The accuracy of freehand dental implant placement is highly operator-dependent, and positional and angular deviations can increase procedural risk. Robotic assistance provides improved consistency and precision in clinical implantology. A CBCT-driven framework was developed to predict the six-degree-of-freedom position and orientation of missing teeth using multivariate regression with occlusal references. Multi-frame registration aligned imaging, simulation, and execution coordinate systems, while collision-free Cartesian trajectories were generated using RRT-Connect/RRT* and refined by seventh-order polynomial smoothing. In vitro validation was performed on high-fidelity 3D-printed jaw models. Mean coronal and apical deviations were 1.03±0.34 and 1.06±0.23mm, respectively, with a mean angular deviation of 3.47°±1.99°, all within accepted clinical tolerance thresholds. The proposed framework enhances the accuracy and repeatability of dental implant surgery, reduces dependence on surgeon experience, and supports safer, more standardised clinical execution.

China is one of the world’s major lychee producers, and the fruit’s soft texture, small size, and thin peel make non-destructive robotic harvesting particularly challenging. Accurate fruit detection, branch segmentation, and precise picking-point localization are critical for enabling automated harvesting in complex natural orchard environments. This study proposes an integrated perception framework for lychee harvesting that combines object detection, density-based clustering, and semantic segmentation. An improved YOLO11s-based detection network incorporating SimAM attention, CMUNeXt feature enhancement, and MPDIoU loss is developed to enhance robustness under illumination variation, occlusion, and scale changes. The proposed detector achieves a precision of 84.3%, recall of 73.2%, and mAP of 81.6%, outperforming baseline models. Density-based clustering is employed to group individual detections into fruit clusters. Comparative experiments demonstrate that MeanShift achieves the highest clustering consistency, with an average Adjusted Rand Index (ARI) of 0.768, outperforming k-means and other baselines. An improved DeepLab v3+ semantic segmentation network with a ResDenseFocal backbone and Focal Loss is designed for accurate branch extraction under complex backgrounds. Finally, a rule-based geometric picking-point localization algorithm is formulated in the image coordinate system by integrating detection, clustering, and branch segmentation results. Experimental validation demonstrates that the proposed framework can reliably localize picking points in two-dimensional images under natural orchard conditions. The proposed method provides a practical perception solution for intelligent lychee harvesting and establishes a foundation for future 3D robotic manipulation and field deployment.

Radio Frequency Identification (RFID) is a core technology in the perception layer of the Passive Internet of Things (Passive IoT). The complex external environments and an increase in the number of tags would cause channel contention and data conflicts during the reading process, which significantly affects the positioning accuracy and reading performance, resulting in the loss of data information stored in the tags. Although there are many methods to improve the reading performance of RFID system, most of them evaluate the reading performance through anti-collision protocol to optimize the position of reader and antenna. However, when moving antennas and readers read multi-tag, the solutions are neither efficient nor reliable. To improve the performance of RFID systems, this paper proposes an improved multi-tag position measuremnet for UNet networks, which can optimize the reading performance of RFID system. Firstly, for the electronic interference in complex environments, an experimental platform assisted visual intelligence is designed to construct the RFID multi-tag position measurement and performance analysis system. Secondly, the image deblurring for the UNet network improved by Implicit Neural Representations (INR) and Residual Fast Fourier Transform (Res FFT) is proposed to improve the image quality of the degraded multi-tag image. Finally, the 3D coordinates of the tags are found by YOLOv9 in the image coordinate system, which are subsequently converted to actual 3D distributions. It can improve the spatial distribution of RFID tags to Combine the prior knowledge of RFID 3D space in visual intelligence with the indicators of reading performance, thereby better guiding more effective physical tags placement methods. Experimental results demonstrate that the PSNR of the proposed method is 30.17 dB, and at least 2% better than the state-of-the-art algorithms, which indicates that the method proposed can accurately obtain the 3D distribution of multi-tag. Our system can capture the multi-tag 3D distribution corresponding to the maximum reading distance, thereby guiding the 3D structure distribution of multi-tag to enhance RFID reading performance.

Three-dimensional (3D) ultrasound imaging offers a larger field of view and enables volumetric measurements. Among the versatile methods, free-hand 3D ultrasound imaging utilizing deep learning networks for spatial coordinate prediction exhibits advantages in terms of simplified device configuration and user-friendliness. However, this imaging method is restricted to predicting the relative spatial transformation between two consecutive 2D ultrasound images, resulting in substantial cumulative errors. When imaging large organs, cumulative errors can severely distort the 3D images. In this study, we proposed a labeling strategy based on the ultrasound image coordinate system, enhancing the network prediction accuracy. Meanwhile, pre-planning the scanning trajectory and using it to guide the network prediction significantly reduced cumulative error. Spinal 3D ultrasound imaging was performed on both healthy volunteers and scoliosis patients. Comparison of reconstruction results across different methods demonstrated that the proposed method improved the prediction accuracy by approximately 40% and reduced the cumulative error by nearly 80%. This method shows promise for application in various deep learning networks and different tissues and is expected to facilitate the broader clinical adoption of 3D ultrasound imaging.

To accurately determine the relationships among sub-aperture cameras in four-aperture infrared bionic compound eye systems and enhance the target-positioning accuracy, addressing the issues that traditional single-aperture infrared cameras suffer from a limited imaging field-of-view, and multi-aperture camera systems fail to utilize all camera combination and exhibit slow target-positioning convergence speed due to neglecting pose differences of sub-eye cameras, static and dynamic target images were captured using the system. Target-positioning methods were then designed and investigated. Spatial and pose weights were assigned based on the spatial positions and rotation angles of the cameras. A reprojection method was employed to establish the relationships between various camera combinations using the conversion relationship between the image and world coordinate systems. A Kalman filter prediction and update method incorporating a spatially weighted projection was proposed for all camera combinations. Single-aperture and dual-target calibrations were performed using a four-aperture infrared bionic compound eye system to determine the actual field-of-view overlap rate, spatial angle, and pose angle. Static target-positioning experiments were performed using the four-aperture infrared bionic compound eye system. The results indicated that the static target-positioning algorithm achieved a horizontal error of 1.76% at a distance of 2660mm, a vertical error of 2.44%, and a spatial depth-estimation error of 0.40%. Ablation experiments were conducted to demonstrate the necessity of incorporating spatial weighting. Further verification through dynamic ball target-positioning experiments indicated that the positioning error in the four-aperture overlapping field-of-view at 2840mm could be controlled within 2.5%.

Dynamic magnetic resonance imaging (MRI) enables in vivo imaging of bone motion during knee movement, but quantifying joint kinematics from these images remains technically challenging due to image quality trade-offs inherent in dynamic acquisition sequences. We aimed to develop a semi-automated pipeline for tracking femoral and tibial motion from sagittal plane CINE MRI during active knee flexion and extension. The performance of the method was evaluated by quantifying: (i) bone boundary alignment error, (ii) frame segmentation processing time, and (iii) consistency of derived osteokinematic parameters, with the latter two compared against manual segmentation. The presented algorithm combines Canny edge detection and connected-component labeling with frame-to-frame transformation optimization to track bone boundaries. The approach was validated in five healthy volunteers performing controlled knee flexion and extension using a dedicated MRI-compatible device. The relative bone displacements measured using the semi-automated approach were qualitatively compared to that from manual segmentation. All bone displacements were defined in the two-dimensional (2D) image coordinate system, with the centroid of the tibial segment tracked relative to the centroid of the femoral segment in the horizontal and vertical directions. The semi-automated tracking method achieved an average alignment error of 0.40± 0.02mm for both bones, with processing time reduced from approximately 15 minutes for manual segmentation to less than 5 minutes for semi-automated segmentation per dataset. Both approaches showed similar relative bone motion patterns, with horizontal displacement of the tibia with respect to the femur ranging between 8 and 28mm and vertical displacement remaining relatively constant at around 57mm through the knee motion cycle. Further analysis revealed that the semi-automated method demonstrated improved precision with smaller standard deviations (SDs) in displacement measurements compared to the manual approach, with horizontal displacements of 1.7-2.7mm vs. 2.2-3.3mm and vertical displacements of 0.7-1.2mm vs. 0.9-1.7mm. These results demonstrate the potential of the semi-automated method for reliable and time-efficient quantification of relative bone positions during volitional knee motion in dynamic MRI protocols. The shorter processing time and the demonstrated reliability of the semi-automated method support its utility for analyzing dynamic MRI data.

Abstract A panoramic image provides a wide-angle field of view representation of a scene up to 360-degree. 360-degree images do not conform to the central projection camera model. One of the important research areas is the calculation of the three-dimensional coordinates of objects from a single 360-degree image more accurately. It is possible to calculate three-dimensional coordinates from a panoramic image using the direct linear transformation. However, distortions in panoramic images created by combining images from multiple fisheye lenses, such as the Ladybug2 camera, are not uniform across the image. In this study, a method for defining a non-uniform image coordinate system is proposed to address this issue. To generate the elliptical coordinate system the singular value decomposition algorithm was applied to the cylindrical coordinate system. The adaptive neuro-fuzzy inference system (ANFIS) method was applied to the elliptical coordinates to obtain a gridded non-uniform coordinate system. Then, three-dimensional coordinates of the control points in the calibration room were calculated from non-uniform coordinates by applying the direct linear transformation method. Finally, the three-dimensional coordinate values calculated from the proposed method were compared with the obtained conventional direct linear transformation method. When the proposed method was compared with the conventional method, it was found that the L4 parameter (488.82%) was calculated more accurately among the eleven direct linear transformation parameters. The results revealed that more accurate three-dimensional coordinate values were calculated with the proposed non-uniform coordinate system.

To achieve automation at the inner corner guard installation station in a steel coil packaging production line and enable automatic docking and installation of the inner corner guard after eye position detection, this paper proposes a binocular vision method based on deep learning for eye position detection of steel coil rolls. The core of the method involves using the Mask R-CNN algorithm within a deep-learning framework to identify the target region and obtain a mask image of the steel coil end face. Subsequently, the binarized image of the steel coil end face was processed using the RGB vector space image segmentation method. The target feature pixel points were then extracted using Sobel edges, and the parameters were fitted by the least-squares method to obtain the deflection angle and the horizontal and vertical coordinates of the center point in the image coordinate system. Through the ellipse parameter extraction experiment, the maximum deviations in the pixel coordinate system for the center point in the u and v directions were 0.49 and 0.47, respectively. The maximum error in the deflection angle was 0.45°. In the steel coil roll eye position detection experiments, the maximum deviations for the pitch angle, deflection angle, and centroid coordinates were 2.17°, 2.24°, 3.53 mm, 4.05 mm, and 4.67 mm, respectively, all of which met the actual installation requirements. The proposed method demonstrates strong operability in practical applications, and the steel coil end face position solving approach significantly enhances work efficiency, reduces labor costs, and ensures adequate detection accuracy.

One of the challenges in calibrating millimeter-wave radar and camera lies in the sparse semantic information of the radar point cloud, making it hard to extract environment features corresponding to the images. To overcome this problem, we propose a track association algorithm for heterogeneous sensors, to achieve targetless calibration between the radar and camera. Our algorithm extracts corresponding points from millimeter-wave radar and image coordinate systems by considering the association of tracks from different sensors, without any explicit target or prior for the extrinsic parameter. Then, perspective-n-point (PnP) and nonlinear optimization algorithms are applied to obtain the extrinsic parameter. In an outdoor experiment, our algorithm achieved a track association accuracy of 96.43% and an average reprojection error of 2.6649 pixels. On the CARRADA dataset, our calibration method yielded a reprojection error of 3.1613 pixels, an average rotation error of 0.8141°, and an average translation error of 0.0754 m. Furthermore, robustness tests demonstrated the effectiveness of our calibration algorithm in the presence of noise.

Optical information synthesis, which fuses LiDAR and optical cameras, has the potential for highly detailed 3D representations. However, due to the disparity of information density between point clouds and images, conventional matching methods based on points often lose significant information. To address this issue, we propose a regional matching method to bridge the differences in information density between point clouds and images. In detail, fine semantic regions are extracted from images by analyzing their gradients. Simultaneously, point clouds are transformed into meshes, where each facet corresponds to a coarse semantic region. Extrinsic matrices are used to unify the point cloud coordinate system with the image coordinate system. The mesh is further subdivided based on the guidance of image texture information to create regional matching units. Within each matching unit, the information density of the point cloud and the image is carefully balanced at a semantic level. The texture features of the image are well preserved in the transformed mesh structure. Consequently, the proposed texture synthesis method significantly enhances the overall quality and realism of the 3D imaging.

Due to the two-dimensional spatial variability of the range model during the missile-borne synthetic aperture radar (SAR) flight along a dive and curve trajectory, the accuracy of traditional algorithms is limited. In this paper, a new improved parametric polar format algorithm (IPPFA) is proposed to process missile-borne SAR imaging with a dive and curve trajectory. The algorithm divides the trajectory into two parts: dive trajectory and curve trajectory. For the dive trajectory processing, a new imaging coordinate system is established to transform the dive trajectory into a horizontal trajectory through coordinate rotation, and the space variation scene height change error term caused by coordinate rotation is embedded into the range variation compensation of the PPFA algorithm. Then, for the curve trajectory processing, based on the precise motion compensation ability of PPFA algorithm, we regard the curve trajectory caused by acceleration as the equivalent quadratic motion error that can be accurately analyzed, and then the equivalent motion error is processed through azimuth interpolation in PPFA. In addition, using the advantage that the equivalent motion error can be accurately expressed analytically, we derive a more accurate space-time spectral relationship to ensure the accuracy of wavefront curvature compensation. Finally, the superiority and effectiveness of the proposed algorithm are verified by comparing the simulation and real-data experiments with other imaging algorithms.

Abstract Feature point extraction plays a key role in visual simultaneous localization and mapping (SLAM) systems. And it remains a major challenge to accurately select static feature points in a complex dynamic environment. To address this issue, this paper proposes an RGB-D SLAM method, referred to as DE-RGBD SLAM, which optimizes feature selection by integrating depth information and effectively utilizes depth data and multi-view geometric information to achieve localization and navigation for mobile robots in dynamic environments. Firstly, the method analyzes prominent feature regions in the image based on colour and depth information captured by an RGB-D camera. It sets adaptive FAST corner detection thresholds according to the grayscale information of these regions while masking other areas. Next, the method obtains in-depth information on the detected feature points in the current frame. It combines their pixel coordinates in the image coordinate system to determine the presence of redundant feature points. Notably, the method can detect some dynamic feature points between consecutive frames. Subsequently, in the camera coordinate system, the method compares the depth information of feature points in the depth image with the epipolar depth estimates derived from the essential matrix to determine whether the features are static and eliminate dynamic feature points. This approach significantly enhances the reliability of static feature points. Finally, the accuracy and robustness of the proposed method are validated through experiments conducted on the public TUM dataset and real-world scenarios compared to state-of-the-art visual SLAM systems.

Foreign object intrusion detection is one of the important means to ensure the safe operation of rail transit. The distance measurement of the intruding foreign object is conducive to reminding the train crew to take corresponding measures in time. This paper proposes a distance detection algorithm for intruding foreign objects based on monocular vision. Firstly, corner detection and threshold segmentation methods are used to obtain the image coordinates of a black-and-white checkerboard. Secondly, regression analysis is employed to establish a conversion model between image coordinates and world coordinates. Finally, an object detection algorithm based on the YOLOv10 network is used for image recognition to obtain the coordinates of the intruding foreign objects in the image coordinate system. The distance information of the intruding objects is detected. Experimental validation shows that the distance of the intruding objects has an error range of -5 to +6 cm in the X-axis direction and -9 to +16 cm in the Y-axis direction, demonstrating high accuracy.

Abstract Visual vibration measurement has emerged in the field of structural health monitoring in recent years, but it still has some shortcomings in terms of resolution, recognition rate and real-time performance. Considering the three aspects of recovering high-frequency image details, improving the compactness of the target bounding box, and reducing the computational time, we use the constructed image super-resolution reconstruction model and target detection model to measure the vibration displacement of the bridge structural model. First, we integrate the Transformer module into the Unet network with a simple structure. The Swin and Global Transformer Unet (SGTU) module constructed in this form can reduce the computational cost while reconstructing the large-resolution feature map target, and it can sharply edge information of the vibration target. We use the framework of the YOLOv5 algorithm as the backbone, and use the GhostBottleneck (GB) module to reduce the time for convolution operations to generate similar features. In addition, the proposed DWCBottleneck (DWCB) fusion module is also able to achieve high-level semantic fusion and network depth expansion with minimal computational cost. Finally, the center point offset of the bounding box predicted by the model can be used to obtain the displacement offset of the object in the image sequence. The position information of the target in the first frame image is used as the reference frame for calculating the offset, and the vibration displacement of the flexible structure in the image coordinate system is obtained by calculating the deviation of the displacement between the remaining frames and the first frame. We perform qualitative and quantitative comparisons in three aspects: video super-resolution reconstruction, visual detection robustness, and sensor vibration measurement displacement using a homemade vibration image dataset. The time–frequency domain displacement curves regressed by the visual vibration measurement algorithm are compared with the curves acquired after accelerometer acquisition, indicating the necessity of super-resolution reconstruction in visual vibration measurement.

A panorama ensures a stunning wide-angle field of view up to 360° representation of a scene, exceeding the limits of a normal photograph. Panoramic cameras satisfy the single-viewpoint characteristic. There are several types of panoramic cameras for 360-degree imaging. Multi-camera panoramic imaging systems pose a difficulty in obtaining a single projection center for the cameras. In a variety of practical implementations of panoramic cameras, it is possible to calculate three-dimensional coordinates from a panoramic image, especially using the Direct Linear Transformation (DLT) method. In this study, not only a defining method of the non-uniform image coordinate system is presented by utilizing the C-Means algorithm for a single panoramic image, captured with a Ladybug2 panoramic camera in a panoramic calibration room but also the use of an elliptical panoramic projection coordinate system is defined by the Singular Value Decomposition method in a panoramic view. The results of the suggested method have been compared with the DLT algorithm for a single panoramic image which defined a conventional photogrammetric image coordinate system. It has been observed that the proposed method provides more accurate results for the 3D coordinate definition.

This article presents an innovative method for planning and tracking the trajectory in the image plane for the visual control of a quadrotor. The community of researchers working on 2D control widely recognizes this challenge as complex, because a trajectory defined in image space can lead to unpredictable movements of the robot in Cartesian space. While researchers have addressed this problem for mobile robots, quadrotors continue to face significant challenges. To tackle this issue, the adopted approach involves considering the separation of altitude control from the other variables, thus reducing the workspace. Furthermore, the movements of the quadrotor (pitch, roll, and yaw) are interdependent. Consequently, the connection between the inputs and outputs cannot be reversed. The task complexity becomes significant. To address this issue, we propose the following scenario: When the quadrotor is equipped with a downward-facing camera, flying at high altitude is sensible to spot a target. However, to minimize disturbances and conserve energy, the quadrotor needs to descend in altitude. This can result in the target being lost. The solution to this problem is a new methodology based on the principle of differential flatness, allowing the separation of altitude control from the other variables. The system first detects the target at high altitude, then plots a trajectory in the image coordinate system between the acquired image and the desired image. It is crucial to emphasize that this step is performed offline, ensuring that the image processing time does not affect the control frequency. Through the proposed trajectory planning, complying with the constraints of differential flatness, the quadrotor can follow the imposed dynamics. To ensure the tracking of the target while following the generated trajectory, the proposed control law takes the form of an Image Based Visual Servoing (IBVS) scheme. We validated this method using the RVCTOOLS environment in MATLAB. The DJI Phantom 1 quadrotor served as a testbed to evaluate, under real conditions, the effectiveness of the proposed control law. We specifically designed an electronic card to transfer calculated commands to the DJI Phantom 1 control joystick via Bluetooth. This card integrates a PIC18F2520 microcontroller, a DAC8564 digital-to-analogue converter, and an RN42 Bluetooth module. The experimental results demonstrate the effectiveness of this method, ensuring the precise tracking of the target as well as the accurate tracking of the path generated in the image coordinate system.

The star sensor is the key component of Celestial Navigation. It measures the autonomous attitude of navigation bodies by observing stars. And it conducts image collection, preprocessing, feature extraction and matching recognition. Aimed to implement the latter two procedures, we first estimate the coordinate of the point which is the intersection point of the optical axis and the celestial sphere. We employ geometrical knowledge to get the relationship between the intersection point and given projection distances. When distances are unknown, we use Newton’s method to approach the exact coordinate of the intersection point. Based on our coordinate calculation method, we are required to find a principle for improving the accuracy of the coordinate. We first establish a projection screening model to obtain star maps. Then we establish four coordinate systems, i.e., the celestial coordinate system, the star sensor coordinate system, the image coordinate system and the pixel coordinate system. Taking the star map at the north celestial pole as an instance, we finish the transformation of coordinate between different systems and search for the factors affecting accuracy of coordinate. Ultimately, we draw the conclusion that the coordinate accuracy improves, when selected stars projection close the centroid of the photosensitive surface. Aimed to implement the matching recognition, we establish a novel feature extraction and matching model. We take the angle between stars and three of their nearest stars as the feature of the central star. Then we extract the feature matrix of the given star table as the feature database. Using the same way, we get the feature matrix of four–star maps. To achieve the last step of matching recognition, we compare the feature matrix of star maps with the given navigation stars. During the process, we employ DBScan clustering algorithm to implement the matching recognition process. We select the cluster center that satisfies the maximum number of matches as the actual location of the identified star map.

The time-domain imaging algorithm can meet the imaging requirements of bistatic synthetic aperture radar(Bi-SAR) system with flexible and complex configurations, and establishing the imaging coordinate system is the key step especially for Fast Time-Domain algorithms. In this paper, to simplify the establishment of imaging coordinate system, the orthogonal elliptical mapping polar (OEMP) coordinate system is proposed and used in the frame of the fast factorized back projection(FFBP) algorithm. Compared to the orthogonal elliptical polar(OEP) coordinate system, the proposed coordinate system has simpler coordinate solving, and can also reduce the computational burden of the FFBP algorithms. The applicability and performance of the OEMP-based FFBP algorithm is verified through simulation experiments.

Currently, the practical implementations of panoramic cameras range from vehicle navigation to space studies due to their 360-degree imaging capability in particular. In this variety of uses, it is possible to calculate three-dimensional coordinates from a panoramic image, especially using the Direct Linear Transformation (DLT) method. There are several types of omnidirectional cameras which can be classified mainly as central and non-central cameras for 360-degree imaging. The central omnidirectional cameras are those which satisfy the single-viewpoint characteristic. Multi-camera systems are usually developed for applications for which two-image stereo vision is not flexible enough to capture the environment surrounding a moving platform. Although the technology based on multi-view geometry is inexpensive, accessible, and highly customizable, multi-camera panoramic imaging systems pose a difficulty in obtaining a single projection center for the cameras. In this study, not only a defining method of the non-uniform image coordinate system is suggested by means of the K-Means algorithm for a single panoramic image, captured with a Ladybug2 panoramic camera in the panoramic calibration room but also the use of an elliptical panoramic projection coordinate system definition by Singular Value Decomposition (SVD) method in panoramic view. The results of the suggested method have been compared with the DLT algorithm for a single panoramic image which defined a conventional photogrammetric image coordinate system.

Abstract Art and design professional education can not keep up with the development speed of the industry, and there is no reference to the market demand data, failing to keep pace with the times, which brings many negative impacts on the employment of students. This paper suggests an art design professional education model that incorporates VR interactive scene technology for this reason. Through the double camera, we simulate the visual principle of the human eye, collect three-dimensional data from VR interactive scenes, and use the image coordinate system and pixel coordinate system conversion operations to pre-process the collected data. The optimal layout solution is sought by adjusting the layout to generate a realistic 3D model of the teaching scene and to realize the construction of the VR interactive scene. The principle and implementation process of the art professional education model supported by the VR interactive scene is elaborated, and the experimental comparison method is used to empirically analyze the art and design professional education model integrating VR technology. The results show that there is a significant difference between the experimental group and the control group in the dimensions of adaptability (0.004) and uniqueness (0.044) of design thinking and the cultivation of design thinking ability (0.016) (P<0.05), which indicates that the art and design professional education model integrating the VR scene interaction technology constructed in this paper has a significant effect on the enhancement of design thinking ability of college students.