The reporter learned on the 30th from the Shenzhen Institute of Advanced Technology, Chinese Academy of Sciences that the team of researcher Song Zhan at the Machine Vision Research Center of the Institute of Integrated Technology has made progress in the field of dynamic three-dimensional facial imaging technology. The research team has independently developed a multi-view 4D high-precision three-dimensional facial imaging system. Compared with traditional 3D facial scanning technology, this system has achieved substantial improvements in accuracy, resolution, and speed.

Song Zhan (right) and team members debugging the multi-view 4D facial scanning system. Photo by Lin Yicheng

According to the introduction, the multi-view 4D facial scanning system developed by the research team consists of three structured-light cameras at different near-infrared wavelength bands. The underlying algorithm adopts the high-frequency fringe displacement encoded 3D reconstruction method proposed by Song Zhan's team, which can achieve a 3D scanning speed of more than 100 frames per second at 1080P (1920*1080 pixels) resolution. The GPU-based highly parallel 3D reconstruction algorithm can achieve a real-time reconstruction speed of up to 300 Hz, with a depth imaging error of less than 0.05 mm.

In order to achieve more complete dynamic 3D facial imaging from multiple angles, the system uses near-infrared light at three different wavelength bands (invisible to the human eye) as the light source, which not only prevents glare to the face but also avoids mutual interference among the projection patterns cast by the projectors of the three sets of equipment, thereby improving imaging completeness.

Song Zhan said that the research team adopted an innovative image enhancement algorithm and, combined with breakthroughs in highly robust fringe encoding and decoding algorithms, improved the phase calculation accuracy for decoding the projected patterns, thereby improving the 3D reconstruction accuracy.

According to the introduction, the research team has currently applied this technology to film special effects, special processing, facial 3D diagnosis and treatment, dynamic 3D vision-guided assembly, and other areas, and has achieved good results.