Below, twilight bird’s-eye view of the scene. At right, highlighted detail. Far right, an exploded view of the vehicle with the victim’s leg resting against parts of the car, wich left marks on the leg that helped recreate the body position.
Moreover, the defendant’s serious visual impairment wouldn’t have allowed him to see the woman from such a distance. In order to prove this theory in a courtroom at the Court of Assizes, we reconstructed the crime scene and also the defendant’s visual capacity.
A hearing was held in February 1999 to prove that the 3D reproduction of the crime scene faithfully represented reality and therefore was valid for the trial.
Our first step was to set up a three-dimensional model of the crime scene. Even though most of the objects making up the scene were by nature extremelysimple to develop, close attention had to be devoted to exact size and location to avoid optical distortions that would have invalidated the reconstruction.
I did a thorough study of the scene and used realworld units in 3D Studio Max r2.5 for modeling, and then checked object size and placement with Max’s Tape helper object and Measure Utility.
The scene was bounded by the edges of the square in front of the cemetery. Once the model was completed, it was crucial to make sure that all materials and surface features corresponded to the real ones.
This was done by taking about 300 photos which were then digitized. One of the most challenging aspects was reproducing the exact colors of the crime scene surfaces.
Great care was given to making the backdrop. I took a series of 18 pictures to cover a 180ø panorama from the accused’s point of view and applied the resulting single combined image to a background half-cylinder. And the same was done for the reverse point of view, so that the two half-cylinders perfectly enclosed the scene at a very wide diameter to avoid distortion. The sets of 18 pictures were taken with a 35mm camera mounted on a tripod with a graduated panoramic head like is used for shooting photos to be stitched into QTVR sequences (the Manfrotto 302 Plus, www.manfrotto.com).
1. The Polizia Scientifica is a police department that investigates with technological methods such as DNA and finger prints. The criminal inquiry is conducted by a magistrate representing the prosecution, helped by the Polizia Scientifica and expert consultants such as mentioned in this article, including myself.
2. Criminal cases in Italy pass through three levels of judgment. The 3D reconstruction told about here was the first to be accepted at all three levels, starting with the Court of Assizes in Novara, going to the Court of Appeal in Turin, and then the Court of Cassation, where the decisions of the first two courts were examined.
Lighting was crucial, since the reconstruction was required specifically to ascertain whether the defendant could have actually seen what he stated. To verify the accuracy of the Max rendering engine, I checked the shadows cast by the sun in situ in three different months, including December–the month in which the murder was committed.
This was done using a luxmeter supplied by the forensics department, and I found that the simulation perfectly matched reality. The two cars were reproduced using construction drawings provided by their manufacturers.
Max Surface Tools was used to model the Lancia Y10, while the Fiat Tipo was modeled using NURBS surfaces and Surface Tools. In order to ensure that the virtual corpse had the same position as the real one, the police doctor worked with us to get a perfect match according to marks on the body. The next stage involved Prof. Fabio Dossi, a renowned optician, who worked with me to reproduce the effect of the defendant’s visual impairment on the screen, based on medical examinations. Images as seen by healthy subjects were modified to represent the visual effect produced in subjects affected by the same condition as the defendant’s.
Each image representing the defendant’s visual field was put out of focus, strictly applying the blurring parameters that the doctor had identified. We thus reproduced what the defendant, given the twilight conditions and his serious sight impairment, could have seen in that specific place and time, the parking lot in front of the Borgomanero Cemetery in Novara, Italy at 5:03pm on 29 December 1997.
To get to this point, much image testing and modification had been done related to specific sight deficiencies. It was a long process, but we were now ready for the hearing. Since all this work had been carried out using my own equipment, five monitors identical to mine were set up in the courtroom to provide satisfactory viewing, and DPS Italia supplied two identical workstations equipped with the PVR (and later dpsReality) video editing card that I had used during development.
At crucial points during the demonstration, Prof. Dossi asked me to display on the monitor what the defendant could see, not in the twilight at the moment of the crime, but in hypothetical full brightness conditions. This showed that not even in such optimum conditions could the defendant have actually seen all he mentioned during his deposition. However, we had not yet proven the validity of the virtual model. The defendant’s impairment needed to be related to his location, which was also determined from his deposition. During model development, we matched the scene shot with a video camera, positioned in the place indicated by the defendant at his eye height, with its twin virtual scene.
Overlapping the two, the images were identical. The court and jury were shown both the actual pictures of the area and their virtual reproductions, so that they could see the perfect correspondence. I also overlapped orthogonal views of the cars’ technical drawings and their 3D reproductions to confirm the modeling accuracy. I then showed a series of overlaps of the actual video made by the forensic department and the reproductions, using a dissolve between them, to show that the 3D scene views matched the real views from several viewpoints.
Finally, I played my best card: a four-minute sequence which provided an overall view of the area to the court, the defense, and the prosecution. By the end of the hearing we had been able to prove and strengthen the prosecution’s case, thanks to a meticulous reconstruction. The sentence was passed down on 23 February 1999. The defendant was found guilty of voluntary manslaughter and was sentenced to 20 years in jail, plus three years probation, and was required to pay 280 million lire (about us$140,000) to the victim’s family. The verdict was subsequently confirmed by the Court of Appeal and the Court of Cassation.
Giuseppe Galliano has a degree in law from the State University of Milan, and has wide experience in working with 3D Studio for DOS and 3DS Max among a wide range of multimedia software and hardware tools, which he uses to provide all manner of production services. You can reach him at or visit www.giuseppegalliano.it (Flash required).