ICP Algorithm Research Articles

Stable 3D Airways Registration from Segmented 3D CT Images

Image registration is an efficient computational technique for diagnosis and diseases monitoring. Lungs airways diseases, such as chronic obstructive pulmonary disease (COPD) and asthma, require periodical checkups and past data comparison to estimate the patient condition. In this work, it is proposed an airways registration algorithm that is performed branch by branch. The branch registration combines the positive aspects from the conventional ICP and skeleton registration algorithms. The registration procedure requires two main modules: point correspondence determination, and transformation determination. The point correspondence determination is the most time consuming and it is performed in high speed using a 3D discrete Voronoi diagram. Copyright ©2015 IF AC.

应用显著纹理特征的医学图像配准

Traditional registration methods based on geometric measurement can not match the medical image with local deformation. to solve the problem, an improved iterative closest points (icp) algorithm about human visual cognitive process is proposed base on the salient texture. firstly, the method establishes the model for the salient texture feature of a medical image based on active appearance model (aam) algorithm, and it gives the feature point with more salient for a larger weight to complete the image match in the first step. then, it introduces the salient texture distance to the traditional space distance. by simulating the human visual cognitive process proposed by gestalt, the linear decreasing weight is used to balance the two kinds of distance measuring methods. with the algorithm, a whole registration is obtained by the geometric distance in the early stage. on the other hand, the feature points of local deformation are accurately registrated with the texture features in the later stage. at last, a series of experiments are performed on real live images. the experiment results show that the algorithm can get a good matching result, and the registration accuracy is 78.82%, increasing by 22.22% as compared with those of other popular algorithms. the experimental results also show that it is not sensitive to the rotation of the images. it concludes that the algorithm solves the registration problems of local deformation in medical organs and achieves higher precision and stronger robustness. ©, 2015, chinese academy of sciences. all right reserved.

Face Detection, Registration and Feature Localization Experiments with RGB-D Face Database

RGB-D Face database is a multimodal database based on Kinect. This paper focuses on pre-processing techniques for depth data that involve face detection, registration and facial feature localization which are very critical for face authentication and identification systems. Faces are detected in the range data, assisted by detection of face on corresponding texture data using Viola-Jones detector. Validated faces using a depth PCA classifier are registered using conventional ICP algorithm. Finally Nose tip and eye corners are localized on registered faces.

Perimeter-Based Polar Scan Matching (PB-PSM) for 2D Laser Odometry

The paper presents Perimeter-Based Polar Scan Matching (PB-PSM), a new 2D scan matching algorithm. The algorithm favors matches with a larger perimeter overlap between the two input scans, while using a robust cost minimization process (using an adaptive direct search method, made possible due to a linear complexity data association technique). PB-PSM is benchmarked against the previously published PSM and PSM-C algorithms, and numerous realizations of the ICP algorithm. Results for convergence, accuracy, and computational speed are discussed. PB-PSM is employed on several laser scan datasets, both existing and in-house. Quantitative comparison of resulting maps is done using a new metric for evaluating occupancy grid maps accuracy, by calculating the average cell distance from the walls of the true map. The relative importance of each novel contribution is quantified using the new metric. Additional qualitative analysis is provided for previously published and relatively large datasets.

Water Hazard Detection Based on 3D LIDAR

Autonomous off-road navigation is a highly complicated task for a robot or unmanned ground vehicle (UGV) owing to the different kinds of obstacles it could encounter. In-particular, water hazards such as puddles and ponds are very common in outdoor environments and are hard to detect even with ranging devices due to the specular nature of reflection at the air water interface. In recent years, many researches to detect the water bodies have been done. But there still has been very little work on detecting bodies of water that could be navigation hazards, especially at night. In this paper, we used Velodyne HDL-64ES2 3D LIDAR to detect water hazard. The approach first analyzes the data format and transformation of 3D LIDAR, and then writes the data acquisition and visualizations algorithm, integrated data based on ICP algorithm. Finally according the intensity distribution identifies the water hazard. Experiments are carried out on the experimental car in campus, and results show the promising performance.

3D scanning of cultural heritage with consumer depth cameras

Three dimensional reconstruction of cultural heritage objects is an expensive and time-consuming process. Recent consumer real-time depth acquisition devices, like Microsoft Kinect, allow very fast and simple acquisition of 3D views. However 3D scanning with such devices is a challenging task due to the limited accuracy and reliability of the acquired data. This paper introduces a 3D reconstruction pipeline suited to use consumer depth cameras as hand-held scanners for cultural heritage objects. Several new contributions have been made to achieve this result. They include an ad-hoc filtering scheme that exploits the model of the error on the acquired data and a novel algorithm for the extraction of salient points exploiting both depth and color data. Then the salient points are used within a modified version of the ICP algorithm that exploits both geometry and color distances to precisely align the views even when geometry information is not sufficient to constrain the registration. The proposed method, although applicable to generic scenes, has been tuned to the acquisition of sculptures and in this connection its performance is rather interesting as the experimental results indicate.

Skyline matching based camera orientation from images and mobile mapping point clouds

Abstract. Mobile Mapping is widely used for collecting large amounts of geo-referenced data. An important role plays sensor fusion, in order to evaluate multiple sensors such as laser scanner and cameras jointly. This requires to determine the relative orientation between sensors. Based on data of a RIEGL VMX-250 mobile mapping system equipped with two laser scanners, four optional cameras, and a highly precise GNSS/IMU system, we propose an approach to improve camera orientations. A manually determined orientation is used as an initial approximation for matching a large number of points in optical images and the corresponding projected scan images. The search space of the point correspondences is reduced to skylines found in both the optical as well as the scan image. The skyline determination is based on alpha shapes, the actual matching is done via an adapted ICP algorithm. The approximate values of the relative orientation are used as starting values for an iterative resection process. Outliers are removed at several stages of the process. Our approach is fully automatic and improves the camera orientation significantly.

Register Aircraft Point-Clouds Based on Point Features

The 3D models obtained from laser radar are often used to analyze the objects’ reconstruction. Due to the target’s structure and the viewing angle, multi-angle scan are demanded to cover target’s whole surface, which means data got from different directions need to be spliced together. The ICP algorithm can register accurately. But if the surface of target is complex, such as aircrafts, the direct use of ICP algorithm is difficult to complete. In this paper, the corresponding points are found by the point features: normal vector, curvature and point feature histogram descriptor. The initial registration with rotation matrix and translation vector calculated from the corresponding points can be achieved. Then, improved ICP algorithm can be used to register accurately. This method is accurate and reliable with the test of aircrafts data.

Automatic registration of vestibular systems with exact landmark correspondence

Shape registration has a wide range of applications in geometric modeling, medical imaging, and computer vision. This paper focuses on the registration of the genus-3 vestibular systems and studies the geometric differences between the normal and Adolescent Idiopathic Scoliosis (AIS) groups. The non-trivial topology of the VS poses great technical challenges to the geometric analysis. To tackle these challenges, we present an effective and practical solution to register the vestibular systems. We first extract six geodesic landmarks for the VS, which are stable, intrinsic, and insensitive to the VS’s resolution and tessellation. Moreover, they are highly consistent regardless of the AIS and normal groups. The detected geodesic landmarks partition the VS into three patches, a topological annulus and two topological disks. For each pair of patches of the AIS subject and the control, we compute a bijective map using the holomorphic 1-form and harmonic map techniques. With a carefully designed boundary condition, the three individual maps can be glued in a seamless manner so that the resulting registration is a homeomorphism with exact landmark matching. Our method is robust, automatic and efficient. It takes only a few seconds on a low-end PC, which significantly outperforms the non-rigid ICP algorithm. We conducted a student’s t-test on the test data. Computational results show that using the mean curvature measure EH, our method can distinguish the AIS subjects and the normal subjects.

Inside Vehicle Inspection System Utilizing a Mobile Robot with LRF Sensor

Automated vehicle inspection is utilized to inspect vehicles in a parking lots. This paper presents an approach for inspecting vehicle inside through car windows to determine any changes since the last inspection using a Laser Range Finder (LRF) sensor. Features of the approach include the detection inside vehicles method, data alignment using an ICP algorithm, inside vehicle data comparison to find any differences since the last inspection. An item identification method has been used to obtain the average height and size of objects inside vehicles to identify such changes. Our approach was shown to successfully detect typical simple items, i.e., bags, notebook PCs, and wallets, used to test the proposed method. Experiments are conducted to demonstrate the efficiency of our approach for inspecting and recognizing objects inside vehicles.

Breast Image Registration for PET-CT and MR based on 3D Surface Matching

In this paper, we explore the PET-MRI 3D surface image registration method for the longitudinal study of cancer. Image registration gives new information to the radiologists by matching images with different modalities. The objective of this research is to develop 3D surface registration for PET/CT and MR images acquired by different medical imaging systems at different times. Our proposed method consists of three stages: 1) global surface matching, 2) fine adjustment stage and 3) image transformation. In this study, we perform the global registration with surface points within CT and MRI using ICP algorithm. Then we perform the fine adjusts stage using ECICP algorithm. Finally, PET image was transformed with various parameters and overlaid onto MR image.

A computational Method for Bilateral Symmetry Recognition in Asymmetrically Scanned Human Faces

This paper presents a new mirroring-and-registration method for the automatic symmetry plane detection of 3D asymmetrically scanned human faces. Once the mirroring of the original data is carried out with respect to the first-attempt symmetry plane, which is estimated by the PCA method, the source point cloud and the mirrored data are registered by the ICP algorithm that minimises a new weighted function. The final symmetry plane obtained approximates in the least-squares sense the midpoints of the lines connecting homologous points randomly chosen. This method is validated by analysing some specifically-designed test cases. The obtained results show that the method is quite insensitive to asymmetries of data resulting from the acquisition process.

Calculation of profile error for complex surface

Evaluation of profile error is important in manufacturing and detection of complex surface. In this paper, an algorithm is presented to calculate surface profile error accurately based on linear quadtree after rough localization and accurate registration. In rough localization, the best transformation is selected from four transformations. Then two algorithms are adopted to realize accurate registration and comparison is performed. One algorithm is the combination of LM algorithm and ICP algorithm. The other algorithm is BFGS algorithm. Experiment result shows that complex surface profile error is evaluated effectively through the proposed algorithm.

선박 블록 조립 후 최적 정도 계산을 위한 알고리즘 연구

본 연구에서는 기존의 블록 조립 후 정도 계산 절차와 블록 조립의 특성을 고려한 최적 정도 계산 알고리즘 개발을 위한 연구를 수행하였다. 여기서 제안된 알고리즘은 생산관리점들 중 특정한 관리점을 기준으로 생산관리점들의 설계와 측정 데이터 사이의 평균제곱근 오차의 합을 목적함수로 가진다. 생산관리점들은 접합면 상의 데이터와 그 외의 데이터로 구분하였으며, 구분된 데이터는 정합 과정에서 사용되어지는 6가지 자유도 조합 결정에 있어 다양한 제약조건 구성과 목적함수 계산에 사용하였다. 목적함수 및 제약조건과 함께 탑재공정을 고려하여 설계와 측정 계산 대상점들 간의 오차가 허용 오차 이내에 포함되는지를 확인하는 과정이 포함되는 점과 점 관계를 이용하는 변형 ICP 알고리즘과 sampling법을 혼합하여 최소 오차 범위를 계산하는 최적 정도 계산 알고리즘을 개발하였다. 실제 공정에서 확인된 블록 측정 데이터를 개발된 알고리즘에 적용한 결과에 따르면 최적 정도 계산의 대상점은 접합면 상의 점들만으로 계산을 수행하는 것보다 전체 점을 대상으로 계산하는 것이 더 작은 오차를 가지며 접합면의 한 점을 고정된 일치점으로 두고 모든 생산관리점들을 대상점으로 계산 하는 것이 최소 오차를 가지는 최적 정도 계산방법이라는 결론을 도출하였다. In this paper, an optimization algorithm for the block assembly accuracy control assessment is proposed with consideration for the current block assembly process and accuracy control procedure used in the shipbuilding site. The objective function of the proposed algorithm consists of root mean square error of the distances between design and measured data of the other control points with respect to a specific point of the whole control points. The control points are divided into two groups: points on the control line and the other points. The grouped data are used as criteria for determining the combination of 6 degrees of freedom in the registration process when constituting constraints and calculating objective function. The optimization algorithm is developed by using combination of the sampling method and the point to point relation based modified ICP algorithm which has an allowable error check procedure that makes sure that error between design and measured point is under allowable error. According to the results from the application of the proposed algorithm with the design and measured data of two blocks data which are verified and validated by an expert in the shipbuilding site, it implies that the choice of whole control points as target points for the accuracy calculation shows better results than that of the control points on the control line as target points for the accuracy of the calculation and the best optimized result can be acquired from the accuracy calculation with a fixed point on the control line as the reference point of the registration.

Spatio-temporal 3D surface matching for hand gesture recognition using ICP algorithm

With emerging new applications like virtual reality, different algorithms for human action and gesture recognition have been proposed. In this paper, a new method for the recognition of moving hand gestures is presented. The proposed algorithm is based on the representation of hand motion as spatio-temporal 3D surfaces. Then, 3D surface matching is used to recognize the hand gesture. To form the spatio-temporal 3D surface of hand motion, we first apply the necessary preprocessing to video frames and extract hand contours. Then, by normalizing and overlapping hand contours in different frames, we construct spatio-temporal 3D surface of the hand gesture. To recognize hand gesture, we match the input 3D surface with surfaces in the database. For this purpose, we utilize ICP algorithm to find and compensate for 3D transformation between surfaces as well as the similarity measure between them. In real-world applications, hand motion is continuous and results in a sequence of disjointed hand gestures, which is called continuous hand gesture. To recognize continuous hand gestures, we propose an algorithm which first estimates probable disjointed gestures in the continuous gesture and then divides iteratively continuous gestures to true disjointed gestures. Finally, by applying a robust algorithm, the continuous gesture is recognized. We tested the proposed algorithm with hand gestures of American sign language and results showed the recognition rate of 94 % for disjointed gestures and 93.9 % for continuous gestures. The experimental results showed the efficiency of the proposed algorithm for hand gestures with noise as well.

Polar-Cartesian Hybrid Transforms: A novel 2D range scan registration algorithm

Registration of range scans is commonly required in many localization and mapping algorithms. In this paper, we introduce a novel approach called Polar-Cartesian Hybrid Transforms for pair-wise registration of range scans. The proposed algorithm iteratively establishes correspondences by searching the points with closest polar angles in the polar coordinate frame. An angular look-up table is constructed based on the properties of the laser range finder to accelerate the searching procedure. In order to speed up the convergence, we compute the difference of polar range of every matched point pair to select the most contributing correspondences. After the correspondences are determined, the transformation is computed in Cartesian coordinate frame using a point-to-line metric. Combining the advantages of the polar and Cartesian coordinate frames, both robustness and efficiency are greatly improved compared with an up-to-date ICP algorithm.

Research on the Application of ICP Algorithm in Spatial Registration of Computer-Aided Surgery

This paper discusses the surface profile method based on image registration, and illustrates the principle and key points of registration algorithm based on surface profile. Canny operator is adopted to extract the surface contour and method based on quaternion and the singular value decomposition is used for space registration of the corresponding points, as well as the closest iterative search (ICP). In order to improve the registration speed, we adopt multiple step iteration method, which has been proved to be a reliable method according to the experiment.

Application of 3D Laser Scanner on Dike Risk Analysis

A dike risk model experiment is introduced in this paper. Researchers use a 3D laser scanner to scan the dike model before and after a simulated flood event and calculate the volume of the erosional fossa. The scan work includes four scan stations located at different places surrounding the dike model. The four point clouds are merged together by ICP (singular value decomposition) and SVD (singular value decomposition) algorithm to create a complete, thorough and detailed virtual 3D model of the experiment dike. Researchers can do kinds survey and calculate work based on the 3D model instead of working on the real dike. The work efficient and data quality are increased significantly.

A Method for Spatial Data Registration Based on PCA-ICP Algorithm

The spatial data which acquired by 3D laser scanning is huge, aiming at the iteration time is long with classic ICP algorithm, a improved registration algorithm of spatial data ICP algorithm which based on principal component analysis (PCA) is proposed in this paper (PCA-ICP), the basic principle and steps of PCA-ICP algorithm are given. The experiment results show that this method is feasible and the iterative time of PCA-ICP algorithm is shorter than classical ICP algorithm.

3D Reconstruction Using Low Precision Scanner

The objective of this work is to use low precision laser sensor and create reasonably precise 3D reconstructions. The 3D reconstruction is executed by combining several point clouds obtained from different viewpoints. The proposed method was developed with three main steps: point cloud registration, error compensation and surface reconstruction. The ICP algorithm is improved to execute the point cloud registration: dynamic distance threshold, weighted distance, rigid body restriction and color information. It is shown that using this improved ICP, the number of point correspondences to evaluate the quadratic error converges to a value. The quadratic error can be determined independently of scene complexity. The point cloud errors are compensated using the consensus surface algorithm with signed distance. The surface is reconstructed using the marching cubes algorithm. Several results are shown to demonstrate the reliability of the proposed method.