期刊名称：ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
印刷版ISSN：2194-9042
电子版ISSN：2194-9050
出版年度：2008
卷号：XXXVII Part B5
页码：805-812
出版社：Copernicus Publications
摘要：The paper describes the first Malaysian Craniofacial soft tissue 3D imaging system which was developed based on the integration of stereophotogrammetry and triangulation-based laser scanning system. The main purposes of developing the imaging system are to provide a non-contact method for craniofacial anthropometric measurement and fast and radiation free 3D modelling of craniofacial soft tissue. The stereophotogrammetric system consists of high resolution digital cameras setup as three stereo cameras placed at the left, front and right sides of the patient. The system was also add-up with another extra two digital cameras setup in convergent mode at bottom left and bottom right of the patient. The combination of all the cameras allowed for the accuracy improvement of craniofacial anthropometry through a novel technique called "natural features technique". In the natural features technique, the images acquired from the camera system were used to digitize the natural features on the human face. Photogrammetric triangulation method was used to calculate the 3D coordinates of the features. The cameras was highly synchronized (0.2miliseconds) using a new external shutter controller. The stereophotogrammetric system was designed to be operated in battery system for mobile data capturing purposes. Apart from the camera system, the developed stereophotogrammetric system was completely designed with the object space control frame. The new patient's chair and photogrammetric control frame has been designed and developed. The object distance is 700mm. Special-built camera calibration device was designed and developed to calibrate each camera individually. The camera was placed at the camera platform to capture eight convergent images of the 3D test field. The self calibration bundle adjustment process was carried out using Australis software to calculate the calibration parameters. The developed stereophotogrammetric system was integrated with the triangulation-based laser scanning system. Two eye-safe Minolta VI-910 laser scanners was setup at right and left side of the patient and near to the stereophotogrammetric system with object distance of 1000mm. For the purposes of scanning the craniofacial morphology, the scanners was setup with middle lens (focal length = 14mm) and fine mode resolution with one scan mode. The scanners scanned one after another with 19 seconds scan period (complete scan). With the optimum setup, two scan images was acquired which covered the craniofacial area (from right ear to left ear and the hair line to bottom part of the chin). The texture data of the craniofacial area was also captured. Both scanners were calibrated using calibrated object. In the data collection session, the patient sited on the chair with the head placed at the middle of the control frame. The complete system was firstly tested using mannequin to determine the accuracy and precision. Both stereo images and scans data was processes separately. The DVP digital stereophotogrammetric workstation was used to carry out the photogrammetric orientation of the stereo images. The vectorization module was used to measure the 3D coordinates of the craniofacial landmarks. The laser scan datasets involved with few data processing steps which included the registration process, merging process, editing process, smoothing process and texturing process. The processing tasks were carried out using the RapidForm 2004 software. At final stage, the craniofacial landmarks measured from stereophotogrammetric system were registered onto the 3D model developed from the laser scanners. The research also involved with the development of the craniofacial database system which used to store the captured datasets. The results show that both stereophotogrammetry and laser scanning system was an effective system to be used in craniofacial mapping. Both systems provide high accuracy non-contact measurement method. The accuracy of the craniofacial landmark measurement is 0.2mm with one standard deviation, while the accuracy of the 3D model is 0.3mm with one standard deviation
