Precise 3D Reconstruction Research Articles

Study on the impact of camera ISP algorithms on stripe structured light 3D reconstruction and the improvement of 3D reconstruction accuracy by photon input

This paper investigates the impact of camera image signal processing (ISP) algorithms on stripe structured light 3D reconstruction and explores the improvement of 3D reconstruction accuracy by photon input. Existing 3D reconstruction technologies have broad applications in fields such as intelligent manufacturing, healthcare, and consumer electronics, but their high precision requirements often cannot be met by current ISP algorithms. By using handheld devices to capture images and combining key techniques such as the multi-step phase shifting method, multi-frequency phase unwrapping method, and triangulation method, this paper conducts an in-depth study on the application of photon input in 3D reconstruction. The study demonstrates that using non-visual information (RAW images) as input can significantly improve reconstruction accuracy, producing more accurate results compared to images processed by ISP. The paper also quantifies the impact of ISP processing on 3D reconstruction results by comparing the depth information and point cloud data of two sets of images. Experimental results show that disabling certain ISP algorithms, such as bilateral noise filtering (BNF), edge enhancement (EEH), and non-local means denoising (NLM), can further improve reconstruction accuracy and reduce errors. In conclusion, this paper proposes a photon image-based 3D reconstruction method that, combined with artificial intelligence technology and differentiable point cloud rendering techniques, holds promise for achieving higher precision and faster 3D reconstruction. This technology is of great significance in practical applications, particularly in the field of industrial close-range 3D reconstruction.

Stereo matching and 3D reconstruction with NeRF supervision for accurate weight estimation in free-swimming fish

Non-invasive methods of accurate estimation of fish weight are essential for biomass assessment, precision feeding, and effective management in aquaculture. However, the lack of large-scale stereo matching datasets and costly depth sensing equipment in underwater scenarios present significant challenges to accurately measure the size of free-swimming fish. This study introduces a novel NeRF-supervised stereo matching network for precise depth estimation and 3D reconstruction of the fish body. The state-of-the-art neural rendering methods are employed to generate stereo training data from image sequences. In this study, the residual network with deformable convolutions was proposed to extract contextual texture features, significantly enhancing disparity estimation accuracy at the edge of the fish body. Additionally, multiscale spatial features are fused using a feature pyramid network (FPN). To enhance the adaptability of disparity search, we introduce a lightweight Triplet Attention module to capture cross-dimensional dependencies prior to constructing the correlation volume. Experimental results demonstrate the effectiveness of the proposed depth estimation method, achieving an impressive result of 92.35% in the similarity metric, compared to the ground truth. Moreover, we developed a weight estimation model based on bass perimeter using instance segmentation and 3D reconstruction techniques. Notably, a robust correlation (correlation coefficient of 0.98) between fish perimeter and weight is observed, highlighting the potential application of our methods to enhance the accuracy of weight estimation for free-swimming fish.

Method for 3D light-field reconstruction based on a spatially free camera

This study introduces an advanced algorithm for 3D light-field displays, utilizing disparity images from non-fixed cameras to achieve precise 3D reconstructions, moving beyond the complexities of traditional methods. Unlike standard approaches that depend on complex 3D reconstruction or specific camera arrangements, this algorithm uses pixel similarity across different viewpoints to determine depth accurately. It integrates camera parameters and light paths to restore light-field information meticulously, ensuring an optimal light distribution. By comparing with the projector image array (PIA) and display effects generated by traditional light-field encoding algorithms, we proved the effectiveness of our algorithm.

Separate and Integrated Data Processing for the 3D Reconstruction of a Complex Architecture

Abstract. In the last few years, data fusion has been an active research topic for the expected advantages of exploiting and combining different but complementary techniques for 3D documentation. The data fusion process consists of merging data coming from different sensors and platforms, intrinsically different, to produce complete, coherent, and precise 3D reconstructions. Although extensive research has been dedicated to this task, we still have many gaps in the integration process, and the quality of the results is hardly sufficient in several cases. This is especially evident when the integration occurs in a later stage, e.g., merging the results of separate data processing. New opportunities are emerging, with the possibility offered by some proprietary tools to jointly process heterogeneous data, particularly image and range-based data. The article investigates the benefits of data integration at different processing levels: raw, middle, and high levels. The experiments are targeted to explore, in particular, the results of the integration on large and complex architectures.

PR3D: Precise and realistic 3D face reconstruction from a single image

AbstractReconstructing the three‐dimensional (3D) shape and texture of the face from a single image is a significant and challenging task in computer vision and graphics. In recent years, learning‐based reconstruction methods have exhibited outstanding performance, but their effectiveness is severely constrained by the scarcity of available training data with 3D annotations. To address this issue, we present the PR3D (Precise and Realistic 3D face reconstruction) method, which consists of high‐precision shape reconstruction based on semi‐supervised learning and high‐fidelity texture reconstruction based on StyleGAN2. In shape reconstruction, we use in‐the‐wild face images and 3D annotated datasets to train the auxiliary encoder and the identity encoder, encoding the input image into parameters of FLAME (a parametric 3D face model). Simultaneously, a novel semi‐supervised hybrid landmark loss is designed to more effectively learn from in‐the‐wild face images and 3D annotated datasets. Furthermore, to meet the real‐time requirements in practical applications, a lightweight shape reconstruction model called fast‐PR3D is distilled through teacher–student learning. In texture reconstruction, we propose a texture extraction method based on face reenactment in StyleGAN2 style space, extracting texture from the source and reenacted face images to constitute a facial texture map. Extensive experiments have demonstrated the state‐of‐the‐art performance of our method.

Spectral line confocal sensor signal analysis and correction: Unlocking reflectance difference sample measurements

Spectral line confocal imaging (LCI) technology enables high-resolution, rapid 3D surface morphology analysis of transparent materials and other samples, positioning it as a leading optical measurement tool. Nevertheless, where the sensor measures samples with significant variations in surface reflectance, the peak signals of the acquired images exhibit low signal-to-noise ratios or become distorted, hampering the peak extraction algorithm from accurately locating their peak positions. Consequently, it's arduous to reconstruct their precise surface topography. To enhance the sensor's dynamic measurement range and adaptability, this paper introduces an adaptive correction approach. It utilizes a regularized non-negative matrix to decompose images with various grey levels. This is followed by weighted frequency filtering and Gaussian enhancement of all the grey layers. Eventually, these layers are fused into one image with high contrast and uniformity. Using this approach, the sensor is capable of achieving precise 3D reconstruction with high-quality peak signals. Additionally, this study experimentally demonstrates the efficacy of the suggested approach by measuring the interdigital electrode and printed circuit boards (PCB). It should be noted that the technique presented in this paper is not limited to spectral line confocal instruments and is also effective for line laser and line structured light instruments.

Deep learning-based frequency-multiplexing composite-fringe projection profilometry technique for one-shot 3D shape measurement

The pursuit of precise and efficient 3D shape measurement has long been a focal point within the fringe projection profilometry (FPP) domain. However, achieving precise 3D reconstruction for isolated objects from a single fringe image remains a challenging task in the field. In this paper, a deep learning-based frequency-multiplexing (FM) composite-fringe projection profilometry (DFCP) is proposed, where an end-to-end absolute phase retrieval network (APR-Net) is trained to directly recover the absolute phase map from a FM composite fringe pattern. The obtained absolute phase map exhibits exceptional precision and is devoid of spectrum crosstalk or leakage disturbance typically encountered in traditional FM techniques. APR-Net is intricately crafted, incorporating a nested strategy along with the concept of centralized information interaction. A diverse dataset encompassing various scenarios and free from spectrum aliasing is assembled to guide the training process. A seven-map loss calculation scheme is employed to guide the training process, of which the efficacy is proved through ablation experiments. In the first qualitative experiment, DFCP demonstrates comparable phase accuracy to ground truths with 47 fewer projected images, outperforming other three methods with mean absolute phase errors of 0.0052 rad, 0.1761 rad, 0.0169 rad, and 0.0139 rad. The second qualitative experiment and the quantitative evaluation respectively prove DFCP’s capability in high dynamic range 3D measurement and in precise 3D measurement, with sphere diameter errors of 0.0780 mm and 0.0726 mm, and a spherical centroid distance error of 0.0555 mm.

OCT-based intra-cochlear imaging and 3D reconstruction: ex vivo validation of a robotic platform.

The small size of the cochlea, and its location deeply embedded in thick temporal bone, poses a challenge for intra-cochlear guidance and diagnostics. Current radiological imaging techniques are not able to visualize the cochlear microstructures in detail. Rotational optical coherence tomography (OCT) fibers show great potential for intra-cochlear guidance. The generated images could be used to map, and study, the tiny cochlear microstructures relevant for hearing. This work describes the design of a rotational OCT probe with an outer diameter of 0.9 mm. It further discusses a robotic system, which features a remote center of motion mechanism, dedicated to the probe's positioning, fine manipulation and stable insertion into the cochlear micro-spaces. Furthermore, the necessary calibration steps for 3D reconstruction are described, followed by a detailed quantitative analysis, comparing the 3D reconstructions using a synthetic, 2:1 scaled scala tympani model with a reconstruction from micro-CT, serving as the ground truth. Finally, the potential of the system is demonstrated by scanning a single ex vivo cadaveric human cochlea. The study investigates five insertions in the same 2:1 scaled tympani model, along with their corresponding 3D reconstruction. The comparison with micro-CT results in an average root-mean-square error of 74.2 µm, a signed distance error of 38.1 µm and a standard deviation of 63.6 µm. The average F-score of the reconstructions, using a distance threshold of 100 and 74.2 µm, resulted in 83.0% and 71.8%, respectively. Insertion in the cadaveric human cochlea showed the challenges for straight insertion, i.e., navigating the hook region. Overall, the system shows great potential for intra-cochlear guidance and diagnostics, due to the system's capability for precise and stable insertion into the basal turn in the scala tympani. The system, combined with the calibration procedure, results in detailed and precise 3D reconstructions.

A smooth surface measurement method by flexible contact using multiple fingers device

Contact measurement technology is commonly used for analysis and modeling of three-dimensional (3D) object. This paper presents a contact measurement method, using a claw type probe based on a set of rotary encoders that can flexibly contact objects with efficiently multi-lines scanning. In this paper, a 3D measurement system by flexible contact of multiple fingers was built. The claw probe was set on a coordinate measurement machine. By moving the position of the claw for interpolation detection to improve the sampling rate, higher precision 3D reconstruction was achieved. Also, a compensation algorithm was proposed for this type detector to improve measuring accuracy. In the experiments, two surface models were produced by 3D printing as the test objects. By comparing the designed surface with the reconstructed surface, it is demonstrated that this system can effectively and softly measure smooth surface, like the back of human body, which is valuable for further study and use in rehabilitation massage, apparel design, and etc.

Guided by model quality: UAV path planning for complete and precise 3D reconstruction of complex buildings

Obtaining high-quality 3D models is essential for building information modelling and city management. Unmanned aerial vehicle (UAV) photogrammetry offers a cost-effective solution, but complex buildings with self-occlusion pose challenges to reconstruction quality. Conventional flight paths often result in incomplete and imprecise models. This study presents a novel multirotor UAV path planning method for achieving complete and precise 3D reconstruction of complex buildings. A coarse model is obtained from common flight paths as input, its quality factors including completeness and precision are considered. By detecting incomplete or occluded areas, additional viewpoints are strategically planned to enhance completeness while adhering to photogrammetric constraints. Furthermore, precision is improved by adding viewpoints based on error ellipsoids of target points. Real-world experiments demonstrate that the proposed method increases incomplete area coverage by up to 21% and enables centimetre-level accuracy reconstruction with reduced uncertainties. The method holds promise for georeferenced inspection and documentation of complex buildings.

Exploring the feasibility of smartphone cameras for 3D modelling of bite patterns in forensic dental identification

The field of bitemark analysis involves examining physical alterations in a medium resulting from contact with teeth and other oral structures. Various techniques, such as 2D and 3D imaging, have been developed in recent decades to ensure precise analysis of bitemarks. This study assessed the precision of using a smartphone camera to generate 3D models of bitemark patterns. A 3D model of the bite mark pattern was created using 3Shape TRIOSTM and a smartphone camera combined with monoscopic photogrammetry. The mesiodistal dimensions of the anterior teeth were measured using Rapidform Explorer and OrtogOnBlender, and the collected data were analyzed using IBM® SPSS® Statistics version 23.0. The mean mesiodistal dimension of the anterior teeth, as measured on the 3D model from 3Shape TRIOSTM and smartphone cameras, was found to be 6.95 ± 0.7667 mm and 6.94 ± 0.7639 mm, respectively. Statistical analysis revealed no significant difference between the two measurement methods, p > 0.05. The outcomes derived from this study unequivocally illustrate that a smartphone camera possessing the specific parameters detailed in this study can create a 3D representation of bite patterns with an accuracy level on par with the outputs of a 3D intraoral camera. These findings underscore the promising trajectory of merging smartphone cameras and monoscopic photogrammetry techniques, positioning them as a budget-friendly avenue for 3D bitemark analysis. Notably, the monoscopic photogrammetry methodology assumes substantial significance within forensic odontology due to its capacity for precise 3D reconstructions and the preservation of critical measurement data.

Nanoscale Examination of Artistic Surfaces Utilizing 3D Non-Contact Optical Scanning Technology

The assessment of painting surfaces at the microscale has been historically impeded by challenges related to limited resolution and accuracy in traditional methodologies. This study pioneers the utilization of non-contact 3D optical scanning technology, meticulously calibrated for nanoscale precision, to unravel the intricate features residing on painting surfaces. The initial phase employs the Point Diffraction Interferometer (PDI) for 3D optical scanning, incorporating meticulously optimized parameters tailored to nanoscale analysis. Subsequent phases involve the application of Phase Shifting Interferometry (PSI) and Holographic Interferometry (HI). PSI is employed to discern morphological alterations, while HI captures the nuanced color and optical characteristics embedded in the painting surfaces. To enhance the continuity of phase information, the Goldstein algorithm is introduced during phase stitching, fortifying the method’s robustness against discontinuities. Further refinement is achieved through the Iterative Closest Point (ICP) algorithm, orchestrating precise 3D data reconstruction. This process encompasses multi-view stereo matching and surface fitting, ensuring a meticulous representation of the painting surface geometry. The study meticulously presents detailed 3D optical scanning results, probing into the painting surface’s performance concerning nanoscale resolution, measurement accuracy, and color consistency. The unveiled findings showcase a remarkable minimum feature capture capability of 1.8 at nanoscale resolution. The quantitative assessment, encapsulated by a Root Mean Square Error (RMSE) ranging from 0.001 to 0.012 for 100 scanned data points, and a Standard Deviation (STD) oscillating between 0.0008 to 0.0018, attests to the method’s efficacy. This effectiveness is underscored by its capacity to deliver a thorough and intricate analysis of painting surface performance at the nanoscale.

High-Resolution Imaging and Morphological Phenotyping of C. elegans through Stable Robotic Sample Rotation and Artificial Intelligence-Based 3-Dimensional Reconstruction.

Accurate visualization and 3-dimensional (3D) morphological profiling of small model organisms can provide quantitative phenotypes benefiting genetic analysis and modeling of human diseases in tractable organisms. However, in the highly studied nematode Caenorhabditis elegans, accurate morphological phenotyping remains challenging because of notable decrease in image resolution of distant signal under high magnification and complexity in the 3D reconstruction of microscale samples with irregular shapes. Here, we develop a robust robotic system that enables the contactless, stable, and uniform rotation of C. elegans for multi-view fluorescent imaging and 3D morphological phenotyping via the precise reconstruction of 3D models. Contactless animal rotation accommodates a variety of body shapes and sizes found at different developmental stages and in mutant strains. Through controlled rotation, high-resolution fluorescent imaging of C. elegans structures is obtained by overcoming the limitations inherent in both widefield and confocal microscopy. Combining our robotic system with machine learning, we create, for the first time, precise 3D reconstructions of C. elegans at the embryonic and adult stages, enabling 3D morphological phenotyping of mutant strains in an accurate and comprehensive fashion. Intriguingly, our morphological phenotyping discovered a genetic interaction between 2 RNA binding proteins (UNC-75/CELF and MBL-1/MBNL), which are highly conserved between C. elegans and humans and implicated in neurological and muscular disorders. Our system can thus generate quantitative morphological readouts facilitating the investigation of genetic variations and disease mechanisms. More broadly, our method will also be amenable for 3D phenotypic analysis of other biological samples, like zebrafish and Drosophila larvae.

Virtual Styling and Fitting using AI

Abstract: The digitization of human forms holds significant relevance across domains like virtual reality, medical imaging, and robot navigation. While sophisticated multi-view systems excel at precise 3D body reconstruction, they remain largely inaccessible to the general consumer market. Seeking more accessible approaches, methods in human digitization explore simpler inputs, such as single images. Among these approaches, pixel-aligned implicit models [1] have garnered attention for their efficiency in capturing intricate geometric details like clothing wrinkles. These models, notably lightweight, employ parametric human body models without necessitating mappings between canonical and posed spaces. PIFu [1], a prime example, translates a parametric body model into pixel-aligned 2D feature maps, offering more comprehensive information than mere (x, y, z) coordinates. This research paper, rooted in PIFu [1], delves into the implementation intricacies of 3D human digitization within a specific domain – 3D digitization for virtual styling and fitting. In today's tech-driven world, online shopping has boomed, offering unparalleled convenience by letting users swiftly browse and buy items from home. However, despite its speed and ease, trying on clothes remains a hurdle. Without a physical store, customers struggle to gauge fit and size, leading to increased returns and order abandonment. To tackle this, a project aims to revolutionize the online clothing shopping experience. By offering a solution that allows virtual try-ons, users can visualize how clothes look and fit before buying, potentially reducing return rates and enhancing the overall shopping journey.

MODEL-BASED MULTI-UAV PATH PLANNING FOR HIGH-QUALITY 3D RECONSTRUCTION OF BUILDINGS

Abstract. Unmanned aerial vehicle (UAV) photogrammetry is widely used for acquiring high-quality 3D models of urban areas. However, the completeness and quality of the reconstructed model can be affected by complex or concave structures when using common flight paths. Furthermore, flight paths with multi-altitude and multi-attitude waypoints can be time-consuming and may not be efficiently implemented by a single drone. To address these challenges, we propose a model-based multi-UAV path planning method for capturing images that enable complete and precise 3D reconstruction of buildings. Our method analyzes the geometry of the input coarse model and plans viewpoints based on the reconstructability related to completeness and precision. By solving a multiple travelling salesmen problem (mTSP), individual flight paths for each UAV are generated. We conducted a real-world experiment comparing the performance and efficiency of the proposed method with two existing solutions. Results the proposed method produces a more complete 3D reconstruction with fewer images compared to the existing methods. It also shows that using two UAVs with the proposed method can significantly reduce the overall time required for 3D reconstruction.

Binocular Structured Light 3-D Reconstruction System for Low-Light Underwater Environments: Design, Modeling, and Laser-Based Calibration

The three dimensional (3D) perception of the low-light underwater environment has always been a major challenge which greatly limits the underwater operations. In this paper, an underwater active vision measurement system based on binocular structured light is designed to achieve high precision 3D reconstruction. Firstly, the fusion technology of binocular camera and laser addresses underwater optical attenuation and feature sparsity. Then, in order to avoid the huge inertia caused by the overall movement of the system, a laser scanner based on the mirror-galvanometer is used to accomplish static scanning of the scene. Subsequently, considering the influence of multiple media, underwater refraction models including monocular imaging model, binocular ranging model, and binocular polar curve constraint model are systematically proposed. What’s more, the conventional checkerboard-based passive visual calibration method is ineffective for low-light waters. Therefore, a simple calibration block is designed, and a new multi-objective laser-based calibration algorithm based on laser geometric constraints is proposed. Finally, the effectiveness of the system is verified by analyzing the 3D reconstruction results of underwater objects.

Okapi-EM: A napari plugin for processing and analyzing cryogenic serial focused ion beam/scanning electron microscopy images.

An emergent volume electron microscopy technique called cryogenic serial plasma focused ion beam milling scanning electron microscopy (pFIB/SEM) can decipher complex biological structures by building a three-dimensional picture of biological samples at mesoscale resolution. This is achieved by collecting consecutive SEM images after successive rounds of FIB milling that expose a new surface after each milling step. Due to instrumental limitations, some image processing is necessary before 3D visualization and analysis of the data is possible. SEM images are affected by noise, drift, and charging effects, that can make precise 3D reconstruction of biological features difficult. This article presents Okapi-EM, an open-source napari plugin developed to process and analyze cryogenic serial pFIB/SEM images. Okapi-EM enables automated image registration of slices, evaluation of image quality metrics specific to pFIB-SEM imaging, and mitigation of charging artifacts. Implementation of Okapi-EM within the napari framework ensures that the tools are both user- and developer-friendly, through provision of a graphical user interface and access to Python programming.

Combining Photogrammetry and Photometric Stereo to Achieve Precise and Complete 3D Reconstruction.

Image-based 3D reconstruction has been employed in industrial metrology for micro-measurements and quality control purposes. However, generating a highly-detailed and reliable 3D reconstruction of non-collaborative surfaces is still an open issue. In this paper, a method for generating an accurate 3D reconstruction of non-collaborative surfaces through a combination of photogrammetry and photometric stereo is presented. On one side, the geometric information derived with photogrammetry is used in areas where its 3D measurements are reliable. On the other hand, the high spatial resolution capability of photometric stereo is exploited to acquire a finely detailed topography of the surface. Finally, three different approaches are proposed to fuse both geometric information and high frequency details. The proposed method is tested on six different non-collaborative objects with different surface characteristics. To evaluate the accuracy of the proposed method, a comprehensive cloud-to-cloud comparison between reference data and 3D points derived from the proposed fusion methods is provided. The experiments demonstrated that, despite correcting global deformation up to an average RMSE of less than 0.1 mm, the proposed method recovers the surface topography at the same high resolution as the photometric stereo.

Fringe projection profilometry method with high efficiency, precision, and convenience: theoretical analysis and development.

Fringe projector profilometry (FPP) is an important three-dimensional (3D) measurement technique, especially when high precision and speed are required. Thus, theoretical interrogation is critical to provide deep understanding and possible improvement of FPP. By dividing an FPP measurement process into four steps (system calibration, phase measurement, pixel correspondence, and 3D reconstruction), we give theoretical analysis on the entire process except for the extensively studied calibration step. Our study indeed reveals a series of important system properties, to the best of our knowledge, for the first time: (i) in phase measurement, the optimal and worst fringe angles are proven perpendicular and parallel to epipolar line, respectively, and can be considered as system parameters and can be directly made available during traditional calibration, highlighting the significance of the epipolar line; (ii) in correspondence, when two sets of fringes with different fringe orientations are projected, the highest correspondence precision can be achieved with arbitrary orientations as long as these two orientations are perpendicular to each other; (iii) in reconstruction, a higher reconstruction precision is given by the 4-equation methods, while we notice that the 3-equation methods are almost dominatingly used in literature. Based on these theoretical results, we propose a novel FPP measurement method which (i) only projects one set of fringes with optimal fringe angle to explicitly work together with the epipolar line for precise pixel correspondence; (ii) for the first time, the optimal fringe angle is determined directly from the calibration parameters, instead of being measured; (iii) uses 4 equations for precise 3D reconstruction but we can remove one equation which is equivalent to an epipolar line, making it the first algorithm that can use 3-equation solution to achieve 4-equation precision. Our method is efficient (only one set of fringe patterns is required in projection and the speed is doubled in reconstruction), precise (in both pixel correspondence and 3D reconstruction), and convenient (the computable optimal fringe angle and a closed-form 3-equation solution). We also believe that our work is insightful in revealing fundamental FPP properties, provides a more reasonable measurement for practice, and thus is beneficial to further FPP studies.

Three-Dimensional Reconstruction of Soybean Canopy Based on Multivision Technology for Calculation of Phenotypic Traits

Precise reconstruction of the morphological structure of the soybean canopy and acquisition of plant traits have great theoretical significance and practical value for soybean variety selection, scientific cultivation, and fine management. Since it is difficult to obtain all-around information on living plants with traditional single or binocular machine vision, this paper proposes a three-dimensional (3D) method of reconstructing the soybean canopy for calculation of phenotypic traits based on multivision. First, a multivision acquisition system based on the Kinect sensor was constructed to obtain all-around point cloud data of soybean in three viewpoints, with different fertility stages of soybean as the research object. Second, conditional filtering and K-nearest neighbor filtering (KNN) algorithms were used to preprocess the raw 3D point cloud. The point clouds were matched and fused by the random sample consensus (RANSAC) and iterative closest point (ICP) algorithms to accomplish the 3D reconstruction of the soybean canopy. Finally, the plant height, leafstalk angle and crown width of soybean were calculated based on the 3D reconstruction of soybean canopy. The experimental results showed that the average deviations of the method was 2.84 cm, 4.0866° and 0.0213 m, respectively. The determination coefficients between the calculated values and measured values were 0.984, 0.9195 and 0.9235. The average deviation of the RANSAC + ICP was 0.0323, which was 0.0214 lower thanthe value calculated by the ICP algorithm. The results enable the precise 3D reconstruction of living soybean plants and quantitative detection for phenotypic traits.