J. Paul Siebert

Human body 3D imaging by speckle texture projection photogrammetry

Describes a non‐contact optical sensing technology called C3D that is based on speckle texture projection photogrammetry. C3D has been applied to capturing all‐round 3D models of the human body of high dimensional accuracy and photorealistic appearance. The essential strengths and limitation of the C3D approach are presented and the basic principles of this stereo‐imaging approach are outlined, from image capture and basic 3D model construction to multi‐view capture and all‐round 3D model integration. A number of law enforcement, medical and commercial applications are described briefly including prisoner 3D face models, maxillofacial and orofacial cleft assessment, breast imaging and foot scanning. Ongoing research in real‐time capture and processing, and model construction from naturally illuminated image sources is also outlined.

Local feature extraction and matching on range images: 2.5D SIFT

This paper presents an algorithm that extracts robust feature descriptors from 2.5D range images, in order to provide accurate point-based correspondences between compared range surfaces. The algorithm is inspired by the two-dimensional (2D) Scale Invariant Feature Transform (SIFT) in which descriptors comprising the local distribution function of the image gradient orientations, are extracted at each sampling keypoint location over a local measurement aperture. We adapt this concept into the 2.5D domain by concatenating the histogram of the range surface topology types, derived using the bounded [-1,1] shape index, and the histogram of the range gradient orientations to form a feature descriptor. These histograms are sampled within a measurement window centred over each mathematically derived keypoint location. Furthermore, the local slant and tilt at each keypoint location are estimated by extracting range surface normals, allowing the three-dimensional (3D) pose of each keypoint to be recovered and used to adapt the descriptor sampling window to provide a more reliable match under out-of-plane viewpoint rotation.

Three-dimensional facial characteristics of Caucasian infants without cleft and correlation with body measurements.

Objective The aim of this study was to characterize the soft tissue facial features of infants without cleft and to report on the correlation between these with weight, length, and head circumference. Design This was a prospective study using a noninvasive three-dimensional (3D) stereophotogrammetry (C3D) system to capture the images of the participants. Landmarks were identified on the 3D facial images. Means and SDs were derived for facial distances and angles. A facial asymmetry score was calculated for each image. Two sample Students t tests, Pearsons correlation coefficients and analysis of covariance were used to ascertain any gender differences and determine whether these could be explained by weight differences. Participants Eighty-three infants, 41 boys and 42 girls, were captured at rest with their lips apart, at approximately 3 months of age. Results Significant sex differences, of 1 to 2 mm, were found in several facial dimensions, such as face height and nose width. The larger facial measurements correlated significantly with body measurements. Analysis of variance confirmed these differences could be explained by differences in weight. There were no sex differences in the nose/mouth width ratios or in any of the angles measured, suggesting that there may be little sex difference in shape. A slight degree of asymmetry in the faces of infants without cleft was detected. Conclusions Comparisons between noncleft controls and infants with cleft should take cognizance of normal age and sex variations in height and weight that occur among infants.

Investigation into accuracy and reproducibility of a 3D breast imaging system using multiple stereo cameras

BACKGROUND The aim of this study was to evaluate the validity of a three-dimensional (3D) multiple stereo camera system for objective breast assessment. METHODS A multiple stereo camera system, which consisted of four pods and eight cameras, two cameras on each pod, developed by Glasgow University, was used. Nine specially shaped plaster breast models were captured once, 3Dmodels were constructed and the volume of each plaster model was measured 10 times by the breast analysis tool (BAT) software. A comparison was conducted with water displacement method, and measurements were repeated 10 times. The breast of six live volunteers was captured six times; from each breast capture, a 3D model was constructed and the volume was measured with BAT software. Breast volume assessment by the water displacement method was repeated six times. RESULTS In all plaster casts, the discrepancies in volume measurements between 3D imaging and water displacement methods did not exceed 40 cc. The overall mean relative difference was 5%. The differences of the two methods were not significant at p = 0.189, overall mean difference: 11.1 cc and 95% confidence interval (CI) was (-6.732, 28.976). In the live models, the differences in breast volume measurements between the 3D imaging and water displacement methods were significant at p ≤ 0.017, overall mean difference: 207.05 cc and 95% CI (56.12, 357.98). Measurements by 3D imaging were consistently smaller. In the live models, 3D imaging overall was a more reproducible method for measuring breast volume than the water displacement method with a standard deviation of 36 units cc(-1) and 62.6 units cc(-1), respectively. CONCLUSIONS The 3D breast imaging system using multiple stereo cameras was accurate for measuring the volumes of breast-shaped plaster models, and it was more reproducible than the water displacement method in live models. 3D imaging is a reliable method for the comparative assessment of breast volume.

3D Thermography Imaging Standardization Technique for Inflammation Diagnosis

We develop a 3D thermography imaging standardization technique to allow quantitative data analysis. Medical Digital Infrared Thermal Imaging is very sensitive and reliable mean of graphically mapping and display skin surface temperature. It allows doctors to visualise in colour and quantify temperature changes in skin surface. The spectrum of colours indicates both hot and cold responses which may co-exist if the pain associate with an inflammatory focus excites an increase in sympathetic activity. However, due to thermograph provides only qualitative diagnosis information, it has not gained acceptance in the medical and veterinary communities as a necessary or effective tool in inflammation and tumor detection. Here, our technique is based on the combination of visual 3D imaging technique and thermal imaging technique, which maps the 2D thermography images on to 3D anatomical model. Then we rectify the 3D thermogram into a view independent thermogram and conform it a standard shape template. The combination of these imaging facilities allows the generation of combined 3D and thermal data from which thermal signatures can be quantified.

Constructing dense correspondences for the analysis of 3D facial morphology

In this paper, we present a method for constructing dense correspondences between 3D open surfaces that is sufficiently accurate to permit clinical analysis of 3D facial morphology. Constructing dense correspondences between 3D models representing facial surface anatomy is a natural extension of landmark-based methods for analysing facial shape or shape changes. Compared to landmark-based methods, dense correspondences sample the entire surface and hence provide a more thorough description of the underlying 3D structures. The method we present here is based on elastic deformation, which deforms a 3D generic model onto the 3D surface of a specific individual. We are then able to construct dense correspondences between different individuals by analysing their corresponding deformed generic models. Validation experiments show that, using only five manually placed landmarks, approximately 95% of triangles on the deformed generic mesh model are within the range of +/-0.5mm to the corresponding original model. The established dense correspondences have been exploited within a principal components analysis (PCA)-based procedure for comparing the facial morphology of a control group to that of a surgically managed group comprising the patients who have been subject to facial lip repair.

Three-dimensional assessment of facial appearance following surgical repair of unilateral cleft lip and palate.

Background and Objective Objective assessment of postsurgical facial asymmetry can be difficult, but three-dimensional (3D) imaging techniques have made this possible. The objective of this study was to assess residual asymmetry in surgically repaired unilateral cleft lip (UCL) and unilateral cleft lip and palate (UCLP) patients and to compare this with noncleft controls. Design Retrospective multicohort comparative study. Patients and Methods Fifty-one 10-year-old children with surgically managed UCLP and 44 children with UCL were compared with a control group of 68 ten-year-olds. The 3D facial models at rest and with maximum smile were created using a 3D imaging system. Asymmetry scores were produced using both anatomical landmarks and a novel method of facial curve analysis. Results Asymmetry for the whole face was significantly higher in both cleft groups compared with controls (P < .001). UCLP asymmetry was higher than UCL (P < .001). In cleft patients, the upper lip and nasal rim were the most asymmetric (P < .001 to .05). Control subjects also displayed a degree of facial asymmetry. Maximum smile did not significantly affect the symmetry of the whole face, but it increased asymmetry of the vermillion border and nasal rim in all three groups (P < .001). Conclusions Despite surgical intervention at an early age, asymmetry remains significant in cleft patients at 10 years of age. Three-dimensional imaging is a noninvasive objective assessment tool that identifies specific areas of the face responsible for asymmetry. Facial curve analysis describes the face more comprehensively and characterizes soft tissue contours.

A head called Richard

This paper describes two preliminary experiments concerned with the construction of a robot head. The initial design and research is aimed at producing a system with two cameras and two microphones on a system capable of operating with the same degrees of freedom and reflex times as its biological counterpart. Whilst the primary goal of the project is develop an anthropomorphic system with the sensory reflex capabilities of a human head, the system will also contain some non-anthropomorphic components. The most obvious of the non-anthropomorphic components is a spatially and temporally programmable light source. Some preliminary results are presented. use this information to direct autonomous behaviour (Brooks, 1989) but these reflexes have not been coupled with vision systems for integrated tracking or gaze control.

Unsupervised clustering in Hough space for recognition of multiple instances of the same object in a cluttered scene

We describe an active binocular vision system that is capable of localising multiple instances of objects of the same-class in different settings within a covert, pre-attentive, visual search strategy. By clustering SIFT-feature matches that have been projected into a non-quantised (i.e. continuous) Hough space we are able to detect up to 6 same-class object instances simultaneously while tolerating up to ~66% of each objects surface being occluded by another object instance of the same-class. Our findings are based on using a database of ~2300 images of synthetically composited and real-world images.

Recognising the clothing categories from free-configuration using Gaussian-Process-based interactive perception

In this paper, we propose a Gaussian Process-based interactive perception approach for recognising highly-wrinkled clothes. We have integrated this recognition method within a clothes sorting pipeline for the pre-washing stage of an autonomous laundering process. Our approach differs from reported clothing manipulation approaches by allowing the robot to update its perception confidence via numerous interactions with the garments. The classifiers predominantly reported in clothing perception (e.g. SVM, Random Forest) studies do not provide true classification probabilities, due to their inherent structure. In contrast, probabilistic classifiers (of which the Gaussian Process is a popular example) are able to provide predictive probabilities. In our approach, we employ a multi-class Gaussian Process classification using the Laplace approximation for posterior inference and optimising hyper-parameters via marginal likelihood maximisation. Our experimental results show that our approach is able to recognise unknown garments from highly-occluded and wrinkled configurations and demonstrates a substantial improvement over non-interactive perception approaches.