Surface Segmentation Research Articles

The segmentation and intelligent recognition of structural surfaces in borehole images based on the U2-Net network.

Structural planes decrease the strength and stability of rock masses, severely affecting their mechanical properties and deformation and failure characteristics. Therefore, investigation and analysis of structural planes are crucial tasks in mining rock mechanics. The drilling camera obtains image information of deep structural planes of rock masses through high-definition camera methods, providing important data sources for the analysis of deep structural planes of rock masses. This paper addresses the problems of high workload, low efficiency, high subjectivity, and poor accuracy brought about by manual processing based on current borehole image analysis and conducts an intelligent segmentation study of borehole image structural planes based on the U2-Net network. By collecting data from 20 different borehole images in different lithological regions, a dataset consisting of 1,013 borehole images with structural plane type, lithology, and color was established. Data augmentation methods such as image flipping, color jittering, blurring, and mixup were applied to expand the dataset to 12,421 images, meeting the requirements for deep network training data. Based on the PyTorch deep learning framework, the initial U2-Net network weights were set, the learning rate was set to 0.001, the training batch was 4, and the Adam optimizer adaptively adjusted the learning rate during the training process. A dedicated network model for segmenting structural planes was obtained, and the model achieved a maximum F-measure value of 0.749 when the confidence threshold was set to 0.7, with an accuracy rate of up to 0.85 within the range of recall rate greater than 0.5. Overall, the model has high accuracy for segmenting structural planes and very low mean absolute error, indicating good segmentation accuracy and certain generalization of the network. The research method in this paper can serve as a reference for the study of intelligent identification of structural planes in borehole images.

DilatedToothSegNet: Tooth Segmentation Network on 3D Dental Meshes Through Increasing Receptive Vision

The utilization of advanced intraoral scanners to acquire 3D dental models has gained significant popularity in the fields of dentistry and orthodontics. Accurate segmentation and labeling of teeth on digitized 3D dental surface models are crucial for computer-aided treatment planning. At the same time, manual labeling of these models is a time-consuming task. Recent advances in geometric deep learning have demonstrated remarkable efficiency in surface segmentation when applied to raw 3D models. However, segmentation of the dental surface remains challenging due to the atypical and diverse appearance of the patients’ teeth. Numerous deep learning methods have been proposed to automate dental surface segmentation. Nevertheless, they still show limitations, particularly in cases where teeth are missing or severely misaligned. To overcome these challenges, we introduce a network operator called dilated edge convolution, which enhances the network’s ability to learn additional, more distant features by expanding its receptive field. This leads to improved segmentation results, particularly in complex and challenging cases. To validate the effectiveness of our proposed method, we performed extensive evaluations on the recently published benchmark data set for dental model segmentation Teeth3DS. We compared our approach with several other state-of-the-art methods using a quantitative and qualitative analysis. Through these evaluations, we demonstrate the superiority of our proposed method, showcasing its ability to outperform existing approaches in dental surface segmentation.

Semi-supervised auxiliary learning for surface defect detection and segmentation of injection-molded products from small image datasets

Semi-supervised auxiliary learning for surface defect detection and segmentation of injection-molded products from small image datasets

Craniofacial phenotyping with fetal MRI: a feasibility study of 3D visualisation, segmentation, surface-rendered and physical models

This study explores the potential of 3D Slice-to-Volume Registration (SVR) motion-corrected fetal MRI for craniofacial assessment, traditionally used only for fetal brain analysis. In addition, we present the first description of an automated pipeline based on 3D Attention UNet trained for 3D fetal MRI craniofacial segmentation, followed by surface refinement. Results of 3D printing of selected models are also presented.Qualitative analysis of multiplanar volumes, based on the SVR output and surface segmentations outputs, were assessed with computer and printed models, using standardised protocols that we developed for evaluating image quality and visibility of diagnostic craniofacial features. A test set of 25, postnatally confirmed, Trisomy 21 fetal cases (24–36 weeks gestational age), revealed that 3D reconstructed T2 SVR images provided 66–100% visibility of relevant craniofacial and head structures in the SVR output, and 20–100% and 60–90% anatomical visibility was seen for the baseline and refined 3D computer surface model outputs respectively. Furthermore, 12 of 25 cases, 48%, of refined surface models demonstrated good or excellent overall quality with a further 9 cases, 36%, demonstrating moderate quality to include facial, scalp and external ears. Additional 3D printing of 12 physical real-size models (20–36 weeks gestational age) revealed good/excellent overall quality in all cases and distinguishable features between healthy control cases and cases with confirmed anomalies, with only minor manual adjustments required before 3D printing.Despite varying image quality and data heterogeneity, 3D T2w SVR reconstructions and models provided sufficient resolution for the subjective characterisation of subtle craniofacial features. We also contributed a publicly accessible online 3D T2w MRI atlas of the fetal head, validated for accurate representation of normal fetal anatomy.Future research will focus on quantitative analysis, optimizing the pipeline, and exploring diagnostic, counselling, and educational applications in fetal craniofacial assessment.

Application of geopolymer in synchronous grouting for reusing of the shield muck in silty clay layer

This study employs geopolymer as a complete replacement for the cement to solidify the shield muck discharged from tunneling in the silty clay layer, thereby creating an innovative synchronous grouting material. This novel grouting material not only utilizes geopolymer as its primary cementitious component but also incorporates shield muck, promoting the efficient utilization of the shield muck. The cementitious process of the grout is initiated by the principles of geopolymerization. Through a series of laboratory tests assessing grout performance, the new grouting material was developed by further optimizing the grout proportion using the response surface method (RSM). This involves a comprehensive analysis of the impacts of various factors such as muck moisture content, foam injection ratio, ground granulated blast furnace slag (GGBS) content, and sand content, along with their interactions on the key response variables, including consistency, fluidity, 2 h bleeding rate, setting time, and 3 d unconfined compressive strength (UCS). The optimal proportion of the new grouting material is found to be 140% muck moisture, 100% GGBS, and 230% sand. Experimental results reveal that the GGBS content has the most significant effect on various properties of the grout, while the interaction between these factors has distinct effects on different grout properties. This novel synchronous grouting material was subsequently employed in the Nanjing Metro Line 6 project, and it was found that both surface settlement and segment uplift meet the specified requirements.

ISRM: introspective self-supervised reconstruction model for rail surface defect detection and segmentation

The problems of intrinsic imbalance of the sample and interference from complex backgrounds limit the performance of existing deep learning methods when applied to the detection and segmentation of rail surface defects. To address these issues, an introspective self-supervised reconstruction model (ISRM) is proposed, which only requires normal samples in the training phase and incorporates the concept of self-supervised learning into an introspective autoencoder. The training framework of ISRM first extracts general features using a pretrained Feature Extractor. Subsequently, a Feature Transformer transfers the features to the target domain. Next, a synthetic defect embedder embeds Bessel-Gaussian random defects into the feature space. Finally, the asymmetric autoencoder reconstructs the rail surface features back into image space. The transformation of pretrained features into target-oriented features helps mitigate domain bias. Since defects exhibit higher commonality in the feature space relative to the image space, embedding synthetic defects into the feature space effectively improves training efficiency. Moreover, the adversarial training architecture enhances the clarity of reconstructed images. The impact of core parameters on the model performance is analyzed through ablation experiments. The results from comparative experiments demonstrate that ISRM achieves 98.5% and 97.2% accuracy on defect detection and segmentation tasks, respectively, reducing the error rate by 11.8% and 3.4% compared to the current state-of-the-art model.

Characterization of a Novel TiF4 Inclusion Complex and in vitro Evaluation of Its Effect on Inhibiting Enamel Demineralization

Introduction: Titanium tetrafluoride (TiF4) is an anticariogenic agent with high remineralizing potential. However, the acidic pH of TiF4 solution can limit its clinical application. The present study aimed to prepare and characterize a new TiF4-dendrimer inclusion complex and evaluate its ability to inhibit enamel demineralization under pH cycling conditions. Methods: PEG-citrate dendrimer and TiF4-dendrimer inclusion complex were synthesized and their molecular structures were evaluated using Fourier-transform Infrared Spectroscopy (FTIR), Hydrogen Nuclear Magnetic Resonance (HNMR), and Liquid Chromatography-Mass Spectrometry (LC-MS) tests. Forty-eight enamel samples were prepared and randomly divided into four groups: distilled water (negative control), TiF4 solution (T), dendrimer solution (D), and TiF4-dendrimer solution (TD). The microhardness of the samples was measured initially. Next, the samples underwent pH cycling, were exposed to the solutions, the microhardness was measured again, and microhardness loss was calculated. EDX analysis was performed on the surface and cross-sectional segments of the samples. Results: The microhardness loss was significantly higher in control (−65.1 ± 6.0) compared to other groups. No significant difference was observed between T (−47.9 ± 5.6) and D (−41.7 ± 12.0) and also D and TD (−40.5 ± 9.4) in this regard. Microhardness loss was significantly higher in T compared to TD group. The TD samples showed similar fluoride and titanium content in both surface and subsurface regions, while the T group had higher concentrations in the surface region. Moreover, the TD solution had a higher pH of 3.4 compared to the T solution’s pH of 1.1. Conclusion: No significant difference was observed between the efficacy of TiF4-dendrimer and TiF4 solution in inhibiting demineralization while TiF4-dendrimer solution had the added advantage of having a higher pH.

Automated flatness assessment for large quantities of full‐scale precast beams using laser scanning

AbstractPrefabrication has been widely used in bridge construction, for which precast beams are produced from a beam yard and constructed with a cast‐in‐suit bridge deck. The developments recently are focusing on large dimensions or large quantities of beam units, which leads to the inevitable challenge of beam quality control. Among them, beam surface flatness as one of the important indicators for manufacturing quality needs to be strictly controlled to ensure structural performance. However, the commonly used methods still rely heavily on manually contact‐based inspection, which is not only time‐consuming but also error‐prone. This paper presents an automated flatness assessment for large quantities of full‐scale precast beams by using terrestrial laser scanning techniques. Through axis calibration, target surface segmentation, and flatness deviation calculation, flatness deviations of the segmented surfaces from the point cloud of the entire beam are calculated, and colored maps are generated to visually assess the flatness quality. The proposed approach was applied to numerical simulated models as well as laboratory‐scale specimens for accuracy validation. Besides, its processing time was also compared to conventional contact‐based inspection during a field test. Moreover, efforts are devoted to applying the proposed method on nearly a thousand precast beams in an in‐field bridge construction project. Results can demonstrate its outperformance not only on accuracy but also the efficiency and practicality when applied to large quantities of full‐scale precast beams.

Classifications of anterior segment structure of congenital corneal opacity in infants and toddlers by ultrasound biomicroscopy and slit-lamp microscopic photographs: an observational study

BackgroundThe structural features have an impact on the surgical prognosis for congenital corneal opacity (CCO). The structural classification system of CCO, however, is lacking. Based on data from ultrasound biomicroscopy (UBM) findings in infants and toddlers with CCO, this research proposed a classification system for the anterior segment structure severity.MethodsMedical records, preoperative UBM images and slit-lamp photographs of infants and toddlers diagnosed with CCO at University Third Hospital between December 2018 and June 2022 were reviewed. According to the anterior segment structural features observed in UBM images, eyes were classified as follows: U1, opaque cornea only; U2, central anterior synechia; U3, peripheral anterior synechia combined with angle closure; and U4, aniridia or lens anomaly. The opacity appearance and corneal vascularization density observed in slit-lamp photographs were assigned grades according to previous studies. The extent of vascularization was also recorded. The corresponding intraocular anomaly classifications and ocular surface lesion severity were analysed.ResultsAmong 81 eyes (65 patients), 41 (50.6%) were right eyes, and 40 (49.4%) were left eyes. The median age at examination was 6.91 months (n = 81, 1.00, 34.00). Two (2.5%) of the 81 eyes were classified as U1, 20 (24.7%) as U2, 22 (27.2%) as U3a, 11 (13.6%) as U3b and 26 (32.1%) as U4. Bilateral CCO eyes had more severe UBM classifications (P = 0.019), more severe dysgenesis (P = 0.012) and a larger angle closure (P = 0.009). Eyes with more severe UBM classifications had higher opacity grades (P = 0.003) and vascularization grades (P = 0.014) and a larger vascularization extent (P = 0.001). Eyes with dysgenesis had higher haze grades (P = 0.012) and more severe vascularization (P = 0.003 for density; P = 0.008 for extent), while the angle closure range was related to haze grade (P = 0.013) and vascularization extent (P = 0.003).ConclusionsThis classification method based on UBM and slit-lamp photography findings in the eyes of CCO infants and toddlers can truly reflect the degree of abnormality of the ocular surface and anterior segment and is correlated with the severity of ocular surface anomalies. This method might provide meaningful guidance for surgical procedure design and prognostic determinations for keratoplasty in CCO eyes.

Magnetotransport and Fermi surface segmentation in Pauli limited superconductors

We report the first theoretical investigation of the spectroscopic, electrical and optical transport signatures of d-wave Pauli limited superconductors, based on a non perturbative numerical approach. We demonstrate that the high magnetic field low temperature regime of these materials host a finite momentum paired superconducting phase. Multi-branched dispersion spectra with finite energy superconducting gaps, anisotropic segmentation of the Fermi surface and spatial modulations of the superconducting order characterizes this finite momentum paired phase and should be readily accessible through angle resolved photo emission spectroscopy, quasiparticle interference and differential conductance measurements. Based on the electrical and optical transport properties we capture the non Fermi liquid behavior of these systems at high temperatures, dominated by local superconducting correlations and characterized by resilient quasiparticles which survive the breakdown of the Fermi liquid description. We map out the generic thermal phase diagram of the d-wave Pauli limited superconductors and provide for the first time the accurate estimates of the thermal scales corresponding to the: (a) loss of (quasi) long range superconducting phase coherence (T c), (b) loss of local pair correlations (T pg), (c) breakdown of the Fermi liquid theory (T max) and cross-over from the non Fermi liquid to the bad metallic phase (T BR). Our thermal phase diagram mapped out on the basis of the spectroscopic and transport properties are found to be in qualitative agreement with the experimental observations on CeCoIn5 and κ-BEDT, in terms of the thermodynamic phases and the phase transitions. The results presented in this paper are expected to initiate important transport and spectroscopic experiments on the Pauli limited d-wave superconductors, providing sharp signatures of the finite momentum Cooper paired state in these materials.

Unsupervised Dual Transformer Learning for 3-D Textured Surface Segmentation.

Analysis of the 3-D texture is indispensable for various tasks, such as retrieval, segmentation, classification, and inspection of sculptures, knit fabrics, and biological tissues. A 3-D texture represents a locally repeated surface variation (SV) that is independent of the overall shape of the surface and can be determined using the local neighborhood and its characteristics. Existing methods mostly employ computer vision techniques that analyze a 3-D mesh globally, derive features, and then utilize them for classification or retrieval tasks. While several traditional and learning-based methods have been proposed in the literature, only a few have addressed 3-D texture analysis, and none have considered unsupervised schemes so far. This article proposes an original framework for the unsupervised segmentation of 3-D texture on the mesh manifold. The problem is approached as a binary surface segmentation task, where the mesh surface is partitioned into textured and nontextured regions without prior annotation. The proposed method comprises a mutual transformer-based system consisting of a label generator (LG) and a label cleaner (LC). Both models take geometric image representations of the surface mesh facets and label them as texture or nontexture using an iterative mutual learning scheme. Extensive experiments on three publicly available datasets with diverse texture patterns demonstrate that the proposed framework outperforms standard and state-of-the-art unsupervised techniques and performs reasonably well compared to supervised methods.

Total Generalized Variation Regularized Piecewise Smooth Mumford-Shah Point Cloud Surface Segmentation

Total Generalized Variation Regularized Piecewise Smooth Mumford-Shah Point Cloud Surface Segmentation

Versatile Curve Design by Level Set with Quadratic Convergence.

Many 3D mesh processing tasks revolve around generating and manipulating curves on surface meshes. While it is intuitive to explicitly model these curves using mesh edges or parametric curves in the ambient space, these methods often suffer from numerical instability or inaccuracy due to the projection operation. Another natural strategy is to adapt spline based tools, these methods are quite fast but are hard to be extended to more versatile constraints and need heavy manual interactions. In this paper, we present an efficient and versatile approach to curve design based on an implicit representation known as the level set. While previous works have explored the use of the level set to generate curves with minimal length, they typically have limitations in accommodating additional conditions for rich and robust control. To address these challenges, we formulate curve editing with constraints like smoothness, interpolation, tangent control, etc., via a level set based variational problem by constraining the values or derivatives of the level set function. However, the widely used gradient flow strategy converges very slowly for this complicated variational problem compared to the classical geodesic one. Thus, we propose to solve it via Newton's method enhanced by local Hessian correction and a trust-region strategy. As a result, our method not only enables versatile control, but also excels in terms of performance due to nearly quadratic convergence and almost linear complexity in each iteration via narrow band acceleration. In practice, these advantages effectively benefit various applications, such as interactive curve manipulation, boundary smoothing for surface segmentation and path planning with obstacles as demonstrated.

RailSegVITNet: A lightweight VIT-based real-time track surface segmentation network for improving railroad safety

To address the challenge of precise foreign object detection on railway tracks, this study proposes RailSegVITNet, an efficient deep learning-based model that aims to accurately locate tracks and provide necessary information. RailSegVITNet integrates lightweight bottleneck blocks, separable self-attention, and feature aggregation to balance real-time performance and accuracy. It follows an encoder–decoder framework, with bottleneck blocks used for feature extraction and separable self-attention integrated at different stages to enhance feature extraction capability. The decoder utilizes feature aggregation to effectively merge multi-resolution features and improve segmentation performance. On the proposed Railway-seg dataset, the RailSegVITNet demonstrates competitive performance, achieving an MIoU (Mean Intersection over Union ) of 91.43% while utilizing only 2.01 G FLOPs (Floating Point Operations) and 1.4 M parameters. Furthermore, compared to popular models such as Topformer, Segformer, Segmenter, and Deeplab on the publicly available Railsem19 dataset, RailSegVITNet maintains its lightweight architecture while achieving comparable or higher segmentation performance. This showcases its potential practical value for efficient identification of foreign object intrusion.

An Automatic Road Surface Segmentation in Non-Urban Environments: A 3D Point Cloud Approach With Grid Structure and Shallow Neural Networks

An Automatic Road Surface Segmentation in Non-Urban Environments: A 3D Point Cloud Approach With Grid Structure and Shallow Neural Networks

DeepAqua: Semantic segmentation of wetland water surfaces with SAR imagery using deep neural networks without manually annotated data

Deep learning and remote sensing techniques have significantly advanced water surface monitoring; however, the need for annotated data remains a challenge. This is particularly problematic in wetland detection, where water extent varies over time and space, demanding multiple annotations for the same area. In this paper, we present DeepAqua, a deep learning model inspired by knowledge distillation (a.k.a. teacher–student model) to generate labeled data automatically and eliminate the need for manual annotations during the training phase. We utilize the Normalized Difference Water Index (NDWI) as a teacher model to train a Convolutional Neural Network (CNN) for segmenting water from Synthetic Aperture Radar (SAR) images. To train the student model, we exploit cases where optical- and radar-based water masks coincide, enabling the detection of both open and vegetated water surfaces. DeepAqua represents a significant advancement in computer vision techniques for water detection by effectively training semantic segmentation models without any manually annotated data. Experimental results show that DeepAqua outperforms other unsupervised methods by improving accuracy by 3%, Intersection Over Union by 11%, and F1-score by 6%. This approach offers a practical solution for monitoring wetland water extent changes without the need of ground truth data, making it highly adaptable and scalable for wetland monitoring.

TEXTURE-BASED SEPARATION TO REFINE BUILDING MESHES

Abstract. In this paper, we present a texture-based separation approach to refine building meshes, which aims to address the challenges of detecting and isolating different objects in an indoor scene mesh. We propose a novel segmentation model based on the materials of the different parts of the scene. The proposed approach uses factorization-based texture segmentation to separate the different materials in the meshes and detect the edges on the segmented texture. To prevent large faces containing multiple materials from being segmented wrongly, the mesh is cut along the texture boundaries and new faces with a single material are created. Finally, we segment the new faces based on the material index and neighbouring faces using a region-growing algorithm. We evaluate the proposed approach on the Matterport indoor dataset, which shows that our approach performs well on detecting boundaries between distinct materials, but over-segments on complex shapes. Our proposed method improves the segmentation of large flat surfaces like posters or rugs.

Learning joint surface reconstruction and segmentation, from brain images to cortical surface parcellation.

Reconstructing and segmenting cortical surfaces from MRI is essential to a wide range of brain analyses. However, most approaches follow a multi-step slow process, such as a sequential spherical inflation and registration, which requires considerable computation times. To overcome the limitations arising from these multi-steps, we propose SegRecon, an integrated end-to-end deep learning method to jointly reconstruct and segment cortical surfaces directly from an MRI volume in one single step. We train a volume-based neural network to predict, for each voxel, the signed distances to multiple nested surfaces and their corresponding spherical representation in atlas space. This is, for instance, useful for jointly reconstructing and segmenting the white-to-gray-matter interface and the gray-matter-to-CSF (pial) surface. We evaluate the performance of our surface reconstruction and segmentation method with a comprehensive set of experiments on the MindBoggle, ABIDE and OASIS datasets. Our reconstruction error is found to be less than 0.52 mm and 0.97 mm in terms of average Hausdorff distance to the FreeSurfer generated surfaces. Likewise, the parcellation results show over 4% improvements in average Dice with respect to FreeSurfer, in addition to an observed drastic speed-up from hours to seconds of computation on a standard desktop station.

Multiple damage segmentation and extraction of wind turbine blades surface under complex background based on SKRT approach

Machine vision detection technology has been widely used in detecting wind turbine blade surface damage, but the complex background often has a significant impact on blade damage detection. Aiming at the complex background blade images caused by various natural background features, complex blade surface features, low contrast features, non-uniform illumination, weather conditions, and other factors, the work proposes an SKRT segmentation method based on K-means clustering and adaptive threshold fusion to segment and extract multiple damage features on the surface of wind turbine blades with complex backgrounds. Firstly, an single-scale retinex(SSR) algorithm is adopted to enhance the contrast between surface damage and image background in the research. Then the image is roughly segmented by K-means clustering to remove most of the background features, and the background mask is replaced with the peak of the remaining feature image pixels to solve the problem of wrong segmentation. Finally, making use of the adaptive threshold local segmentation method to accurately segment the blade surface damage. The experimental results show that the proposed SKRT segmentation method can significantly improve the segmentation accuracy of wind turbine blade surface damage with complex background, and the accuracy, intersection over union, and F-measure value are increased by 37.20%, 35.71%, and 28.69%, respectively. The method performs better robustness to multiple damage feature segmentation.

Design and Implementation of a Hardware and Software System for Visual Assessment of Bituminous Coating Quality

Typically, the quality of the bitumen adhesion in asphalt mixtures is assessed manually by a group of experts who assign subjective ratings to the thickness of the residual bitumen coating on the gravel samples. To automate this process, we propose a hardware and software system for visual assessment of bituminous coating quality, which provides the results both in the form of a discrete estimate compatible with the expert one, and in a more general percentage for a set of samples. The developed methodology ensures static conditions of image capturing, insensitive to external circumstances. This is achieved by using a hardware construction designed to provide capturing the samples at eight different illumination angles. As a result, a generalized image is obtained, in which the effect of highlights and shadows is eliminated. After preprocessing, each gravel sample independently undergoes surface semantic segmentation procedure. Two most relevant approaches of semantic image segmentation were considered: gradient boosting and U-Net architecture. These approaches were compared by both stone surface segmentation accuracy, where they showed the same 77% result and the effectiveness in determining a discrete estimate. Gradient boosting showed an accuracy 2% higher than the U-Net for it and was thereby chosen as the main model when developing the prototype. According to the test results, the evaluation of the algorithm in 75% of cases completely coincided with the expert one, and it had a slight deviation from it in another 22% of cases. The developed solution allows for standardizing the data obtained and contributes to the creation of an interlaboratory digital research database.