Computer generated images in 3D – Les Globes de Mercator de l'UNIL http://wp.unil.ch/mercator/en/ Le récit d'une découverte à l'Université de Lausanne Fri, 30 Nov 2018 12:39:58 +0000 en-US hourly 1 https://wordpress.org/?v=5.5.1 Georeferencing the images http://wp.unil.ch/mercator/en/georeferencing-the-images/ Sun, 31 Jan 2016 11:22:58 +0000 http://wp.unil.ch/mercator/georeferencement-images/ March 2016 By digital georeferencing the high-definition images, a 3D model is created to reproduce each of the spheres virtually [...]]]> Once these high-resolution images were made in the photo lab of the School of Forensic Science (ESC), the next step consisted in georeferencing them digitally. By doing so, they could then be used to create a 3D model to reproduce the globes virtually. Alexandre Hirzel, a specialist at UNIL’s Geography IT Systems (SIG) unit, was put in charge of this task.

Below is an explanation of the various operations involved in modelling the terrestrial globe using the ArcGIS Desktop 1.3 (ArcMap) application. The process was identical for the celestial globe.

1. Graticule2. Optimisation3. Aligning the grids4. Cropping5. Weighting6. Use
Réseau de longitudes (en rouge) décalé de 14.1° vers l'ouest (en bleu).

Alexandre Hirzel © UNIL

A network of longitudes (in red) was created based on Mercator’s prime meridian passing through Fuerteventura in the Canary Islands. This represents a 14.1° westward shift in relation to today’s prime meridian passing through Greenwich (in blue).

Positionnement des 42 images sur certains croisements de longitudes et latitudes.

Alexandre Hirzel © UNIL

The process used to photograph the globes produced 42 images centred on selected longitude and latitude intersections.

Zone utilisable par image pour que la couleur, la lumière et la netteté soient homogènes.

Alexandre Hirzel © UNIL

The colour, lighting and sharpness of the images had to be as uniform as possible. A usable circular area was therefore determined for each one.

Géoréférencement d'une image des images.

Alexandre Hirzel © UNIL

Each rectangular photo was georeferenced so that it could then be laid precisely over the virtual globe.

To correct the distortions due to perspective as much as possible, the globe was displayed using the World Vertical Perspective projection from an altitude of 35,800,000 metres. On scale, this altitude corresponds to the distance from the camera to the globe.

georef-diaporamaThe graticule intersections on each image were then matched with those of the virtual grid. Each of these anchor points made it possible to improve the georeferencing.

Since the globes are not perfectly spherical, a non-parametric alignment using the Spline algorithm yielded better results.

On average about 30 of these anchor points were needed to ensure an optimal alignment.

To highlight this process, in the above animation the point focal is shifted in relation to the centre of the globe. Actually, the globe was centred so that its geometry would resemble the conditions of the image capture as closely as possible.

georef-diaporama2The centre of each image was cropped so as to keep only the most uniform portion in terms of colour, lighting and sharpness.

The cropping process was repeated for all 42 photos.

Even so, the round tiles were not perfectly matched and this was particularly visible at their edges. To reduce these discrepancies, the tiles were assembled in a mosaic dataset.

The “blend” function made it possible to weight the tiles’ display in relation to the distance from their centre and thus to fit them together more seamlessly.

Now the Mercator globe is georeferenced, users can process the information at will, for example by changing the system of coordinates:

 

 

Find out more
]]> An original picture-taking system http://wp.unil.ch/mercator/en/an-original-picture-taking-system/ Sat, 30 Jan 2016 11:22:38 +0000 http://wp.unil.ch/mercator/dispositif-de-prise-dimages-original/ February 2016 The specialists at the School of Forensic Science have to solve a number of problems inherent in photographing a spherical surface systematically and precisely [...]]]> Photographing a spherical surface systematically and accurately is no easy task. The challenges involved include controlling luminosity (chiefly glare) and standardising image capture (register).

To overcome these hurdles, Eric Sapin and Quentin Milliet, both specialists at the School of Forensic Science (ESC), came up with an original system. The spheres’ rotation was controlled by two cross lasers equipped with levels. The beam of one laser passed through the camera’s viewfinder to centre each image perfectly. The other beam was projected to the side to control the globes’ horizontality and verticality permanently.

The pictures followed the latitudes and longitudes starting from Mercator’s prime meridian, which passes through Fuerteventura in the Canary Islands (14.1° W on today’s globes). The 42 final images, centred on selected intersections in the grid, covered each globe entirely.

Positioning the spheres

Each image was captured in relation to the horizon corresponding to a given meridian thanks to the sighting and horizontality lasers, which made it possible to check that the picture remained stable. The first register aligned the globe’s horizon with the real horizon indicated by the laser beam. The second beam, passing through the camera’s viewfinder, showed the optical axis with the latitude and longitude intersection placed at the centre of the image.

Technical parameters
  • Mamiya Phase One P65+ camera
  • Macro MF 120 mm 1.4 lens
  • Aperture F/22
  • Size of sensor 40.4 x 53.9 mm
  • 8984 x 6732 active pixels
Lighting

To solve the glare problem inherent in photographing a sphere, special lighting was used to yield images with uniform luminosity, colour and sharpness. Four lamps were positioned at 45° angles equidistantly from the centre of the image indicated by the laser beam passing through the camera’s viewfinder.

The lamps were shone on the centre in alternating flashes, first from angles 1 and then from angles 2. The two series of 84 images were then paired, processed to even the luminosity and colours and assembled using Photoshop to create images with no flash glare. Each final image was assembled from a pair by combining the four quadrants with no direct glare around a central point. Next, the images were numbered to facilitate their arrangement and a viewer’s recognition of the regions they covered. Altogether, 423 useful images were obtained for each globe and 84 final high-resolution images would be used for georeferencing.

Depth of field and focusing

For each image, the area photographed was 20 x 20 cm, including a useful central section roughly 16 cm2 to obtain optimal sharpness, uniform luminosity and acceptable distortions due to the curvature of the spheres. The camera was focused one-third behind the first sharp plane and two-thirds in front of the last sharp plane in order to set the field depth on the area captured. The first sharp plane was at the centre of the globes and the last sharp plane on the periphery.

See methods and parameters
Close
Find out more
]]>