86 results about "Bundle adjustment" patented technology

The invention discloses a loopback SLAM (Simultaneous Localization and Mapping) method based on three-dimensional Gaussian sputtering and multi-camera input. According to the invention, fusion of three-dimensional Gaussian sputtering and multi-camera input is utilized to realize autonomous panoramic data acquisition and scene reconstruction with higher acquisition efficiency; according to the method, the constraint is constructed by using the overlapped part between the cameras, so that more accurate camera pose estimation and high-quality scene modeling are realized; according to the method, timestamp attributes are added to gauss, the gauss are classified according to the timestamp attributes, and the loopback is rapidly and effectively detected through different classes of gauss proportions under a current frame camera view angle; after the loopback is detected, the camera pose is adjusted, the problem of camera drifting is solved, meanwhile, a Gaussian map can be updated according to adjustment of the camera pose, and an accurate three-dimensional model is kept. According to the two-stage binding adjustment strategy provided by the invention, the global camera pose is finely adjusted by using the multi-view rendering image loss and the pose image constraint.

The invention provides a self-adaptive multi-modal odometer dynamic fusion method which specifically comprises the following steps: extracting edge and plane features of a laser point cloud, and generating a laser pose factor by utilizing iterative motion compensation and residual optimization; multi-scale angular point features are extracted from an original image, and visual pose factors are generated in combination with optical flow tracking and local light beam adjustment. And a dynamic factor graph is constructed, laser and visual pose factors are used as constraints, an iSAM2 engine is used for optimization solution, and a global pose is obtained. Calculating the residual mean value and variance of Lagrange factors in real time, dynamically checking the state of the sensor based on chi-square test, triggering failure response during continuous overrun, gradually removing abnormal factors, keeping historical data, aligning timestamps through SE (3) manifold interpolation, and optimizing external parameter calibration. And after the sensor is recovered, factors are re-introduced through progressive weight fusion, and a sliding window mechanism is adopted to smooth the track. The method provides a robust and high-precision fusion algorithm for the field of robots and automatic driving.

The invention relates to an extraterrestrial celestial body boundary-free scene three-dimensional reconstruction method based on a Gaussian radiation field, and the method comprises the steps: extracting visual features of an input image, and obtaining internal parameters, external parameters and scene sparse point cloud of a camera through feature matching and cluster adjustment; an expanded dense point cloud is calculated according to a light plane intersection mode, a series of three-dimensional Gaussian distributions are constructed, and pixel colors in a given visual angle direction are rendered in a rasterization mode; aiming at a rendering result, guiding a Gaussian position to be accurately close to a scene surface by utilizing pre-estimated depth prior in an optimization process; a scene surface is accurately captured in combination with an explicit and implicit model joint representation mode; and rebuilding a fine and complete three-dimensional mesh model on the surface of the extraterrestrial celestial body by using a marching tetrahedron algorithm according to an optimization result. According to the method, fine three-dimensional reconstruction of the surface of the extraterrestrial celestial body in a boundary-free scene can be realized, the modeling capability of a Gaussian radiation field in the boundary-free scene is enhanced, and the precision and integrity of three-dimensional reconstruction of the surface of the extraterrestrial celestial body are improved.

The invention relates to a multi-small-ball auxiliary multi-camera calibration method and system based on deep learning, and the method comprises the steps: obtaining a plurality of groups of image data sets, and extracting the image coordinates of all speckle mark points on a calibration part of each group of images from the image data sets through a target detection algorithm; based on the image coordinates and the geometric information of the calibration component, solving a projection matrix of multiple cameras by using a multi-view geometric relationship and an Euclidean structure recovery algorithm; taking the projection matrix as an initial value, and performing nonlinear optimization on the projection matrix by using a bundle adjustment algorithm to obtain an optimized calibration result; and performing world coordinate system matching on the optimized calibration result and a fixed checkerboard calibration plate to realize multi-camera pose calibration. The method disclosed by the invention is simple and practical, has good robustness, and can quickly and accurately identify the calibration object.

The invention relates to the field of computer vision and mobile robot navigation, and discloses an object-level semantic vision SLAM method and system based on a mixed attention mechanism target detection network and an ellipsoid model. The system comprises five parallel thread modules, namely a semantic perception module, a visual tracking and repositioning module, a local mapping and fusion module, a loopback detection module and a global consistency optimization module. According to the method, local and global features of an image are extracted in parallel through a target detection network, and high-precision semantic observation is output; when the tracking is lost, the dual geometric constraint of the 2D internally tangent ellipsoid-3D object ellipsoid is utilized, and the P3P algorithm and the IoU cost function are matched to realize rapid relocation. In the mapping process, Gaussian-Wasserstein distance is introduced to measure semantic re-projection errors, and a joint objective function containing map point geometric errors and object semantic errors is constructed to carry out bundle adjustment. According to the method, the problem of feature extraction failure in motion blur and weak texture scenes is effectively solved, and the construction precision of the semantic map and the survivability of the system are remarkably improved.

The invention discloses an enhanced 3D (three-dimensional) surface reconstruction method based on 3D (three-dimensional) Gaussian Splitting. Global consistent depth priori is obtained through virtual stereo pair rendering; constructing a factor graph and introducing a cross-view geometry / luminosity consistency constraint to form local beam adjustment loss; the prior is used as a learnable parameter to be combined with 3DGS to be optimized, and meanwhile, the Pull loss is assisted to pull low-credibility pixels; and finally, multi-loss function end-to-end training is adopted. The method comprises the following steps of: in Tanksamp; the method has the advantages that F1 is equal to 0.58 and DTU Chamfer is equal to 0.48 mm on a Temples data set, the training time is only 20 min, compared with the prior art, geometric accuracy SOTA and speed magnitude improvement are achieved at the same time, and the method is suitable for VR / AR, robot and industrial measurement scenes.

The invention provides a NeRF visual SLAM mapping method based on ORB tracking and three-plane hash coding. Firstly, ORB feature point extraction and feature matching are carried out, key frames are obtained through screening, and an initial sparse map is constructed: local common-view key frames are screened and optimized through a local beam adjustment method, and light sampling is carried out on the key frames in a local common-view key frame window; performing local optimization on scene features corresponding to the key frames in the NeRF map by using a NeRF optimization method based on three-plane hash coding; after loopback is detected, global light beam adjustment method optimization is carried out on the global key frame, and a NeRF optimization method based on three-plane Hash coding is used for carrying out global optimization on scene features of a NeRF map so as to guarantee the global consistency of the constructed map; and finally, extracting surface information of the scene from the NeRF map, and generating a three-dimensional reconstruction model. The combination of three-plane Hash coding and a multi-layer perceptron is used as NeRF map scene representation, and scene details can be reconstructed efficiently and delicately.

The invention discloses a track alignment monitoring method and device based on binocular vision measurement, and the method comprises the steps: fixing a marker marked with a coded image at each track segment, enabling a binocular camera to move along a measurement track, and carrying out the shooting at different measurement points; carrying out image preprocessing on the shot photo sequence, carrying out target detection, and extracting a coded image; calibrating internal parameters of the binocular camera and external parameters of each measuring point; obtaining a three-dimensional coordinate of each marker in a local binocular coordinate system according to the internal parameters of the binocular camera, and converting the three-dimensional coordinate into a three-dimensional coordinate of each marker in a global coordinate system by adopting the external parameters of the binocular camera; and performing beam adjustment method optimization on all three-dimensional coordinates of the markers with consistent unique mark numbers under the global coordinate system to obtain optimized three-dimensional coordinates of all the markers under the global coordinate system, and gathering according to a sequence to obtain a three-dimensional linear curve of the measurement orbit. The method is high in precision and free of environmental interference.

The invention relates to a structured light three-dimensional measurement technology, and provides a system joint calibration optimization method fusing three-dimensional offset modeling and phase identity constraint for an industrial field low-precision target scene. The method comprises the following core steps: introducing a three-dimensional offset for a target feature point, and compensating spatial deformation and eliminating scale ambiguity through zero-mean constraint; sub-pixel matching is realized based on a phase identity principle, and projection constraint is enhanced to reduce dependence on a target; and constructing a joint reprojection error function, and adjusting and synchronously optimizing system parameters and offset by means of a bundle method. The prior art depends on a high-precision plane target, and when the target has manufacturing errors, installation warping or pasting deformation, calibration errors can be transmitted through an imaging link, so that the measurement precision is reduced, and target deformation and geometric errors of a projector are difficult to compensate at the same time. The method breaks through the limitation of a traditional inverse camera model, does not need to depend on a high-precision glass target, reduces the hardware cost, and constructs a cooperative compensation mechanism of a target manufacturing error and an imaging parameter. Experiments show that after explicit compensation is carried out on the local deformation of the target by adopting the three-dimensional offset factor, the flatness error gt is detected; on a paper target of 10 microns, the re-projection error of the camera is reduced by 79.6%, and the re-projection error of the projector is reduced by 74.8%; sub-pixel matching is achieved through phase identity constraint, multi-parameter collaborative optimization of bundle adjustment is combined, the high-precision measurement requirement of an industrial site is met, and an efficient solution is provided for high-precision three-dimensional measurement.

Techniques are described for automated operations to analyze visual data from images acquired in multiple rooms of a building to generate building information that may include a floor plan for the building, such as by analyzing visual overlap between those images to determine information that includes global inter-image pose and locations of walls and optionally other structural elements, and by using the generated building information in further automated manners. In some situations, the described techniques include using a combination of a trained diffusion transformer machine learning model and a bundle adjustment optimizer to determine global inter-image pose and wall location data and to use that data to generate a resulting floor plan for the building, such as to operate in parallel or with the bundle adjustment optimizer as a layer within the diffusion model that provides guidance for its automated determinations.

The invention relates to the technical field of three-dimensional reconstruction, in particular to a coarse aggregate space engraving three-dimensional reconstruction method based on a sparse structure. Constructing an unshielded multi-view image acquisition system, and acquiring a three-dimensional contour image of the coarse aggregate particles; acquiring a parameter matrix of image acquisition equipment in the unshielded multi-view image acquisition system; according to the parameter matrix of the image acquisition equipment, a projection matrix of the image acquisition equipment is calibrated by adopting a bundle adjustment and nonlinear optimization method; and establishing a sparse voxel space, and performing three-dimensional reconstruction on the coarse aggregate particles in the sparse voxel space by adopting a space engraving algorithm according to the three-dimensional contour image of the coarse aggregate particles and the calibrated projection matrix. According to the reconstruction method provided by the invention, a non-shielding multi-view imaging system and a space engraving algorithm are adopted, so that high-precision three-dimensional contour reconstruction of coarse aggregate particles can be realized, and the accuracy and efficiency of reconstruction are remarkably improved.

The invention relates to the technical field of planet remote sensing data processing and three-dimensional reconstruction, and discloses a Mars three-dimensional terrain reconstruction method based on feature enhancement and multi-view stereo matching, which comprises the following steps: firstly, obtaining and preprocessing Mars stereo image data; secondly, extracting texture features based on a gray-level co-occurrence matrix, and performing region segmentation by using local entropy to distinguish a high texture region from a low texture region; then, parameters such as the size of a matching block and a similarity threshold value are adaptively adjusted according to a segmentation result, and a multi-dimensional mixed feature descriptor is generated to perform homonymy point matching; a coupling threshold model of the solar incident angle and the terrain roughness is constructed, and parallax calculation of the shadow and the steep terrain is optimized; and finally, combining cross-scale parallax propagation and global bundle adjustment to reconstruct a three-dimensional terrain model. According to the method, the technical problems that the matching reliability of areas such as weak textures and shadows on the surface of Mars is insufficient and the model is discontinuous are solved, and the integrity, the self-adaptability and the global precision of the reconstruction model are improved.

A joint calibration optimization method applied to underground space mainly comprises the following steps: establishing an IMU pre-integration model, and deducing continuous and discrete forms of the IMU pre-integration model for IMU measurement preprocessing; an IMU error parameter residual error is added in bundle adjustment based on the mahalanobis distance to serve as an optimization item, and a Jacobian matrix of the optimization item is deduced; in order to solve the problem that the covariance matrix is irreversible due to the increase of IMU error parameters, a method for regularizing the covariance matrix is adopted, so that the excessive singularity of the matrix can be prevented, and the reversibility of the matrix is improved; visual point feature reprojection errors and visual line feature reprojection errors serve as error factors to be introduced into an optimization function, more constraints and information can be provided to help optimization of a sensor, and through utilization of gray information, a calibration system can have higher robustness in an underground space region with poor texture. By constructing the joint optimization function, joint optimization of a plurality of error factors can be realized, and the calibration result is optimized.

System and method are provided for improving camera pose accuracy in augmented reality (AR) systems. The method detects and processes inconsistencies in captured camera poses by identifying locally rigid pose groups and matching visual features between images both within and across these groups. The process involves triangulating 3D landmarks within pose groups, establishing correspondences between groups, and performing bundle adjustment to optimize camera poses. This systematic approach enables the generation of accurate 3D models by detecting and correcting pose drift through feature matching, landmark triangulation, and global pose optimization across multiple camera positions and orientations.

The invention discloses a fabric-like sewing type cavity self-healing endoscope scene reconstruction method. The method comprises the following steps: carrying out initialization modeling on a two-dimensional Gaussian scene; tracking the pose of the camera; gaussian extension and key frame sampling strategy; a mapping mapping and hole perception completion module; light beam adjustment optimization of fusion smooth constraint is carried out; according to the method, firstly, continuous and dense representation of an endoscope scene is realized through two-dimensional Gaussian scene modeling and a Gaussian extension mechanism; 2, a cavity sensing and complementing mechanism is adopted, so that the mapping process has a fabric-like sewing type self-healing capability; thirdly, light beam adjustment optimization of normal smooth constraint is introduced, and the phenomena of local geometric roughness and Gaussian scene discontinuity are relieved; the overall frame is designed in a modularized mode and can be seamlessly integrated with an existing two-dimensional Gaussian mapping and light beam adjustment system; and 5, the method is suitable for a complex tubular organ scene. According to the method, topological perception and self-healing complementation of the mapping process can be realized in a complex tubular organ scene, and the continuity, stability and geometric quality of an endoscope three-dimensional reconstruction result are improved.

The invention provides a space debris three-dimensional reconstruction method based on grazing observation imaging, belongs to the field of three-dimensional reconstruction, solves the problem of how to accurately predict the future activity trend of space debris, and comprises the following steps: obtaining an optical observation image after meeting a grazing observation constraint; sFM preprocessing, multi-view internal feature point extraction, multi-view feature point matching and bundle adjustment are carried out on the optical observation image to obtain sparse point cloud and pose information; dividing the preprocessed image into a training set, a test set and a verification set; and inputting the sparse point cloud and the pose information as initial conditions and the training set into an implicit neural expression function of NeRF at the same time, mapping the input into a feature vector after position coding correlation positioning, training the MLP, and outputting and rendering an optimal fitting result to obtain a three-dimensional reconstruction optical observation image. The working state of the space debris is evaluated by analyzing the three-dimensional reconstruction optical observation image, and the future activity trend of the space debris is predicted.

This invention belongs to the field of image generation, specifically relating to a method and system for generating 360° images of subway vehicles based on image stitching. The method includes: acquiring source images collected along the subway vehicle; obtaining effective matching point pairs through feature extraction matching and spatial topological similarity filtering; selecting the camera pair with the strongest geometric constraints as a benchmark; using incremental triangulation to recover the camera pose and 3D point cloud; in global bundle adjustment, optimization is performed by combining reprojection error and linear regularization terms based on the geometric principal axes of principal component analysis, with the regularization term weights adjusted according to the average distance from the point cloud to the principal axes; after optimization, the source images are projected onto a cylindrical canvas, and the optimal image gradient is selected based on gradient information entropy to construct a gradient field; a seamless 360° panoramic image is generated by solving the Poisson equation. This invention can effectively eliminate mismatches, suppress 3D reconstruction offset, and generate high-quality seamless panoramic images.

The invention belongs to the field of space target imaging, and discloses a space target three-dimensional reconstruction method for inhibiting dark color interference through combination of brightness and transparency, which is used for solving the problems of poor imaging quality and inaccurate pose estimation of a space target. According to the method, multiple requirements of improving the quality of an input image, adjusting a pose and training a three-dimensional model are jointly completed, and low-quality imaging of an incomplete imaging system is preprocessed based on lucky imaging; and then, dense feature points are extracted by using a pre-training network based on no detector, estimation of a three-dimensional initial point cloud and an initial pose of a camera is carried out, and pose joint adjustment is carried out by combining RANSAC and Bundle AdJustlement. During training, an MLP-based optimizer is used for calculating the deviation of pose adjustment, meanwhile, training of parameters of three-dimensional Gaussian primitives is carried out, and after various physical constraints are added, a one-stop assembly line can be implemented. According to the method, the good basic shape and the sharp edge of the model are kept, the good anti-noise capability is achieved, and meanwhile the real-time performance and the robustness are achieved.

Examples described herein provide a method for trajectory estimation using an image sequence. The method includes receiving an image sequence of an environment from a camera moving relative to the environment. The method further includes extracting and matching features from images of the image sequence. The method further includes determining a relative orientation of the images of the image sequence. The method further includes determining orientation parameters of the camera using sequential image resection. The method further includes performing, using the orientation parameters, bundle adjustment to generate refined orientation parameters of the camera. The method further includes estimating a trajectory of the camera relative to the environment based at least in part on the refined orientation parameters by performing loop closure.

This invention discloses a novel target localization method for a stitched mapping camera, comprising: obtaining the interior orientation elements of each sub-image of the camera based on the calibration parameters of the stitched mapping camera; obtaining the exterior orientation elements of each sub-image based on the POS data of each exposure point in the aerial photography operation and the transformation relationship between each sub-image; establishing an overlap relationship table between each sub-image based on the image overlap rate and the transformation relationship between each sub-image; finding each pair of sub-images with an overlap relationship based on the overlap relationship table, extracting corresponding feature points within the overlapping area of ​​the sub-images, and performing feature matching to obtain the image plane coordinates of the corresponding feature points in each sub-image; treating each sub-image with obtained image plane coordinates as an independent camera, performing aerial triangulation, and using bundle adjustment to jointly solve for the ground positioning result. This invention avoids the accuracy loss caused by the near single-center projection processing of the stitched mapping camera, thus improving the ground positioning accuracy.

The invention relates to the technical field of image processing, and discloses a landslide surface crack three-dimensional evolution unmanned aerial vehicle image extraction and quantification method and system, and the method comprises the steps: carrying out the bundle adjustment of an ultrahigh overlapping degree aerial image of a landslide surface, and obtaining an image shooting position, an image attitude parameter and a sparse three-dimensional point cloud; constructing a high-precision digital earth surface model; performing digital mosaic on the aerial image with the ultrahigh overlapping degree to obtain an orthoimage; performing semantic segmentation on the orthoimage to obtain a crack binary segmentation map of the orthoimage so as to identify a main crack and a secondary crack; performing three-dimensional geometric parameter analysis on the main crack and the secondary crack to obtain a crack three-dimensional space attribute parameter; carrying out evolution trajectory tracking on the crack three-dimensional space attribute parameters to obtain evolution data so as to construct a visual map of the landslide surface; according to the invention, the efficiency of extracting and quantifying the three-dimensional evolution unmanned aerial vehicle image of the landslide surface crack can be improved.

A method, system, and apparatus for an unmanned aerial vehicle (UAV) to autonomously reconstruct overflown terrain and detect safe landing sites. A UAV autonomously acquires on-board pose estimates from an on-board visual-inertial-range odometry method during flight. The on-board pose estimates are utilized as a pose prior and to regain metric scale during three-dimensional (3D) reconstruction. The on-board pose estimates are corrected based on a bundle adjustment approach using previously acquired images. 3D reconstruction is performed based on multiple captured images taken from an on-board camera. Range data from the multiple captured images is fused into a multi-resolution height map. A safe landing site on the terrain is detected based on the multi-resolution height map.

Methods, systems, and apparatus for performing bundling adjustment using epipolar constraints. A method includes receiving image data from a headset for a particular pose. The image data includes a first image from a first camera of the headset and a second image from a second camera of the headset. The method includes identifying at least one key point in a three-dimensional model of an environment at least partly represented in the first image and the second image and performing bundle adjustment. Bundle adjustment is performed by jointly optimizing a reprojection error for the at least one key point and an epipolar error for the at least one key point. Results of the bundle adjustment are used to perform at least one of (i) updating the three-dimensional model, (ii) determining a position of the headset at the particular pose, or (iii) determining extrinsic parameters of the first camera and second camera.

The present application relates to the technical field of map saliency localization, and particularly relates to a saliency localization method based on learning. The saliency localization method based on learning provided by the present application adopts a saliency prediction model to simulate the mechanism in a SLAM framework. A saliency map is predicted according to a saliency model, and the model can capture scene semantics and geometric information. The value of the saliency map is used as the weight of a feature point in a traditional bundle adjustment method. Detailed experiments conducted on the KITTI and EuRoc datasets with the most advanced algorithms show that the algorithm proposed by the present application is superior to existing algorithms in indoor and outdoor environments, and significantly improves the positioning accuracy and robustness.

The invention belongs to the technical field of visual simultaneous localization and mapping, and particularly discloses a binocular visual simultaneous localization and mapping method and device based on bionic divergence. According to the invention, the method comprises the steps: carrying out the equal cutting of the overlapping region of a plurality of images according to the cutting width, and carrying out the independent feature extraction of an effective cutting region; after a re-projection error is calculated by combining an observed camera device, parameter optimization is carried out according to the re-projection error based on a double-map beam adjustment optimization algorithm of an external parameter constraint. Through the above mode, a bionic divergent layout is taken as a core, a divergent binocular configuration is constructed through two camera devices which swing outwards and partially overlap view fields, overall view field expansion is realized, equivalence of virtual camera devices is completed by adopting a divergent binocular modeling strategy, and depth estimation and wide-angle sensing capability are optimized. In combination with an external parameter constrained double map beam adjustment optimization algorithm, the system structure consistency is ensured and the calculation redundancy is reduced, so that the positioning precision and the rotation robustness can be effectively improved.

Systems and techniques are provided for image processing for calibration. For example, a computing device can obtain image frames of a three-dimensional (3D) scene, each image frame including a plurality of two-dimensional (2D) points. Each 2D point corresponds to a 3D point. The computing device can determine a subset of 3D points by applying bundle adjustment on a set of 3D points (distributed over a field of view of the camera) and on fixed parameters of a camera. The computing device can determine inlier points from the subset of 3D points with a reprojection error less than a threshold value. The computing device can determine a final 3D points and final parameters of the camera by applying the bundle adjustment on the inlier points and on a prior set of camera parameters. The computing device can apply, to the camera, final intrinsic parameters of the final parameters of the camera.

The application relates to the field of computer vision and mobile robot navigation, and discloses an object-level semantic visual SLAM method and system based on a hybrid attention mechanism target detection network and an ellipsoid model. The system comprises five parallel thread modules of semantic perception, visual tracking and repositioning, local mapping and fusion, loop detection and global consistency optimization. The application extracts local and global features of an image in parallel through a target detection network, and outputs high-precision semantic observations; when tracking is lost, a 2D inscribed ellipse-3D object ellipsoid dual geometric constraint is used to realize fast repositioning in cooperation with a P3P algorithm and an IoU cost function. In the mapping process, a Gaussian-Wasserstein distance is introduced to measure semantic re-projection errors, and a joint objective function containing map point geometric errors and object semantic errors is constructed to perform bundle adjustment. The application effectively solves the problem of feature extraction failure in motion blur and weak texture scenes, and significantly improves the construction accuracy of a semantic map and the survival ability of the system.

The present invention discloses a monocular vision SLAM method for a dynamic environment, including: 1) initialization, including reading the first two frames of an image sequence, extracting and matching ORB feature points, establishing a world coordinate system, setting monocular scale information, establishing an initial map, constructing a key frame sequence and a key frame sliding window; 2) tracking a reference frame and estimating an initial pose; 3) eliminating dynamic feature points and optimizing the pose; 4) according to the tracking result, inserting a key frame and tracking a reference key frame, and then constructing map points and inserting them into the map; 5) eliminating dynamic map points on the sliding window and performing local bundle adjustment optimization; 6) eliminating redundant key frames and map points; 7) if the device computing power is sufficient, performing global bundle adjustment optimization; 8) repeating steps 2)-7) until all image frames in the sequence are processed. The present invention solves the problems of weak robustness and poor accuracy in positioning and mapping when the vision SLAM method is applied to a dynamic environment.

This invention discloses a calibration-free panoramic camera and laser point cloud registration method and system. The method includes: acquiring laser point clouds and a panoramic image; splitting the panoramic image into multiple pinhole images and performing feature extraction and matching, then recalculating the matching results back into the panoramic image; reconstructing the sparse point cloud using a motion recovery structure and obtaining the initial pose of the panoramic image; registering the sparse point cloud and the laser point cloud using an iterative nearest-point algorithm, calculating seven parameters to achieve preliminary alignment; and improving the bundle adjustment method by combining the plane normal information corresponding to the sparse point cloud with the geometric constraints of the laser point cloud to achieve precise alignment between the panoramic image and the laser point cloud. This invention does not rely on any external calibration parameters, IMU information, or prior scene information. Through image self-reconstruction and laser constraint fusion, it significantly improves the registration accuracy and geometric consistency of 3D reconstruction, and is suitable for multi-source point cloud fusion reconstruction in scenarios without auxiliary information.

The invention provides a fringe projection trinocular vision calibration method based on phase reliability weighting, and the method comprises the steps: constructing a three-dimensional point set for each feature circle by taking a pixel coordinate in a neighborhood of the feature circle as a plane coordinate and taking a corresponding phase value as a height coordinate; carrying out plane fitting on the point set and obtaining a normal vector; the phase normal vector difference of each point is obtained, a frequency statistical graph is generated, curve fitting is carried out, and a self-adaptive segmentation threshold value is obtained according to a preset exponential decay curve; dividing the error curve into three sections by utilizing a self-adaptive segmentation threshold value and endowing the three sections with weights; on the basis of accurate geometric constraint provided by a calibration plate, 'phase normal vector difference 'is introduced as a physical criterion of the quality of feature points, a weight model is constructed through a piecewise function, and adaptive weighting is performed on the feature points with different qualities in multi-view beam adjustment optimization, so that the multi-view beam adjustment optimization is realized. Therefore, the calibration precision of the projector in the trinocular system is obviously improved.