]. For this project it was required from Bosch the use of a Lenticular Lenses Display. With the use of an array of a lenticular lens on top of a Liquid Crystal Display (LCD) it is possible to “send” a set of pixels columns for each eye separately, displayed in flat LCD based on the light refraction [4]. To produce the right Pixel Assignment Matrix (PAM), which has the responsibility to guide the stereo image generation to be showed in the 3D Display, several parameters must be in count, which are required by the lenticular lenses configuration, such as: the number of Lens Per Inch (LPI), the lens’s slanted angle, viewing distance, Field Of View (FOV) and others. The PGU receives. as an input, the lenticular lens’s parameters, the optical parameters previously mentioned and the chosen scene, generating a 3D image, which was executed by the Central Unit Processing (CPU), which generates a huge rendering time, around 1.3 seconds which is less than 1 frame per second. The solution, to reduce this rendering time, will pass to migrate the processing from CPU to the Graphical Processing Unit (GPU), which is present in the graphics card. This solution also requires a pre-processing of the PAM which will releases a lot of processing during the executing of the pixel assignment on sub-pixel level. To implement this solution on sub-pixel level, through the GPU programming, it will be use OpenGL libraries [5] and Python [6] 2.7 as programming language. For the rendering process of the stereo image, it will be applied the rasterization method. The lenticular lenses can be used to provide two-view (one point of view) or multi-view (several points of views). For these AR-HUD, it be used for two-view, with an eye-tracking system as future feature system, which will allow the inclination and the adjustment of the 3D display, based on the location of the driver’s eyes. |
|---|