Hybrid Facial Features with Application in Person Identification

This paper presents the hybrid facial feature with identification and verification based on facial images. A query facial image had been taken under different conditions of the facial image of the same person (as the query). The query facial image database was constructed. We have used the technique of three-dimensional (3D) Dlib facial landmarks using a direct linear transform technique. A set of absolute affine invariance had been constructed from a series of the 3D landmark quadruplets, which make the facial identification robust to affine geometric transformation. These 3D facial features serve as a coarse feature depending on each individual facial structure. The construct of the 2D detail features represents the edge facial image confined between the 2D Dlib landmarks. The similarity of the 2D feature is achieved by aligning the 2D query edge image against that of the reference edge image. The geometric transformation matrix is estimated from the 2D Dlib landmarks, where correspondence is well established. An identification/verification cost function using a combination of local 2D facial features and global 3D facial features is utilized to verify and identify a query facial image against a candidate facial image(s). The performance of the algorithm yielding an area of 99.97% perfect classification is represented as a value under the receiver operating characteristic (ROC) curve.

Swimmer’s Head Detection Based on a Contrario and Scaled Composite JTC Approaches

In order to accompany the swimming coaches in evaluating high-level swimmers, we developed a prototype for instantaneous speed estimation. To achieve this, we proposed and validated, in a previous work, a swimmer tracking system based on data fusion. However, the initialization phase is done manually, and our aim, in this paper, is to automate this process. First, we propose a region of interest localization module that allows the detection of the first appearance of the swimmer in the lane as well as the restriction of the region of interest around him. This module is based on the method a contrario which consists of modeling the random noise corresponding to the water and detecting the structured movement relative to the swimmer motion. To do that, we calibrate the pool using DLT (Direct Linear Transform) technique, extract the concerned lane, apply the frame difference approach to detect the moving objects, and then decompose the lane into blocs and classify them into swimmer motion or noise. Second, in order to detect the swimmer’s head, we propose the Scaled Composite JTC which is based on the NL-JTC correlation technique. The input plane of this latter includes a target and a reference image. The first is the region of interest detected by the method a contrario. The second consists of a Scaled Composite Reference. The tests conducted on real video sequences of French swimming championships (Limoges 2015) showed very good results in terms of region of interest localization and swimmer’s head detection which allows a reliable initialization for the tracking system.

A novel method for robust markerless tracking of rodent paws in 3D

Studying animal locomotion improves our understanding of motor control and aids in the treatment of motor impairment. Mice are a premier model of human disease and are the model system of choice for much of basic neuroscience. Placement of the tips of appendages, here paws, is typically critical for locomotion. Tracking paws from a video is difficult, however, due to frequent occlusions and collisions. We propose a method and provide software to track the paws of rodents. We use a superpixel-based method to segment the paws, direct linear transform to perform 3D reconstruction, a 3D Kalman filter (KF) to solve the matching problem and label paws across frames, and spline fits through time to resolve common collisions. The automated method was compared to manual tracking. The method had an average of 2.54 mistakes requiring manual correction per 1000 frames with a maximum of 5.29 possible errors while these values were estimates of the expected errors. We present an algorithm and its implementation to track the paws of running rodents. This algorithm can be applied to different animals as long as the tips of the legs can be differentiated from the background and other parts of the body using color features. The presented algorithm provides a robust tool for future studies in multiple fields, where precise quantification of locomotor behavior from a high-speed video is required. We further present a graphical user interface (GUI) to track, visualize, and edit the tracking data.

Study on Camera Pose Estimation

A common need in photogrammetry, robotics and computer vision is performing camera pose estimation. A comparative analysis is presented here for three classical and representative algorithms, including direct linear transform (DLT), EPNP and Cayley method, each of which computes the translation and rotation matrix using non-iterative method with six or more point correspondences. The comparison shows qualitative and quantitative experimental results to determine (1) the accuracy and robustness under the influence of different levels of noise, (2) the accuracy, robustness and efficiency for different sizes of point correspondences.

Study on the Technologies of Rapid Three-Dimension Optical Measurement for Spatial Tubular Joints Deformation

To solve the problem of deformation measurement for spatial complex tubular joints, a non-contact three-dimension optical method is proposed. Firstly, some artificial targets are pasted on the deformation area before loading. Then the three-dimension coordinates of those targets are calculated by three-dimension reconstruction technique including collinear equation, image orientation based on the coplanar equation, direct linear transform, epipolar geometric constraint and bundle adjustment. Finally, each single deformation stage is aligned to the basic stage, by tracking and comparing the three-dimension coordinates of the homonymous artificial targets, the deformation vectors are obtained. The precision evaluation experiments prove that the proposed measurement system could achieve accuracy of 0.1mm/m. Real scale model experiment shows that the proposed method could fulfill the efficiency and accuracy requirement in three-dimension deformation measurement of the spatial complex tubular joints. In addition, this method has the advantage of non-contact.

An Improved Calibration Method for Optical 3D Measuring System Using Cross Target Based on Bundle Adjustment Algorithm

This paper presents an accurate calibration method of binocular 3D measurement systems for industrial on-site inspection, which uses a cross target with ring coded points. The cross target can be used to calibrate large-scale field-of-view stereo measurement systems and obtain higher measurement precision conveniently. The world coordinates of these ring coded points are not required. The calibration initial value is computed by using the relative orientation method and the Direct Linear Transform (DLT) method of photogrammetry. The bundle adjustment algorithm is used to optimize the calibration parameters as well as the 3D coordinates of the ring coded points. Experiment results show that the RMS error of the reprojection in our method is less than 0.05 pixels and the measurement error is 0.011 mm compared with the Coordinate Measuring Machine (CMM).

Geomatics for Precise 3D Breast Imaging

Canadian women have a one in nine chance of developing breast cancer during their lifetime. Mammography is the most common imaging technology used for breast cancer detection in its earliest stages through screening programs. Clusters of microcalcifications are primary indicators of breast cancer; the shape, size and number may be used to determine whether they are malignant or benign. However, overlapping images of calcifications on a mammogram hinder the classification of the shape and size of each calcification and a misdiagnosis may occur resulting in either an unnecessary biopsy being performed or a necessary biopsy not being performed. The introduction of 3D imaging techniques such as standard photogrammetry may increase the confidence of the radiologist when making his/her diagnosis. In this paper, traditional analytical photogrammetric techniques for the 3D mathematical reconstruction of microcalcifications are presented. The techniques are applied to a specially designed and constructed x-ray transparent Plexiglas phantom (control object). The phantom was embedded with 1.0 mm x-ray opaque lead pellets configured to represent overlapping microcalcifications. Control points on the phantom were determined by standard survey methods and hand measurements. X-ray films were obtained using a LORAD M-III mammography machine. The photogrammetric techniques of relative and absolute orientation were applied to the 2D mammographic films to analytically generate a 3D depth map with an overall accuracy of 0.6 mm. A Bundle Adjustment and the Direct Linear Transform were used to confirm the results.