3D-Harris Feature-Based CPD Registration for Freeform Surface System Alignment Errors

The typical probability based point pattern matching method is coherent point drift (CPD) algorithm, which treats one point set as centroids of a Gaussian mixture model, and then fits it to the other. It uses the expectation maximization framework, where the point correspondences and transformation parameters are updated alternately. However, the anti-outlier performance of CPD is not robust enough as outliers have always been involved in the operation until the CPD converges. Hence, an automatic outlier suppression (AOS) mechanism is proposed. First, outliers are judged by a matching probability matrix. Then, transformation parameters are fitted using accurate matching point sets. Finally, the Gaussian centroids are forced to move coherently by this transformation model. AOS-CPD can efficiently improve the anti-outlier performance of rigid CPD. Furthermore, CPD is applied to image matching. A new local changing information descriptor-relative phase histogram (RPH) is designed and RPH-AOS-CPD is proposed to embed RPH measurement into AOS-CPD as a constraint condition. RPH-AOS-CPD makes full use of grayscale information besides having an excellent anti-outlier performance. The experimental results based on both synthetic and real data indicate that compared with other algorithms, AOS-CPD is more robust to outliers and RPH-AOS-CPD offers a good practicability and accuracy in image matching applications.

Point set registration is a key problem in many computer vision tasks. The goal of point set registration is to match two sets of points and estimate the transformation parameter that maps one point set to the other. Among the many published registration methods, the recently proposed Coherent Point Drift (CPD) algorithm stands out for its accuracy. In this paper we show that by casting CPD in the Bayesian framework we can obtain even better results. In particular, in case of large translation amount, our proposed mathod has much less number of iterations than CPD without any loss of accuracy. Experimental results confirms the advantages of the proposed method and shows an overall speedup when compared with the CPD method.

Ship surveillance using multiple remote sensing sensors becomes more and more vital presently. Among the various sensors, space-borne Synthetic Aperture Radar (SAR) is optimal for its high resolution over wide swaths and all-weather working capabilities. Meanwhile, Automatic Identification System (AIS) is efficient to provide ship navigational information. Limited to the progress of ship surveillance using SAR image only, the integration of them significantly benefits more. Data association is the fundamental issue. Many algorithms have been developed including the Nearest-Neighbour (NN) algorithm, the Joint Probabilistic Data Association (JPDA) method, and the Multiple Hypothesis Testing (MHT) approach. Ship positions derived from SAR image can be associated with the ones provided by AIS. State-of-the-art method (NN algorithm) is proved to be feasible. But it faces more challenges under adverse circumstances, such as high-density-shipping condition. We investigate the non-rigid Point Pattern Matching (PPM) method to solve this problem. To the best of our knowledge, this paper is the first to introduce non-rigid PPM to the data association of SAR and AIS. On the basis of introduction to the data association, Coherent Point Drift (CPD) algorithm is investigated. Experiments are carried out and the results illustrate that the CPD algorithm achieves higher accuracy and outperforms state-of-the-art method, especially under high-density-shipping condition.

We present a feature point detection algorithm which we use for non-rigid registration, illustrated for breast images (mammography, MRI). By associating the continuous intrinsic dimensionality of image structure with the output of a scale saliency algorithm, breast boundary points can be separated from internal feature points. Correspondences established for the breast boundary and internal feature points respectively are used to drive two recent non-rigid registration techniques: polyaffine transformation and coherent point drift registration. Experimental results are presented for digital breast tomosynthesis and 3D breast MRI, and in all case achieve good spatial alignments.

Ultra-precision freeform surfaces are complex surfaces that possess non-rotational symmetry and are widely used in advanced optics applications. Due to the geometrical complexity of optical freeform surfaces, there is, as yet, a lack of generalized surface characterization methods which measure various types of ultra-precision freeform surfaces with sub-micrometer form accuracy and surface finish in the nanometer range. To make good this deficiency, a generalized approach for the measurement and characterization of ultra-precision freeform surfaces, named the intrinsic surface feature-based method (ISFM), is presented in this paper. The ISFM makes use of intrinsic surface properties (e.g. curvatures, normal vectors, torsion and intrinsic frames) to conduct data matching or uses some algorithms to search for correspondences such as correlation functions. The method is experimentally verified through a series of measurement experiments. The results show that the proposed ISFM is capable of addressing the deficiencies and limitations of traditional freeform surface characterization methods which are susceptible to outliers and to uncertainty due to the geometry of the freeform surfaces. ISFM is a generalized methodology which is not dependent on the type of freeform surface being characterized.

This letter presents an effective approach for synthetic aperture radar (SAR) image registration. Corresponding edge features are extracted by exploring the local spatial relationship around corresponding speeded-up robust feature points. Only edge pairs that are matched in the aspect of distance and orientation to the center feature points are taken as effective edge features. For edge-point set matching, we introduce the coherent point drift (CPD) registration method. CPD is able to maintain the global topological structure of edges and is robust to outliers. We refine the original CPD by taking our confidence on edge points as prior information. A new objective function is deduced, resulting in a better registration performance. Experimental results show that the proposed method performs well in terms of registration robustness and accuracy.

Assessment of the Implant Geometry in Fractionated Interstitial HDR Breast Brachytherapy

The advantages of the freeform surface is obviously in optical system application: higher image quality, larger view field, simplified system and more design degrees of freedom. But, the large application of the freeform surface is limited by the difficulty of its manufacture. Measurement is the base of optical fabrication. In order to achieve the high precision freeform surface, the surface should be measured accurately. High precision freeform surface measurement technology is one of the most important problem in the field of optical fabrication. A new measurement method has been proposed which is different from the existing free-form surface measurement instruments, and it is proved in theory that can achieve high precision measurement for free-form surfaces. In this paper, a complete mathematical model is established for the probe part through ray tracing, so as to obtain the object-image relationship data of light passing through each optical element. Using the point set, the wave image of each surface is obtained, the optical path difference of the light is calculated, and the corresponding interference diagram is drawn. This process is principle verification. Contrary to the simulation process, in the process of measuring, the interferogram is obtained first, and then the deviation value is calculated.

While for non-reflecting surfaces a variety of optical techniques is available that allow the flexible geometric measurement of free-form surfaces, established approaches for the testing of specular surfaces are limited to basic geometries or slight deviations from an assumed reference geometry. As not only the intensified use of aspheric optics but also the increasing quality standards for technical surfaces call for an enhanced measurement range, the authors have developed two related techniques for the direct three-dimensional measurement of specular reflecting surfaces. These techniques are based on the observation of the mirror image of a grid-like reference structure and apply principles that are well-known from measurement systems for non-reflecting surfaces, such as photogrammetry and structured illumination, to the evaluation of specular surfaces. While the first of these approaches works with an active triangulation process based on one camera and a pseudo three-dimensional reference structure, the second one utilises a stereo-photogrammetric camera system in conjunction with a merely two-dimensional reference structure. Both systems allow the unambiguous measurement of reflecting free-form surfaces and may, by the use of multiple wavelength and photogrammetric stitching techniques, be extended to the measurement of rather complex geometries. Besides the fundamental mode of operation of this so-called reflection grating photogrammetry, the properties of a suitable reference structure will be presented in this contribution. Furthermore the photogrammetric calibration procedure and the used calibration models will be discussed. Finally the measurement uncertainty is evaluated based on both, experimental and theoretical considerations.

High-quality aspherical and freeform surfaces are in high demand, and the high-accuracy form measurement of such surfaces is a challenging task. To explore the current status of form measurement systems for complex surfaces such as aspheres and freeforms, interlaboratory comparison measurements are performed. This study presents the pseudonymized results obtained using three different surfaces (metal asphere, glass asphere, toroidal surface) in a total of six different round robins. These results were taken from a total of 13 different measurement instruments based on 9 different measurement principles and operated at 12 different laboratories. They were analyzed using a sophisticated procedure that was first developed in 2018 and then refined and tested on simulated data in 2022 to address the challenges of such a comparison at this level of accuracy. In the current study, we applied these refined methods to data acquired from tactile and optical point measurements as well as from optical areal measurements. As there are no absolutely measured and very well characterized reference standard aspherical and freeform surfaces available at the accuracy level of a few tens of nanometers root-mean-square, the approximated true forms of the surfaces were derived from the measurements and indicate the manufacturing accuracy of the surface forms. Then, the measurement’s differences to the approximated true forms were analyzed, which directly indicate the systematic measurement errors of the instruments. By also comparing the approximated true forms from the two different round robins for each surface, additional insights into the reliability and stability of these so-called virtual reference topographies were gained.

Mobile laser scanning (MLS) systems are often used to efficiently acquire reference data covering a large-scale scene. The terrestrial laser scanner (TLS) can easily collect high point density data of local scene. Localization of static TLS scans in mobile mapping point clouds can afford detailed geographic information for many specific tasks especially in autonomous driving and robotics. However, large-scale MLS reference data often have a huge amount of data and many similar scene data; significant differences may exist between MLS and TLS data. To overcome these challenges, this paper presents a novel deep neural network-based localization method in urban environment, divided by place recognition and pose refinement. Firstly, simple, reliable primitives, cylinder-like features were extracted to describe the global features of a local urban scene. Then, a probabilistic framework is applied to estimate a similarity between TLS and MLS data, under a stable decision-making strategy. Based on the results of a place recognition, we design a patch-based convolution neural network (CNN) (point-based CNN is used as kernel) for pose refinement. The input data unit is the batch consisting of several patches. One patch goes through three main blocks: feature extraction block (FEB), the patch correspondence search block and the pose estimation block. Finally, a global refinement was proposed to tune the predicted transformation parameters to realize localization. The research aim is to find the most similar scene of MLS reference data compared with the local TLS scan, and accurately estimate the transformation matrix between them. To evaluate the performance, comprehensive experiments were carried out. The experiments demonstrate that the proposed method has good performance in terms of efficiency, i.e., the runtime of processing a million points is 5 s, robustness, i.e., the success rate of place recognition is 100% in the experiments, accuracy, i.e., the mean rotation and translation error is (0.24 deg, 0.88 m) and (0.03 deg, 0.06 m) on TU Delft campus and Shanghai urban datasets, respectively, and outperformed some commonly used methods (e.g., iterative closest point (ICP), coherent point drift (CPD), random sample consensus (RANSAC)-based method).

One of the essential problems in the measurement of the freeform surfaces on a coordinate measuring machine is to design appropriate sampling plans to improve the industrial practice in terms of the tradeoff between the sampling accuracy and the efficiency. This paper presents a curve network sampling strategy to approximate the measured surface within a required accuracy while minimizing the cost and time for the measurement by adaptively deriving the optimal sampling locations. The method iteratively extracts two sets of iso-planar curves along two different directions on the parts to form a curve network, which is used to reconstruct the measured surfaces based on the Gordon surface fitting method. Two criteria are integrated to determine the locations of the sampled curves in the sampling process, including the surface complexity and the deviation of the reconstructed surfaces from the CAD model. Both the computer simulation and the actual measurement are conducted to verify the superior sampling efficiency of the proposed method to the conventional raster fashion sampling in measuring freeform surfaces.

A calibration method for stereo vision sensor with large FOV based on 1D targets

Compared with conventional high-end optical systems, application of freeform optics offers many advantages. Their widespread use, however, is held back by the lack of a suitable measurement method.The NANOMEFOS project aims at realizing a universal freeform measurement machine to fill that void.The principle of operation of this machine requires a novel sensor for surface distance measurement, the development and realization of which is the objective of the work presented in this thesis. The sensor must enable non-contact, absolute distance measurement of surfaces with reflectivities from 3.5% to 99% over 5 mm range, with 1 nm resolution and a 2s measurement uncertainty of 10 nm for surfaces perpendicular to the measurement direction and 35 nm for surfaces with tilts up to 5°. To meet these requirements, a dual-stage design is proposed: a primary measurement system tracks the surface under test by translating its object lens, while the secondary measurement system measures the displacement of this object lens. After an assessment of various measurement principles through comparison of characteristics inherent to their principle of operation and the possibilities for adaptation, the differential confocal measurement has been selected as the primary measurement method. Interferometry is used as secondary measurement method. To allow for correction of tilt dependent error through calibration, a third measurement system has been added, which measures through which part of the aperture the light returns. An analytical model of the differential confocal measurement principle has been derived to enable optimization. To gain experience with differential confocal measurement, a demonstrator has been built, which has resulted in insights and design rules for prototype development. The models show satisfactory agreement with the experimental results generated using the demonstrator, thus building confidence that the models can be applied as design and optimization tools. Various properties that characterize the performance of a differential confocal measurement system have been identified. Their dependence on the design parameters has been studied through simulations based on the models. The results of this study are applied to optimize the sensor for use in NANOMEFOS. An optical system has been designed in which the interferometer and the differential confocal systems are integrated in a compact design. The optical path of the differential confocal system has been folded using prisms and mirrors so that it can be realized within the allotted volume envelope. For the same reason, many components are adapted from commercially available parts or are custom made. An optomechanical and mechatronic design has been made around the optical system. A custom focusing unit has been designed that comprises a guidance mechanism and actuator to enable tracking of the surface. To achieve a low measurement uncertainty, it aims at accurate motion, high bandwidth and low dissipation. The lateral position of the guidance reproduces within 20 nm and from the frequency response, it is expected that a control bandwidth of at least 800 Hz can be realized. Power dissipation depends on the form of the freeform surface and is a few mW for most expected trajectories. Partly custom electronics are used for signal processing, and to drive the laser and the focusing unit. Control strategies for interferometer nulling, focus locking and surface tracking have been developed, implemented and tested. Various tests have been performed on the system to evaluate the performance. Calibrations must be carried out to achieve the required measurement uncertainty. One calibration is based on a new method to measure tilt dependency of distance sensors. The sensor realized has 5 mm measurement range, -2.5 µm to 1.5 µm tracking range, sub-nanometer resolution, and a small-signal bandwidth of 150 kHz. Using the test results, the 2s measurement uncertainty after calibration is estimated to be 4.2 nm for measurement of rotationally symmetric surfaces, 21 nm for measurement of medium freeform surfaces and 34 nm for measurement of heavily freeform surfaces. To test the performance of the machine with the sensor integrated, measurements of a tilted flat have been carried out. In these measurements, a tilted flat serves as a reference freeform with known surface form. The measurements demonstrate the reduction of tilt dependent error using the new calibration method. A tilt robust, single point distance sensor with millimeter range and nanometer uncertainty has been developed, realized and tested. It is installed in the freeform measurement machine for which it has been developed and is currently used for the measurement of optical surfaces.

A high accuracy and flexible method for calibrating intrinsic parameters of CCD camera is presented. The intrinsic parameters of camera is determined from collinear constraint and computed through bundle adjustment algorithm. Calibration is carried out by moving an optical feature point with CMM and forming a virtual stereo calibration board. The images of this virtual board are taken by the calibrated camera which in different locations and orientations. Due to take infrared LED as feature point and the light intensity of these feature points can be automatically controlled according to the distance between camera and feature point, the imaging feature points have uniform intensity profile and high contrast with background, and hence the calibration accuracy is improved. This method has been used for calibrating the Kodak MegaPlus 1.4i CCD cameras, which used in large-scale stereo vision coordinate online measurement system, and obtained higher measurement precision.