Lynnerup Niels, Andersen Marie, Lauritsen Helle Petri
Institute of Forensic Medicine, University of Copenhagen, Frederik d V's Vej 11, DK-2100 Copenhagen, Denmark.
Leg Med (Tokyo). 2003 Sep;5(3):156-60. doi: 10.1016/s1344-6223(03)00054-3.
We present the results of a preliminary study on the use of 3-D software (Photomodeler) for identification purposes. Perpetrators may be photographed or filmed by surveillance systems. The police may wish to have these images compared to photographs of suspects. The surveillance imagery will often consist of many images of the same person taken from different angles. We wanted to see if it was possible to combine such a suite of images in useful 3-D renderings of facial proportions.Fifteen male adults were photographed from four different angles. Based on these photographs, a 3-D wireframe model was produced by Photomodeler. The wireframe models were then rotated to full lateral and frontal views, and compared to like sets of photographs of the subjects. In blind trials, 9/15 of the wireframe models were assigned to the correct sets of photographs. In five/15 cases, the wireframe models were assigned to several sets, including the correct set. Only in one case was a wireframe model not assigned to a correct set of photographs at all.
我们展示了一项关于使用三维软件（Photomodeler）进行身份识别的初步研究结果。犯罪者可能会被监控系统拍照或录像。警方可能希望将这些图像与嫌疑人的照片进行比对。监控图像通常会包含从不同角度拍摄的同一个人的多张图像。我们想看看是否有可能将这样一组图像组合成有用的面部比例三维渲染图。从四个不同角度对15名成年男性进行了拍照。基于这些照片，Photomodeler生成了一个三维线框模型。然后将线框模型旋转到完全侧视图和正视图，并与受试者的类似照片组进行比较。在盲测中，15个线框模型中有9个被分配到正确的照片组。在15个案例中的5个案例中，线框模型被分配到了几个组，包括正确的组。只有一个案例中的线框模型完全没有被分配到正确的照片组。
