Matching Points Research Articles

Efficient object recognition under cluttered scenes via descriptor-based matching and single point voting

This paper addresses the problem of recognizing multiple objects and multiple instances from point clouds. Whereas existing methods utilize descriptors on 3D fields or pointwise voting to achieve this task, our framework takes advantage of both descriptor-based and voting-based schemes to realize more robust and efficient prediction. Specifically, we propose a novel and robust descriptor called an orientation-enhanced fast point feature histogram (OE-FPFH) to describe points in both the object model and scene, and further to build the correspondence set. The OE-FPFH integrates an orientation vector through mining the geometric tensor of the local structure of a surface point, which is more representative than the original FPFH descriptor. To improve voting efficiency, we devise a novel single-point voting mechanism (SPVM), which constructs a unique local reference frame (LRF) on a single point using the orientation vector. The SPVM takes as input the corresponding point set and can generate a pose candidate for each correspondence. The process is realized by matching LRFs from two corresponding points. All pose candidates are subsequently divided into clusters and aggregated using the K-means clustering algorithm to deduce the poses for different objects or instances in the scene. Experiments on three challenging datasets demonstrate that our method is effective, efficient, and robust to occlusions and multiple instances.

Vibration Position Detection of Robot Arm Based on Feature Extraction of 3D Lidar.

In the process of construction, pouring and vibrating concrete on existing reinforced structures is a necessary process. This paper presents an automatic vibration position detecting method based on the feature extraction of 3D lidar point clouds. Compared with the image-based method, this method has better anti-interference performance to light with reduced computational consumption. First, lidar scans are used to capture multiple frames of local steel bar point clouds. Then, the clouds are stitched by Normal Distribution Transform (NDT) for preliminary matching and Iterative Closest Point (ICP) for fine-matching. The Graph-Based Optimization (g2o) method further refines the precision of the 3D registration. Afterwards, the 3D point clouds are projected into a 2D image. Finally, the locations of concrete vibration points and concrete casting points are discerned through point cloud and image processing technologies. Experiments demonstrate that the proposed automatic method outperforms ICP and NDT algorithms, reducing the mean square error (MSE) by 11.5% and 11.37%, respectively. The maximum discrepancies in identifying concrete vibration points and concrete casting points are 0.059 ± 0.031 m and 0.089 ± 0.0493 m, respectively, fulfilling the requirement for concrete vibration detection.

Self-assembly of amphiphilic homopolymers grafted onto spherical nanoparticles: complete embedded minimal surfaces and a machine learning algorithm for their recognition.

By means of computer modelling, the self-assembly of amphiphilic A-graft-B macromolecules, grafted onto a spherical nanoparticle, is studied. In a solvent, that is poor for side pendants, the macromolecules self-assemble into thin membrane-like ABBA bilayers deviated from spherical nanoparticles. The bilayers form morphological structures that depend on the grafting density and macromolecular polymerization degree and can be referred to as the classical family of complete embedded minimal surfaces. The plane disk, catenoid, helicoid, Costa and Enneper surfaces as well as "double" helicoid and "complex minimal surface" were identified, and the fields of their stability were defined. The surfaces can be grouped according to the sequences of conformal transformations that transform them into each other. These surfaces arise in different experiments situationally. Results are summarized in a pie diagram constructed using a machine learning algorithm based on matching grafting points with a specially created planar graphic image.

Multi-line laser 3D reconstruction method based on spatial quadric surface and geometric estimation

The traditional multi-line laser 3D reconstruction is based on binocular epipolar constraints and the equation of the laser optical plane. Firstly, matching points are identified based on the epipolar constraints. Then, the correct matching points are selected based on the optical plane equation of the multi-line laser. Finally, 3D reconstruction is performed using the selected matching points. In actual production processes involving multi-line lasers, noticeable distortion occurs due to the projection of Diffractive Optical Elements (DOE) at a large angle. This results in curved laser light planes and curved reflections on the measured objects. Under such circumstances, efficiently screening matching points using the traditional method based on the laser plane equation becomes challenging. Additionally, inherent noise in multi-line laser systems introduces errors in extracted laser center coordinates, making it impossible to directly obtain high-precision 3D data. To address these issues, this paper proposes a method that utilizes spatial quadric surfaces and geometric estimation for completing multi-line laser 3D reconstructions. By analyzing the distortion principle of DOE and the positional relationship after optical diffraction, the multi-line laser manifests as a quadric surface on the optical output plane. Consequently, by calibrating the equations of the quadric surface and applying binocular polar constraints, suitable matching points for the multi-line laser can be chosen. Once an accurate matching point is identified, a minimum geometric distance estimation can be established based on the distance constraint between the point and its corresponding epipolar line. This distance represents the separation between the left and right camera’s laser center points and their respective epipolar lines. Utilizing this estimate of the minimum geometric distance, a refined calculation can be performed to better satisfy epipolar constraints and obtain new matching points. Ultimately, utilizing these new matching points enables the completion of 3D reconstruction for multi-line lasers. In comparison with methods relying on spatial plane and epipolar constraints, our algorithm exhibits a superior degree of matchability and accuracy.According to multiple groups of test data, the average error before optimization is 0.3126 mm, and can be increased to 0.0336 mm after optimization.

Preparation of novel and efficient arylphosphonate flame retardants for simultaneously enhancement of fire safety and UV-shielding properties of transparent thermoplastic polyurethane

Thermoplastic polyurethane (TPU) is highly flammable and UV aging, limiting its application in the field of new energy and electronic device. In this work, a novel macromolecular aromatic phosphonate Bis(4-(((diphenylphosphinyl)oxy)methyl)phenyl)phenylphosphonate (DMP) containing both P-C and P=O was successfully prepared by nucleophilic substitution reaction. The DMP was introduced into the TPU matrix to prepare multifunctional TPU composites. Condensed and gas phase analysis have shown that the flame retardant mechanism of DMP was mainly influenced by flame retardant inhibition and the barrier effect of char layer. Only 5 wt% DMP enabled TPU to obtain the UL-94V-0 rating with the LOI of 27.6 %. Compared with pure TPU, the HRR and THR of TPU/5 wt% DMP were decreased by 19.8 % and 13.2 %. Meanwhile, the matching of the melting point for DMP and the processing temperature for TPU increased its dispersion in TPU matrix. Due to the excellent compatibility between DMP and TPU, TPU/DMP composites almost retained their original transparency and ductility. In addition, TPU/DMP composites exhibited excellent UV resistance, obtaining 100 % UV shielding at UV-B wavelengths and up to 97 % UV isolation at UV-A wavelengths. The study introduced a novel approach for preparing multifunctional flame retardant additives, and opened up wide application prospects for high-performance flame retardant TPU composites in emerging fields such as rail transit and electronic packaging.

High-accuracy fringe projection profilometry without phase unwrapping based on multi-view geometry constraints

The number of fringes and phase unwrapping in fringe projection profilometry result in two key factors. The first is to avoid the problems of excessive fringe patterns, and the second is phase ambiguity. This paper presents a three-dimensional (3D) measurement method without phase unwrapping. This method benefits from the geometric constraints and does not require additional images. Meanwhile, epipolar rectification is performed to calibrate the rotation matrix relationship between the new plane of the dual camera and the plane of the projector. Subsequently, using depth constraints, the point pairs with incorrect 3D positions are effectively eliminated, and the initial parallax map is obtained by establishing epipolar lines of the left and right matching points in the projector domain, obtaining the intersection points, and setting up the threshold for filtering. Finally, a function combining the modulation intensity and phase is proposed to refine the parallax map such that the 3D result is insensitive to phase error. The standard step block and standard ball were used to verify the validity of the proposed method, and the experimental results showed that the root mean square error of the method was 0.052 mm.

High-Precision Heterogeneous Satellite Image Manipulation Localization: Feature Point Rules and Semantic Similarity Measurement

Misusing image tampering software makes it easier to manipulate satellite images, leading to a crisis of trust and security concerns in society. This study compares the inconsistencies between heterogeneous images to locate tampered areas and proposes a high-precision heterogeneous satellite image manipulation localization (HSIML) framework to distinguish tampered from real landcover changes, such as artificial constructions, and pseudo-changes, such as seasonal variations. The model operates at the patch level and comprises three modules: The heterogeneous image preprocessing module aligns heterogeneous images and filters noisy data. The feature point constraint module mitigates the effects of lighting and seasonal variations in the images by performing feature point matching, applying filtering rules to conduct an initial screening to identify candidate tampered patches. The semantic similarity measurement module designs a classification network to assess RS image feature saliency. It determines image consistency based on the similarity of semantic features and implements IML using predefined classification rules. Additionally, a dataset for IML is constructed based on satellite images. Extensive experiments compared with existing SOTA models demonstrate that our method achieved the highest F1 score in both localization accuracy and robustness tests and demonstrates the capability for handling large-scale areas.

PPDM++: Parallel Point Detection and Matching for Fast and Accurate HOI Detection.

Human-Object Interaction (HOI) detection aims to understand human activities by detecting interaction triplets. Previous HOI detection methods adopt a two-stage instance-driven paradigm. Unfortunately, many non-interactive human-object pairs generated by the first stage are the main obstacle impeding HOI detectors from high efficiency and promising performance. To remedy this, we propose a novel top-down interaction-driven paradigm, detecting interactions first and bridging interactive human-object pairs through interactions. We formulate HOI as a point triplet human point, interaction point, object point and design a Parallel Point Detection and Matching (PPDM) framework. We further take advantage of two-stage methods and propose a novel framework, PPDM++, that detects the interactive human-object pairs by PPDM, then extracts region features for each pair to predict actions. The core of PPDM/PPDM++ is to convert the instance-driven bottom-up paradigm to an interaction-driven top-down paradigm, thus avoiding additional computation costs from traversing a tremendous number of non-interactive pairs. Benefiting from the advanced paradigm, PPDM/PPDM++ has achieved significant performance gains with high efficiency. PPDM-DLA-34 has achieved 19.94 mAP with 42 FPS as the first real-time HOI detector, and PPDM++-SwinB achieves 30.1 mAP with 17 FPS on HICO-DET dataset. We also built an application-oriented database named HOI-A, a supplement to the existing datasets.

Deformation tracking of honeycomb structure based on image skeletonization and branch point matching

Honeycomb structures have attracted much attention in various engineering fields due to its superiorities in high specific strength, high specific stiffness and excellent energy-absorbing characteristics. Therefore, it is very important to obtain the deformation state of honeycomb structure in the studies of its manufacturing process and mechanical behavior. In this study, a simple and efficient strategy for tracking the deformation of thin-walled honeycomb structure based on image skeletonization and branch points matching is presented. Principle and process of the new proposed strategy are first detailed, including image skeletonization, branch points selection, matching expansion and deformation calculation, etc. Simulations and experiments with compression and tensile deformations are performed to verify the efficiency of the proposed strategy. The results indicate that the displacement measurements based on the proposed strategy are able to provide subpixel-level accuracy, even though some interference branch points are generated during deformation. In addition, the limitations of the proposed strategy are discussed, which points out the train of thought for the subsequent research.

Deformation measurement of aero-engine rotating twisted blade based on multi-vision overlapping area feature registration method

Due to the complexity of the spatial surface of the twisted blade of the aero-engine, there are occlusions and shadows between the blades, resulting in visual blind spots. This causes the loss of certain information in the deformation measurement of twisted blade. A deformation measurement method of rotating twisted blades based on multi-vision overlapping area feature registration is proposed. It aims to solve the problem of limited field of view (FOV) in the deformation measurement of twisted blade, and overcome the problems of small measurement range and difficult measurement of traditional contact sensor method in the rotating state of the blade. First, a calibration and image-matching method based on multi-vision digital image correlation method is proposed. Under the premise of ensuring the simultaneous calibration of multi-vision camera stereo calibration, the matching of image detection points on the twisted blade surface is realized. Second, a three-dimensional reconstruction method based on the feature registration of the overlapping area of the FOV is proposed. It is used to solve the problem that the traditional image stitching method is prone to mismatching and to provide complete point cloud information for the deformation field calculation. Finally, the strain of each point is solved by piecewise point-by-point fitting of the local subdomains of the discrete displacement data, which suppresses the amplification of noise and improves the accuracy of blade strain calculation. In this paper, the deformation measurement experiment is carried out for the rotating twisted blade (up to 6000 rpm). The experimental results show that the proposed method can achieve a displacement measurement accuracy of 5 μm and the strain measurement accuracy of 50 με. This method can provide an important guarantee for the accurate evaluation and safe operation of the key performance parameters of the engine.

An improved SLAM algorithm for substation inspection robot based on the fusion of IMU and visual information

In the past, manual inspection was often used for equipment inspection in indoor environments such as substation rooms and chemical plant rooms. This way often accompanies high labor intensity, low inspection efficiency, and low safety, which is difficult to meet the increasingly stringent requirements of indoor equipment operation and maintenance management. For dealing with these issues, a VIORB-SLAM2 algorithm based on the integration of IMU and visual information, was proposed by this paper. Firstly, the IMU data and image data were integrated to restore scale information of cameras, and then an error function was established to enhance the algorithm’s robustness. Secondly, in order to improve the accuracy of the algorithm, the random sampling consensus method was used to eliminate the wrong matching points in feature point matching, and the normalized cross-correlation matching was employed to constrain key frame matching conditions. Finally, through the iterative closest point method to stitch the point clouds, a dense map for navigation was constructed. The experimental results show that the algorithm designed by this paper has solved the shortcomings of applying the ORB-SLAM2 algorithm to indoor inspection robots while achieving high positioning accuracy, which can be combined with other algorithms in the field of artificial intelligence for object detection and semantic map construction in the future.

An enhanced efficiency 170–260 GHz frequency doubler based on three points resonance matching technique

This paper presents a full-band frequency doubler (170–260 GHz) with enhanced efficiency based on Schottky diodes. To achieve broadband matching between the matching circuit and the Schottky diode, a three points resonance matching technique (TPRMT) is employed. This technique is developed by considering the embedded impedance matching at three distinct frequency points: upper, middle, and lower, thereby guiding the design of the peripheral matching circuits. In comparison to single frequency point matching, this approach effectively suppresses sideband performance degradation, resulting in improved efficiency. Additionally, the peripheral matching circuits are meticulously designed to fulfill a resonance clockwise impedance trajectory, facilitating matching at the aforementioned three frequency points. Due to the limitation of the input power source, measurements were conducted only within the 170–237 GHz frequency band. The measured results show that the doubler achieves an efficiency ranging from 8.4 % to 14.1 % under an input power of 130–175 mW. The good consistency between simulation and measurement validates the significant advantages of TPRMT.

High-accuracy 3D reconstruction of step edge with ray aliasing based on projective parallel single-pixel imaging

Optical 3D measurement technology especially fringe projection profilometry (FPP) is widely utilized in industrial production. However, invalid mixed phases are common when measuring step edges in industrial parts which leads to outliers in reconstructed point cloud and ultimately causes reduction of valid point cloud and dimensional measurement errors. This paper first analyzes reasons for failure of step edge measurement in FPP and illustrates feasibility of parallel single-pixel imaging (PSI) applied to eliminate ray aliasing of step edge. Further for a good balance of accuracy and efficiency, we propose an optimized projective parallel single-pixel imaging (pPSI) method combing with edge detection in wrapped phase generated by pPSI patterns and adaptive bi-direction projection strategy to accurately identify direct correspondence matching points of step edge. Experimental results verified that the proposed method can achieve high accuracy of 3D reconstruction on step edge while taking less measurement time.

A Hybrid Approach for Image Acquisition Methods Based on Feature-Based Image Registration.

This paper presents a novel hybrid approach to feature detection designed specifically for enhancing Feature-Based Image Registration (FBIR). Through an extensive evaluation involving state-of-the-art feature detectors such as BRISK, FAST, ORB, Harris, MinEigen, and MSER, the proposed hybrid detector demonstrates superior performance in terms of keypoint detection accuracy and computational efficiency. Three image acquisition methods (i.e., rotation, scene-to-model, and scaling transformations) are considered in the comparison. Applied across a diverse set of remote-sensing images, the proposed hybrid approach has shown marked improvements in match points and match rates, proving its effectiveness in handling varied and complex imaging conditions typical in satellite and aerial imagery. The experimental results have consistently indicated that the hybrid detector outperforms conventional methods, establishing it as a valuable tool for advanced image registration tasks.

Match Point: Nuclear Medicine Imaging for Recurrent Thyroid Cancer in TENIS Syndrome-Systematic Review and Meta-Analysis.

Background/Objectives: Disease recurrence and resistance to radioiodine (RAI) therapy are major challenges in the management of differentiated thyroid cancer (DTC). In particular, the TENIS (Thyroglobulin Elevated Negative Iodine Scintigraphy) syndrome, characterised by elevated thyroglobulin (Tg) serum levels in addition to a negative radioiodine whole body scan (WBS), complicates disease monitoring and treatment decisions. Conventional imaging techniques often fail to detect disease in WBS-negative patients with rising Tg levels, leading to limitations in therapeutic intervention. This systematic review and meta-analysis aims to evaluate the diagnostic accuracy of nuclear imaging modalities in detecting disease recurrence in patients with the TENIS syndrome and to provide insights to guide therapeutic approaches in this complex clinical scenario. Methods: A comprehensive search of PubMed/MEDLINE and EMBASE databases up to March 2024 was performed according to PRISMA guidelines. Eligible studies were selected, and quality assessment was performed with the QUADAS-2 tool. For each study, relevant data were extracted and synthesised. A meta-analysis of the diagnostic accuracy of [18F]FDG PET/CT was performed, and patient-based pooled sensitivity and specificity were calculated using a random-effects model. Statistical heterogeneity between studies was assessed using the I2 statistic. Results: Of the 538 studies initially identified, 22 were included in the systematic review, of which 18 were eligible for meta-analysis. The eligible studies, mainly focused on [18F]FDG PET/CT, showed variable sensitivity and specificity for the detection of RAI-refractory thyroid cancer lesions. For [18F]FDG PET/CT, pooled estimates displayed a sensitivity of 0.87 (95% CI: 0.82-0.90) and a specificity of 0.76 (95% CI: 0.61-0.86), with moderate heterogeneity between studies. Conclusions: [18F]FDG PET/CT remains central in the detection of disease recurrence in patients with the TENIS syndrome. The emergence of novel radiopharmaceuticals with specific molecular targets is a promising way to overcome the limitations of [18F]FDG in these patients and to open new theranostics perspectives. This review highlights the great potential of nuclear medicine in guiding therapeutic strategies for RAI-refractory thyroid cancer.

Digital image correlation calculation method for RGB-D camera multi-view matching using variable template

Multi-view matching is an indispensable method for digital image correlation 3D reconstruction. However, as the camera viewing angle changes, traditional DIC methods can lead to the accumulation of matching errors and an increase in mismatches, thereby affecting the 3D reconstruction quality. This paper proposes a digital image correlation method for multi-view RGB-D images based on variable template matching. Initial positioning is achieved by dynamically adjusting the size and sampling direction of the variable template. Sub-pixel registration is then performed using the inverse compositional algorithm, followed by multi-view point cloud fusion using the Iterative Closest Point (ICP) registration method. Experimental results show that the iteration speed of point cloud registration is increased by 30%, and the cumulative error is reduced by 62.5%. These improvements demonstrate that the proposed method is suitable for multi-viewpoint matching scenarios, achieving excellent results in multi-view feature point matching and multi-view point cloud fusion, significantly enhancing 3D reconstruction accuracy and efficiency.

Automatic personal identification using a single CT image.

Computer vision (CV) mimics human vision, enabling computers to automatically compare radiological images from recent examinations with a large image database for unique identification, crucial in emergency scenarios involving unknown patients or deceased individuals. This study aims to extend a CV-based personal identification method from orthopantomograms (OPGs) to computed tomography (CT) examinations using single CT slices. The study analyzed 819 cranial computed tomography (CCT) examinations from 722 individuals, focusing on single CT slices from six anatomical regions to explore their potential for CV-based personal identification in 69 procedures. CV automatically identifies and describes interesting features in images, which can be recognized in a reference image and then designated as matching points. In this study, the number of matching points was used as an indicator for identification. Across six different regions, identification rates ranged from 41/69 (59%) to 69/69 (100%) across over 700 possible identities. Comparison of images from the same individual achieved higher matching points, averaging 6.32 ± 0.52% (100% represents the maximum possible matching points), while images of different individuals averaged 0.94 ± 0.15%. Reliable matching points are found in the teeth, maxilla, cervical spine, skull bones, and paranasal sinuses, with the maxillary sinuses and ethmoidal cells being particularly suitable for identification due to their abundant matching points. Unambiguous identification of individuals based on a single CT slice is achievable, with maxillary sinus CT slices showing the highest identification rates. However, metal artifacts, especially from dental prosthetics, and various head positions can hinder identification. Radiology possesses a multitude of reference images for a CV database, facilitating automated CV-based personal identification in emergency examinations or cases involving unknown deceased individuals. This enhances patient care and communication with relatives by granting access to medical history. Unknown individuals in radiology or forensics pose challenges, addressed through automatic CV-based identification methods. A single CT slice highlighting the maxillary sinuses is particularly effective for personal identification. Radiology plays a pivotal role in automated personal identification by leveraging its extensive image database.

Highly robust thermal infrared and visible image registration with canny and phase congruence detection

When a dual-optical vehicle-mounted camera is photographed, visible and thermal infrared (VI-T) images cannot be registered, and realising fine matching of the images to solve for the optimal affine parameters is thus key to this research. Visible and thermal image matching is of great interest owing to the presence of challenging scale transformations, nonlinear radial aberrations, and robustness under illumination conditions. Thus, this paper presents a novel algorithm, namely, Highly Robust Thermal Infrared and Visible Image Registration with Canny and Phase Congruence Detection (HRCP). Firstly, the image is pre-processed, and feature point detection is performed by combining adaptive thresholding GW-Canny edge detection and phase coherence (PC), which increases the edge information. Secondly, SURF and FAST feature detection at the maximum distance of PC maps can obtain stable and reliable feature points, which are robust to scale invariance and rotation invariance. Finally, the optimal affine parameter is solved using the least square method with several iterations. Compared to existing matching methods, we find that VI-T image matching is particularly sensitive to both lighting factors and scale-transformed nonlinear radial distortion and that HRCP is robust in performing matching. For the randomly selected dataset, the average image matching time is reduced by 3.265 s, the number of correct image matches increases by 44 pairs on average, the root mean square error improves to 1.830, and the average error improves to 2.815. The proposed method not only improves the inaccuracy of radiation-insensitive feature transform matching in both low and high light, but also achieves a higher accuracy and an increase in the number of feature matching points, which facilitates robust VI-T image registration.

NMC3D: Non-Overlapping Multi-Camera Calibration Based on Sparse 3D Map.

With the advancement of computer vision and sensor technologies, many multi-camera systems are being developed for the control, planning, and other functionalities of unmanned systems or robots. The calibration of multi-camera systems determines the accuracy of their operation. However, calibration of multi-camera systems without overlapping parts is inaccurate. Furthermore, the potential of feature matching points and their spatial extent in calculating the extrinsic parameters of multi-camera systems has not yet been fully realized. To this end, we propose a multi-camera calibration algorithm to solve the problem of the high-precision calibration of multi-camera systems without overlapping parts. The calibration of multi-camera systems is simplified to the problem of solving the transformation relationship of extrinsic parameters using a map constructed by multiple cameras. Firstly, the calibration environment map is constructed by running the SLAM algorithm separately for each camera in the multi-camera system in closed-loop motion. Secondly, uniformly distributed matching points are selected among the similar feature points between the maps. Then, these matching points are used to solve the transformation relationship between the multi-camera external parameters. Finally, the reprojection error is minimized to optimize the extrinsic parameter transformation relationship. We conduct comprehensive experiments in multiple scenarios and provide results of the extrinsic parameters for multiple cameras. The results demonstrate that the proposed method accurately calibrates the extrinsic parameters for multiple cameras, even under conditions where the main camera and auxiliary cameras rotate 180°.

RTMFusion: An enhanced dual-stream architecture algorithm fusing RGB and depth features for instance segmentation of tomato organs

Accurate and rapid acquisition of semantically annotated point clouds from tomato cultivation regions is crucial for autonomous tomato harvesting robots. This entails the robot’s ability to perceive tomato point clouds from various locations and identify the stems and pedicels connecting the fruits. However, amidst background interference from similar stem and pedicel features, the heterogeneous growth postures of tomato clusters and the slender and short nature of pedicels pose challenges in semantic point cloud segmentation and matching of different tomato fruits, pedicels, and stems. Leveraging the salient feature representation of depth images in fine contours and distance dimensions, this paper introduces RTMFusion, a lightweight dual-stream backbone instance segmentation model that integrates color images and depth images. A Depth Enhanced Feature Fusion module is designed for multi-modal feature fusion. Fruit-bearing organ matching is facilitated by a method based on distance constraints and conditional search. Key characteristics are extracted from semantic point clouds, and an optimal condition search is conducted on candidates meeting constraint conditions. On the test dataset, our model improved the mask mAP by 4.6 % compared to the baseline, achieving 72.1 % with a speed of 17.1 ms. Our matching method achieved 99.2 % pedicel matching accuracy and 97.3 % stem matching accuracy on 150 greenhouse samples. These results demonstrate the effectiveness of accurately identifying and matching tomato organs in real-world scenarios, addressing the challenge from complex greenhouse environments. This integrated visual system enhances the success rate of visual identification and the precision of cutting position localization for tomato harvesting robots.