RECONSTRUCTION OF THREE-DIMENSIONAL COMPUTER MODELS FROM DIGITIZED STEREO PICTURES
The research reported here was carried out by Ecotech Design Ltd under contract to the Joint Research Centre of the European Communities, Ispra Establishment, as part of the Energy Management in Habitat Programme. One aspect of the Energy Management in Habitat Programme is the Energy Auditing activity which aims to investigate techniques for making energy audits of existing buildings quicker, more reliable and cost effective. An essential part of any energy audit is the collection of the building geometrical data for input to a thermal simulation model. This is likely to be a time consuming process, not helped by the fact that in many buildings original plans will not be available. The JRC therefore decided to investigate the possibility of using images of the building to provide the required dimensional data for the building envelope. The use of stereo pairs of images was proposed as a way of obtaining the three-dimensional geometry of the envelope, which could then be used in a thermal simulation model or for solar shading calculations. From the beginning it was decided to base the thermal simulation on a model that could be run on a microcomputer so that the techniques developed could in principle be used by any small company of building surveyors performing energy audits. Since Ecotech already had a thermal model (SPIEL) and a graphic drawing/editing program (SCRIBE), which would also be essential for this work, running on microcomputers, this company was chosen to develop a stereo-reconstruction algorithm and link it to the existing programs. The stereo reconstruction algorithm was to operate on line drawings which could be generated either from digitized video images, or from photographs using a digitizing tablet. Meanwhile the JRC was to investigate the possibilities of obtaining digitized TV images and reducing them to line drawings by using one of the now emerging vision systems for microcomputers. The research began by investigating the relatively simple case of pairs of stereo images with parallel axes of view. A program was written in C which was successful in converting high resolution pairs of images into models. But the parallel case had a number of limitations when dealing with images of a lower resolution, so algorithms were developed for the much more complicated case of images with converging axes of view. Another program was produced and tested with a variety of types of images.
Bibliographic Reference: COMPUTER AIDED DESIGN 1986, LONDON (UK), SEPT. 2-5, 1986 WRITE TO CEC LUXEMBOURG, DG XIII/A2, POB 1907 MENTIONING PAPER E 32976 ORA
Availability: Can be ordered online
Record Number: 1989125056300 / Last updated on: 1987-11-01
Available languages: en