Registration Of Scans Research Articles

GPU-Accelerated 3D Normal Distributions Transform

The three-dimensional (3D) normal distributions transform (NDT) is a popular scan registration method for 3D point cloud datasets. It has been widely used in sensor-based localization and mapping applications. However, the NDT cannot entirely utilize the computing power of modern many-core processors, such as graphics processing units (GPUs), because of the NDT’s linear nature. In this study, we investigated the use of NVIDIA’s GPUs and their programming platform called compute unified device architecture (CUDA) to accelerate the NDT algorithm. We proposed a design and implementation of our GPU-accelerated 3D NDT (GPU NDT). Our methods can achieve a speedup rate of up to 34 times, compared with the NDT implemented in the point cloud library (PCL).

Lidar-Level Localization With Radar? The CFEAR Approach to Accurate, Fast, and Robust Large-Scale Radar Odometry in Diverse Environments

This paper presents an accurate, highly efficient, and learning-free method for large-scale odometry estimation using spinning radar, empirically found to generalize well across very diverse environments -- outdoors, from urban to woodland, and indoors in warehouses and mines - without changing parameters. Our method integrates motion compensation within a sweep with one-to-many scan registration that minimizes distances between nearby oriented surface points and mitigates outliers with a robust loss function. Extending our previous approach CFEAR, we present an in-depth investigation on a wider range of data sets, quantifying the importance of filtering, resolution, registration cost and loss functions, keyframe history, and motion compensation. We present a new solving strategy and configuration that overcomes previous issues with sparsity and bias, and improves our state-of-the-art by 38%, thus, surprisingly, outperforming radar SLAM and approaching lidar SLAM. The most accurate configuration achieves 1.09% error at 5Hz on the Oxford benchmark, and the fastest achieves 1.79% error at 160Hz.

Monitoring of erosive tooth wear: what to use and when to use it

Although we are increasingly recognising the need to assess patients for accelerated rates of tooth wear progression, it is often difficult to do so within a feasible diagnostic window. This paper aims to provide evidence-based timelines which a diagnosing clinician can expect to assess tooth wear progression in study models, clinical indices, clinical photographs and visually with intraoral scans. It also discusses new technologies emerging for the quantitative assessment of tooth wear, timelines for diagnosis, and caveats in the 3D scan registration and analysis process.

Accuracy comparison of bilateral versus complete arch interocclusal registration scans for virtual articulation

Statement of problemAlternatives to the bilateral interocclusal registration scanning technique to improve virtual articulation have not been fully investigated. PurposeThe purpose of this in vitro study was to compare the accuracy of virtually articulating digital casts by using bilateral interocclusal registration scans versus a complete arch interocclusal scan. Material and methodsA set of maxillary and mandibular reference casts were hand-articulated and mounted on an articulator. The mounted reference casts were scanned, and the maxillomandibular relationship record was scanned 15 times using 2 different scanning techniques, the bilateral interocclusal registration scan (BIRS) and complete arch interocclusal registration scan (CIRS), with an intraoral scanner. The generated files were transferred to a virtual articulator, and each set of scanned casts was articulated using BIRS and CIRS. The virtually articulated casts were saved as a set and transferred to a 3-dimensional (3D) analysis program. The scanned casts were set in the same coordinate system as the reference cast and overlaid on top of the reference cast for analysis. Two anterior and 2 posterior points were selected to determine points of comparison between the reference cast and test casts virtually articulated with BIRS and CIRS. The mean discrepancy between the 2 test groups and the anterior and posterior mean discrepancy within each group were tested for significance by using the Mann-Whitney U test (α=.05). ResultsA significant difference was found between the virtual articulation accuracy of BIRS and CIRS (P<.001). The mean deviation for BIRS was 0.053 ±0.051 mm and that for CIRS was 0.265 ±0.241 mm. Furthermore, significant differences were found between the anterior and posterior deviations in both BIRS (P=.020) and CIRS (P<.001). The mean deviation for BIRS was 0.034 ±0.026 mm in the anterior and 0.073 ±0.062 mm in the posterior. The mean deviation for CIRS was 0.146 ±0.108 mm anteriorly and 0.385 ±0.277 mm posteriorly. ConclusionsBIRS was more accurate than CIRS for virtual articulation. Moreover, the alignment accuracy of anterior and posterior sites for both BIRS and CIRS exhibited significant differences, with the anterior alignment exhibiting better accuracy in relation to the reference cast.

Prediction of adaptive strategies based on deformation vector field features for MR-guided adaptive radiotherapy of prostate cancer.

The selection of adaptive strategies in MR-guided adaptive radiotherapy (MRgART) usually relies on subjective review of anatomical changes. However, this kind of review may lead to improper selection of adaptive strategy for some fractions. The purpose of this study was to develop prediction models based on deformation vector field (DVF) features for automatic and accurate strategy selection, using prostate cancer as an example. 100 fractions of 20 prostate cancer patients were retrospectively selected in this study. Treatment plans using both adapt to position (ATP) strategy and adapt to shape (ATS) strategy were generated. Optimal adaptive strategy was determined according to dosimetric evaluation. DVFs of the deformable image registration (DIR) of daily MRI and CT simulation scans were extracted. The shape, first order statistics, and spatial features were extracted from the DVFs, subjected to further selection using the minimum redundancy maximum relevance (mRMR) method. The number of features (Fn ) was hyper-tuned using bootstrapping method, and then Fn indicating a peak area under the curve (AUC) value was used to construct three prediction models. According to subjective review, the ATS strategy was adopted for all 100 fractions. However, the evaluation results showed that the ATP strategy could have met the clinical requirements for 23 (23%) fractions. The three prediction models showed high prediction performance, with the best performing model achieving an AUC value of 0.896, corresponding accuracy (ACC), sensitivity (SEN) and specificity (SPC) of 0.9, 0.958, and 0.667, respectively. The features used to construct prediction models included four features extracted from y direction of DVF (DVFy ) and mask, one feature from z direction of DVF (DVFz ). It indicated that the deformation along the anterior-posterior direction had a greater impact on determining the adaptive strategy than other directions. DVF-feature-based models could accurately predict the adaptive strategy and avoid unnecessary selection of time-consuming ATS strategy, which consequently improves the efficiency of the MRgART process.

Intraoral scanning of the edentulous jaw without additional markers: An in vivo validation study on scanning precision and registration of an intraoral scan with a cone-beam computed tomography scan.

A fully digital approach to oral prosthodontic rehabilitation requires the possibility of combining (i.e., registering) digital documentation from different sources. This becomes more complex in an edentulous jaw, as fixed dental markers to perform reliable registration are lacking. This validation study aimed to evaluate the reproducibility of 1) intraoral scanning and 2) soft tissue-based registration of an intraoral scan with a cone-beam computed tomography (CBCT) scan for a fully edentulous upper jaw. Two observers independently performed intraoral scans of the upper jaw in 14 fully edentulous patients. The palatal vault of both surface models was aligned, and the inter-observer variability was assessed by calculating the mean inter-surface distance at the level of the alveolar crest. Additionally, a CBCT scan of all patients was obtained and a soft tissue surface model was generated using patient-specific gray values. This CBCT soft tissue model was registered with the intraoral scans of both observers, and the intraclass correlation coefficient (ICC) was calculated to evaluate the reproducibility of the registration method. The mean inter-observer deviation when performing an intraoral scan of the fully edentulous upper jaw was 0.10 ± 0.09 mm. The inter-observer agreement for the soft tissue-based registration method was excellent (ICC=0.94; 95% confidence interval, 0.81-0.98). Even when teeth are lacking, intraoral scanning of the jaw and soft tissue-based registration of an intraoral scan with a CBCT scan can be performed with a high degree of precision.

Rotational-motion measurement of the sacroiliac joint using upright MRI scanning and intensity-based registration: is there a sex difference?

PurposeThe sacroiliac joint (SIJ) has attracted increasing attention as a source of low back and groin pain, but the kinematics of SIJ against standing load and its sex difference remain unclear due to the difficulty of in vivo load study. An upright magnetic resonance imaging (MRI) system can provide in vivo imaging both in the supine and standing positions. The reliability of the mobility of SIJ against the standing load was evaluated and its sex difference was examined in healthy young volunteers using an upright MRI.MethodStatic (reliability) and kinematic studies were performed. In the static study, a dry bone of pelvic ring embedded in gel form and frozen in the plastic box was used. In the kinematic study, 19 volunteers (10 males, 9 females) with a mean age of 23.9 years were included. The ilium positions for the sacrum in supine and standing positions were measured against the pelvic coordinates to evaluate the mobility of the SIJ.ResultsIn the static study, the residual error of the rotation of the SIJ study was < 0.2°. In the kinematic study, the mean values of SIJ sagittal rotation from supine to standing position in males and females were − 0.9° ± 0.7° (mean ± standard deviation) and − 1.7° ± 0.8°, respectively. The sex difference was statistically significant (p = 0.04). The sagittal rotation of the SIJ showed a significant correlation with the sacral slope.ConclusionThe residual error for measuring the SIJ rotation using the upright MRI was < 0.2°. The young healthy participants showed sex differences in the sagittal rotation of the SIJ against the standing load and the females showed a larger posterior rotation of the ilium against the sacrum from the supine to standing position than the males. Therefore, upright MRI is useful to investigate SIJ motion.

The impact of cone beam computer tomography field of view on the precision of digital intra-oral scan registration for static computer-assisted implant surgery: A CBCT analysis.

Registration of intra-oral surface scans to cone beam computer tomography (CBCT) is critical in the digital workflow for static computer-aided implant surgery (sCAIS). This study aimed to assess the impact of CBCT field of view (FoV) on the precision of digital intra-oral scan registration. Cone beam computer tomography data and intra-oral scans from 20 patients were included. Small FoV CBCT's were created by digitally segmenting a large FoV into three sextants. Virtual implant planning was performed. Digital intra-oral scans were repeatedly registered onto their corresponding large and small FoV CBCT datasets. The distances and angulations between the matching implant positions of each repeated registration were used to determine the precision of the registration process. Wilcoxon Signed Rank Paired Tests were used to compare the differences between large FoV and small FoV. The threshold for statistical significance was set at p= .05. Differences in 3D implant position based on the registration precision between small FoV and large FoV present at both the implant entry point (0.37 ± 0.25 mm vs 0.35 ± 0.23 mm, p=.482) and implant tip (0.49 ± 0.34 mm vs 0.37 ± 0.24 mm, p < .001). Differences in overall angular precision were observed between small FOV and large FoV (1.43 ± 1.36° vs 0.51 ± 0.38°, p < .001). CBCT with a small FoV is accompanied by greater precision errors in intra-oral scan registration. However, when sufficient well-distributed teeth are visible in small FoV CBCT, the precision of digital intra-oral scan registration appears to be within clinically acceptable limits for sCAIS.

Robust landmark-based brain shift correction with a Siamese neural network in ultrasound-guided brain tumor resection.

In brain tumor surgery, tissue shift (called brain shift) can move the surgical target and invalidate the surgical plan. A cost-effective and flexible tool, intra-operative ultrasound (iUS) with robust image registration algorithms can effectively track brain shift to ensure surgical outcomes and safety. We proposed to employ a Siamese neural network, which was first trained using natural images and fine-tuned with domain-specific data to automatically detect matching anatomical landmarks in iUS scans at different surgical stages. An efficient 2.5D approach and an iterative re-weighted least squares algorithm are utilized to perform landmark-based registration for brain shift correction. The proposed method is validated and compared against the state-of-the-art methods using the public BITE and RESECT datasets. Registration of pre-resection iUS scans to during- and post-resection iUS images were executed. The results with the proposed method shows a significant improvement from the initial misalignment ([Formula: see text]) and the method is comparable to the state-of-the-art methods validated on the same datasets. We have proposed a robust technique to efficiently detect matching landmarks in iUS and perform brain shift correction with excellent performance. It has the potential to improve the accuracy and safety of neurosurgery.

Mitigating Shadows in LIDAR Scan Matching Using Spherical Voxels

In this paper we propose an approach to mitigate shadowing errors in Lidar scan matching, by introducing a preprocessing step based on spherical gridding. Because the grid aligns with the Lidar beam, it is relatively easy to eliminate shadow edges which cause systematic errors in Lidar scan matching. As we show through simulation, our proposed algorithm provides better results than ground-plane removal, the most common existing strategy for shadow mitigation. Unlike ground plane removal, our method applies to arbitrary terrains (e.g. shadows on urban walls, shadows in hilly terrain) while retaining key Lidar points on the ground that are critical for estimating changes in height, pitch, and roll. Our preprocessing algorithm can be used with a range of scan-matching methods; however, for voxel-based scan matching methods, it provides additional benefits by reducing computation costs and more evenly distributing Lidar points among voxels.

Reflections of Mythological Elements Used by Phrygians on Regional Tourism

Bu çalışmada, Frigya Bölgesinde yer alan Afyon, Kütahya ve Eskişehir kapsamında, Frig toplumuna ait kültürel ögelerin turizm alanında izdüşümlerine bakılmıştır. İzdüşümler belirlenirken, geleneksel mitolojilerden yararlanılmıştır. Bu amaçla, Frigya Bölgesinin tanıtımı için hazırlanmış olan web sayfaları üzerinden bir analiz gerçekleştirilmiştir. Analizlerde Frigya’ya ait kültürel ögelerin, bu bölgedeki turistik amaçlı kullanım durumu değerlendirilmiştir. Çalışmanın verileri, Afyon, Kütahya ve Eskişehir illerinin Belediye ve Valilik web sayfaları (toplam 49 web sayfası) üzerinden toplanmıştır. Araştırma kapsamında toplanan tüm veriler “Tarama Kayıt Formuna” manuel olarak kayıt edilmiştir. Taramalar, Aralık 2021 ile Şubat 2022 tarihleri arasında tamamlanmıştır. Yapılan incelemeler sonucunda elde edilen bazı mitolojik unsurların benzer özellikler gösterdiği tespit edilmiş ve bu ifadeler birleştirilerek yedi kategori belirlenmiştir. Kategoriler; gezilecek yerler, turistik işletmeler, müzeler, ekonomi, gastronomi, yerel ürünler, sanat ve festivaller olarak adlandırılmıştır. Araştırma sonuçları; bu üç şehrin gezilecek yerlerinin adlandırılmasında, müzelerinin temasında, gastronomisinde, turistik işletmelerdeki salon ve alan adlarında, yöresel ürünlerde, ekonomik yapıda, sanat ve festivallerde Friglerin izlerinin hüküm sürdüğünü göstermektedir.

3DupIC: An Underwater Scan Matching Method for Three-Dimensional Sonar Registration

This work presents a six degrees of freedom probabilistic scan matching method for registration of 3D underwater sonar scans. Unlike previous works, where local submaps are built to overcome measurement sparsity, our solution develops scan matching directly from the raw sonar data. Our method, based on the probabilistic Iterative Correspondence (pIC), takes measurement uncertainty into consideration while developing the registration procedure. A new probabilistic sensor model was developed to compute the uncertainty of each scan measurement individually. Initial displacement guesses are obtained from a probabilistic dead reckoning approach, also detailed in this document. Experiments, based on real data, demonstrate superior robustness and accuracy of our method with respect to the popular ICP algorithm. An improved trajectory is obtained by integration of scan matching updates in the localization data fusion algorithm, resulting in a substantial reduction of the original dead reckoning drift.

Computer-aided design and manufacture of a milled Twin-block: Workflow and use in a clinical patient. Have we entered the digital era?

• CAD/CAM technology was used to design and build a Twin-block appliance. • The mandibular advancement needed was simulated digitally using a virtual articulator. • This article shows a successful example of a milled twin-block fabrication and use. This article aimed to describe the fabrication and clinical use of a milled Twin-block with an expansion screw using intraoral scanning and digital bite registration without an analog impression or construction bite. Computer-aided design and manufacturing technology was used to design and build the appliance. The mandibular advancement needed was simulated digitally using a virtual articulator. The appliance was milled, and the metal components, including a prefabricated expansion screw, were inserted manually after the milling process had been completed. The Twin-block was used to treat a 9-year-old girl presenting an Angle Class II Division 1 malocclusion, a deepbite, and an overjet of 10 mm. The accurate fitting and activation of the computer-aided design and manufacturing appliance, the biocompatibility of the material used, and the achieved clinical outcome indicate that everyday orthodontic functional appliance therapy has entered a new era.

Pairwise versus Transitive Inverse Consistent Longitudinal Rigid Registration of Magnetic Resonance Images of Athletes With Repetitive Non-Concussive Head Injuries: Effects on Regional Distributions of Diffusion Measures.

Accurate alignment of longitudinal diffusion weighted imaging (DWI) scans of a subject is necessary to investigate longitudinal changes in DWI-derived diffusion measures such as fractional anisotropy (FA), mean diffusivity (MD), and quantitative anisotropy (QA). Currently, studies investigating these changes in the context of repetitive non-concussive head injuries (RHIs) perform pairwise rigid registration of all scans of a subject to the first scan or any other reference scan or template. Prajapati et.al 1 show that this strategy of performing pairwise rigid registration lead to a discrepancy in the rigid transformations. To eliminate this discrepancy, they propose performing transitive inverse consistent rigid registration of the longitudinal scans, and they analyze the impact of this approach on the mean values of the local/regional estimates of these diffusion measures. In this work, we further analyze the impact of transitive inverse consistent rigid registration on the distributions (CDFs) of the local/regional estimates of diffusion measures. We identify the regions (among the 48 anatomically defined regions by the JHU DTI-based white matter atlas2,3) that show significant differences in the CDFs obtained using pairwise inverse consistent and transitive inverse consistent rigid registration by performing the two sided Kolmogorov-Smirnov(KS) hypothesis test. We find that for MD and QA, there are certain subjects that have five or more regions with significant differences in the CDFs. Further, these are the same subjects for which Prajapati et.al 1 found regions with 2%-4% differences in the mean values of these diffusion measures. Thus, our results further strengthen the recommendation made by Prajapati et.al 1 to employ transitive inverse consistent rigid registration when investigating local/regional longitudinal changes in diffusion measures.

A new approach to symmetric registration of longitudinal structural MRI of the human brain

BackgroundThis paper presents the Automatic Temporal Registration Algorithm (ATRA) for symmetric rigid-body registration of longitudinal T1-weighted three-dimensional MRI scans of the human brain. This is a fundamental processing step in computational neuroimaging. New methodThe notion of leave-one-out consistent (LOOC) landmarks with respect to a supervised landmark detection algorithm is introduced. An automatic algorithm is presented for identification of LOOC landmarks on MRI scans. Multiple sets of LOOC landmarks are identified on each volume and a Generalized Orthogonal Procrustes Analysis of the landmarks is used to find a rigid-body transformation of each volume into a common space where the volumes are aligned precisely. ResultsQualitative and quantitative evaluations of ATRA registration accuracy were performed using 2012 volumes from 503 subjects (4 longitudinal volumes/subject), and on a further 120 volumes acquired from 3 normal subjects (40 longitudinal volumes/subject). Since the ground truth registrations are unknown, we devised a novel method for showing that ATRA’s registration accuracy is at least better than 0.5 mm translation or 0.5° rotation. Comparison with existing method(s)In comparison with existing methods, ATRA does not require any image preprocessing (e.g., skull-stripping or intensity normalization) and can handle conditions where rigid-body motion assumptions are not true (e.g., movement in eyes, jaw, neck) and brain tissue loss over time in neurodegenerative diseases. In a systematic comparison with the FSL FLIRT algorithm, ATRA provided faster and more accurate registrations. ConclusionsThe algorithm is symmetric, in the sense that any permutation of the input volumes does not change the transformation matrices, and unbiased, in that all volumes undergo exactly one interpolation operation, which precisely aligns them in a common space. There is no interpolation bias and no reference volume. All volumes are treated exactly the same. The algorithm is fast and highly accurate.

All-Weather, sub-50-cm, Radar-Inertial Positioning

Deploying automated ground vehicles beyond the confines of sunny and dry climes will require sub-lane-level positioning techniques that use radio waves, rather than near-visible light radiation. Like human sight, LiDAR and optical cameras perform poorly in low-visibility conditions. We present and demonstrate a novel technique for robust, sub-50-cm, urban ground vehicle positioning based on all-weather sensors. The technique incorporates a computationally-efficient, globally-optimal radar scan registration algorithm within a larger estimation pipeline that fuses data from commercially-available, low-cost, automotive radars, low-cost inertial sensors, vehicle motion constraints, and, when available, precise GNSS measurements. We evaluate the performance of the presented technique on an extensive and realistic urban dataset derived from all-weather sensors. Comparison against ground truth shows that during 60 min of GNSS-denied driving in the urban center of Austin, TX, the technique maintains 95th-percentile errors below 50 cm in horizontal position and 0.5 in heading.

MRI Distortion Correction and Robot-to-MRI Scanner Registration for an MRI-Guided Robotic System.

Magnetic resonance imaging (MRI) guided robotic procedures require safe robotic instrument navigation and precise target localization. This depends on reliable tracking of the instrument from MR images, which requires accurate registration of the robot to the scanner. A novel differential image based robot-to-MRI scanner registration approach is proposed that utilizes a set of active fiducial coils, where background subtraction method is employed for coil detection. In order to use the presented preoperative registration approach jointly with the real-time high speed MRI image acquisition and reconstruction methods in real-time interventional procedures, the effects of the geometric MRI distortion in robot to scanner registration is analyzed using a custom distortion mapping algorithm. The proposed approach is validated by a set of target coils placed within the workspace, employing multi-planar capabilities of the scanner. Registration and validation errors are respectively 2.05 mm and 2.63 mm after the distortion correction showing an improvement of respectively 1.08 mm and 0.14 mm compared to the results without distortion correction.

Target-based terrestrial laser scan registration extended by target orientation

Abstract In almost all projects, in which terrestrial laser scanning is used, the scans must be registered after the data acquisition. Despite more and more new and automated methods for registration, the classical target-based registration is still one of the standard procedures. The advantages are obvious: independence from the scan object, the geometric configuration can often be influenced and registration results are easy to interpret. When plane black-and-white targets are used, the algorithm for estimating the target center fits a plane through the scan of a target, anyway. This information about the plane orientation has remained unused so far. Hence, including this information in the registration does not require any additional effort in the scanning process. In this paper, we extend the target-based registration by the plane orientation. We describe the required methodology, analyze the benefits in terms of precision and reliability and discuss in which cases the extension is useful and brings a relevant advantage. Based on simulations and two case studies we find out that especially for registrations with bad geometric configurations the extension brings a big advantage. The extension enables registrations that are much more precise. These are also visible on the registered point clouds. Thus, only a methodological change in the target-based registration improves its results.

Individualized cyclic mechanical loading improves callus properties during the remodelling phase of fracture healing in mice as assessed from time-lapsed in vivo imaging

Fracture healing is regulated by mechanical loading. Understanding the underlying mechanisms during the different healing phases is required for targeted mechanical intervention therapies. Here, the influence of individualized cyclic mechanical loading on the remodelling phase of fracture healing was assessed in a non-critical-sized mouse femur defect model. After bridging of the defect, a loading group (n = 10) received individualized cyclic mechanical loading (8–16 N, 10 Hz, 5 min, 3 × /week) based on computed strain distribution in the mineralized callus using animal-specific real-time micro-finite element analysis with 2D/3D visualizations and strain histograms. Controls (n = 10) received 0 N treatment at the same post-operative time-points. By registration of consecutive scans, structural and dynamic callus morphometric parameters were followed in three callus sub-volumes and the adjacent cortex showing that the remodelling phase of fracture healing is highly responsive to cyclic mechanical loading with changes in dynamic parameters leading to significantly larger formation of mineralized callus and higher degree of mineralization. Loading-mediated maintenance of callus remodelling was associated with distinct effects on Wnt-signalling-associated molecular targets Sclerostin and RANKL in callus sub-regions and the adjacent cortex (n = 1/group). Given these distinct local protein expression patterns induced by cyclic mechanical loading during callus remodelling, the femur defect loading model with individualized load application seems suitable to further understand the local spatio-temporal mechano-molecular regulation of the different fracture healing phases.

Efficient LiDAR Odometry for Autonomous Driving

LiDAR odometry plays an important role in self-localization and mapping for autonomous navigation, which is usually treated as a scan registration problem. Although having achieved promising performance on the KITTI odometry benchmark, the conventional tree-based neighbor search still has the difficulty in dealing with the large-scale point cloud efficiently. The recent spherical range image-based method enjoys the merits of fast nearest neighbor search by spherical mapping. However, it is not very effective to deal with the ground points nearly parallel to LiDAR beams. To address these issues, we propose a novel efficient LiDAR odometry approach by taking advantage of both non-ground spherical range images and bird's-eye-view maps for ground points. Moreover, a range adaptive method is introduced to robustly estimate the local surface normal. Additionally, a very fast and memory-efficient model update scheme is proposed to fuse the points and their corresponding normals at different time-stamps. We have conducted extensive experiments on the KITTI odometry benchmark and UrbanLoco dataset, whose promising results demonstrate that our proposed approach is effective.