250 results about "Pyramid" patented technology

The invention relates to an OpenGL (Open Graphics Library)-based large-scale three-dimensional model object quick pickup method. The method comprises the following steps: converting an interactive operation of a user into a three-dimensional space structure; a CPU constructs octree space acceleration structures based on a bounding box of an object in a large-scale three-dimensional model, and leaf nodes, intersecting with a pickup ray or a pickup view pyramid, of an axis alignment bounding box in each octree space acceleration structure are screened out to serve as candidate leaf nodes; collecting triangles contained in all candidate leaf nodes, and constructing a candidate triangle set; dynamically allocating thread groups and threads of the GPU based on the number of candidate triangles in the candidate triangle set; and performing highlight display on the candidate triangles intersected with the pickup ray or the pickup view cone, and feeding back IDs of the corresponding candidate triangles. According to the method, the data transmission quantity and the video memory access pressure are remarkably reduced, the pickup response time is greatly shortened in a large-scale three-dimensional model scene, and the requirements of real-time interaction and visual operation can be met.

The invention discloses a three-dimensional scene reconstruction method driven by monocular implicit depth features. The method comprises the following steps: extracting multi-view monocular depth features through a frozen monocular depth encoder; realizing cross-view feature alignment by using a geometry-vision similarity fusion module; generating a multi-scale monocular implicit depth feature through a multi-scale feature adaptive and processing module; a context modeling module and a feature pyramid network are adopted to extract multi-view context features; carrying out one-stage fusion with a monocular implicit depth feature to obtain a fusion feature; generating a depth map through the multi-view stereo depth estimation network; under the constraint of a depth map, generating spatial point features through multi-source feature projection sampling; performing two-stage fusion with the fusion features to generate point-level implicit features; optimizing the features through a frequency-space local-global gating mechanism; respectively inputting the optimization features into a Gaussian sputtering explicit rendering branch and a volume implicit rendering branch; and carrying out weighted fusion on the rendering result to generate a new view image.

The invention relates to the technical field of computer vision, in particular to a complex scene object recognition system based on multi-scale feature fusion, which comprises a feature extraction module, a Riemannian geometric feature pyramid module, a candidate region generation module and an instance recognition and positioning module, and is characterized in that the feature extraction module extracts multi-scale spatial features; the Riemannian geometric feature pyramid module maps spatial features to a Riemannian manifold space, calculates curvature distribution of feature manifolds, dynamically adjusts feature extraction levels and fusion strategies, and realizes geometric consistent fusion of different scale features through geodesic line feature fusion; the candidate region generation module generates an object candidate region based on the fusion features; and the instance identification and positioning module completes final object classification and positioning, introduces a differential geometry theory to construct a Riemannian geometric feature pyramid, realizes geometric representation and adaptive fusion of features, solves the problems of scale change, partial occlusion, background interference and the like in a complex scene, and effectively improves the accuracy and robustness of object identification.

The invention relates to an efficient concrete crack detection method, device and system. The method comprises the following steps: acquiring an image of a to-be-detected concrete beam; generating a hierarchical grid backbone network of redundant features based on linear transformation, and performing feature extraction processing on the to-be-detected concrete beam image to obtain multi-scale feature information; performing feature fusion processing on the multi-scale feature information based on a geometric perception dynamic feature pyramid network to obtain fused feature information; based on task alignment of the dynamic detection head and classification and positioning processing of the fusion feature information, the target crack information corresponding to the to-be-detected concrete beam image is obtained, the concrete crack detection accuracy is improved, the weak texture feature extraction and retention capability is significantly enhanced, the leak detection risk of fine cracks is effectively reduced, and the detection efficiency is improved. The anti-interference capability under a complex background is enhanced, the false detection rate is reduced, mismatching of classification and positioning tasks is avoided, and the crack positioning precision is improved.

The invention discloses a small target detection method based on a bidirectional enhancement mechanism and loss optimization, and the method comprises the steps: replacing a key lower sampling layer with a space depth conversion convolution (SPDConv) module in a backbone network part, so as to maintain the fine-grained features of a small target; an HGBlockPPA module integrating multi-scale pooling and double-path attention is introduced at the tail end of the network, and the expression ability of defect features is enhanced. In the feature fusion stage, a small target enhanced pyramid structure is constructed, a high-resolution shallow-layer feature map P2 is introduced and fused, a bidirectional feature propagation path is established, and feature refining and small target feature response enhancement are realized in combination with a multi-layer feature splicing and re-parameterization convolution module. Meanwhile, a Shape-NWD-SD composite loss function is designed, and a bounding box is modeled as Gaussian distribution. According to the method, the real-time performance of the model is kept, and meanwhile, the detection recall rate and the positioning precision of the small target are remarkably improved.

The invention provides an anti-shake motor, a camera module and electronic equipment, and relates to the technical field of shooting equipment. The anti-shake motor includes a base and a first bracket. The first support is movably connected with the base and can rotate around a first shaft relative to the base. The first axis is parallel to the second direction. The anti-shake motor further comprises a positioning ball. The positioning ball is movably connected between the base and the first support. The base is provided with a positioning groove. The positioning groove comprises three limiting faces. And the three limiting surfaces form three side surfaces of the triangular pyramid. The positioning balls and the positioning grooves are arranged in the second direction, the positioning balls are movably connected into the positioning grooves and make contact with the three limiting faces at the same time, and the first shaft penetrates through the center of the space defined by the three limiting faces. According to the anti-shake motor, positioning of the first shaft is achieved through cooperation of the positioning balls and the positioning grooves, friction force between the first support and the base is small, the first support is not prone to clamping stagnation when moving relative to the base, and the control precision of the anti-shake motor is high.

The invention discloses a novel structure construction method using a gradient triple-period minimal surface lattice. The novel structure construction method comprises the following steps: defining a mathematical relationship of an entity part of a TPMS lattice according to a selected scene; the standard TPMS lattice in the original cube is deformed and matched into the pyramid space, a transition structure is formed, and a final TPMS distortion structure is output; connecting the upper-layer small unit lattices with the lower-layer large unit lattices by the TPMS twisted structure to generate gradient lattices without geometric defects; the gradient between different unit numbers is controlled by controlling the size of the top plane of the pyramid; the size of the transition area is adjusted, the transformation completion distance of the substructure is changed, and then the elastic modulus is adjusted and controlled; the number of layers of the base structure in the transition area is increased, and the buffering distance of structure transformation is prolonged. The problems that in the prior art, a gradient lattice transition area has geometric defects, the adjustable range of elastic modulus is small, and stress is concentrated are solved, and the stability and durability of the structure are improved.

The invention discloses an optical flow estimation method, device, equipment and medium, and the method comprises the steps: extracting the multi-scale features of an image through a pyramid network structure, and obtaining the initial optical flow of each level through feature warping processing and cost body calculation; calculating an uncertainty map of each level of optical flow based on the initial optical flow; calculating the fusion weight of the current layer by adopting the uncertainty map of the current layer and the fused uncertainty map of the previous layer, and carrying out weighted fusion to obtain the fusion optical flow of the current layer; calculating the motion gradient consistency loss based on the optical flow gradient difference of the adjacent pyramid layers; constructing a total loss function based on the endpoint error loss, the uncertainty loss and the motion gradient consistency loss of the initial optical flow; and training the optical flow estimation network by using the total loss function to obtain an optimized optical flow estimation model, and performing optical flow prediction on the to-be-estimated image. According to the invention, the overall precision, robustness and interpretability of optical flow estimation in a complex scene are improved.

The invention discloses a construction method and system of a three-layer inclined quadrangular pyramid grid structure, and relates to the technical field of building steel structure construction. The method comprises the steps of building an intelligent monitoring control system, carrying out BIM collaborative assembly and calibration, carrying out multi-dimensional graded lifting, carrying out adaptive attitude correction, carrying out BIM collaborative in-place recording and the like. The corresponding system comprises an intelligent monitoring control module, a BIM collaborative management module, a hydraulic synchronous lifting execution module, a posture correction module and an unloading in-place module. Through deep integration of a distributed control network and a BIM technology, dynamic monitoring, self-adaptive adjustment and visual management of the construction process can be realized, the accuracy, safety and efficiency of construction of the three-layer inclined quadrangular pyramid grid structure are improved, and the method is suitable for large-span steel structure grid installation engineering.

The invention provides a multi-scale leaf point cloud completion method based on multi-attention mechanism collaboration, and belongs to the technical field of deep learning, and the method comprises the steps: segmenting independent leaf three-dimensional point cloud data according to obtained plant three-dimensional point cloud data, and carrying out the diversity and preprocessing of the leaf three-dimensional point cloud data, taking the preprocessed three-dimensional point cloud data of the blade as the input quantity of a point cloud complementation network, and then extracting the local structure and global semantic features of the incomplete blade point cloud through a multi-resolution encoder integrating a triple attention mechanism and a coordinate attention mechanism; afterwards, a pyramid decoder integrated with a multi-scale attention mechanism realizes progressive complementation of a blade point cloud incomplete part based on multiple resolutions, a discriminator receives point clouds with different resolutions, generation quality is optimized through adversarial training, chamfering distance loss is adopted as a multi-stage reconstruction loss constraint complementation whole process, and the quality of the blade point clouds is improved. The defect that an existing method can only guarantee similar geometrical shapes and cannot achieve detail completion is overcome.

The invention discloses a stainless steel pipe fitting weld defect positioning method based on deep learning. The method comprises the following steps: acquiring original image data of a stainless steel pipe fitting weld; inputting the input image tensor into an improved YOLOv10-S model in the collaborative double-branch structure, and generating a multi-scale feature map; setting a deformable context aggregation block, and outputting an enhanced feature map; generating a fusion feature set by adopting a path aggregation and feature pyramid fusion structure; setting a multi-scale confidence distillation head, and outputting a final bounding box coordinate; and inputting the input image tensor into the OfficientAD model to generate an abnormal thermodynamic diagram, carrying out space alignment and confidence fusion on the bounding box coordinates and the abnormal thermodynamic diagram, and outputting the category, position and size information of the weld defect. According to the method, high-precision positioning and multi-scale abnormal fusion identification of the weld defects of the stainless steel pipe fitting are realized.

The invention relates to the technical field of supply chain product quality inspection, in particular to an AI model-based supply chain intelligent quality inspection and defect identification system, which comprises a texture feature adaptive unit, a defect identification dynamic adjustment unit, an AI defect identification unit and a control output unit, the texture feature adaptive unit extracts texture density, contrast and edge gradient direction distribution of a chip resistor grey-scale map in real time, the texture density, the contrast and the edge gradient direction distribution are dynamically compared with a historical batch feature library to generate texture feature difference mapping, the defect identification dynamic adjustment unit nonlinearly adjusts a gradient deviation threshold value through an S-shaped response function according to the difference mapping, and the defect identification accuracy is improved. The AI defect identification unit adopts a convolutional neural network, outputs a defect classification result through parallel branch processing, spatial pyramid pooling and secondary verification, controls the output unit to transmit the result to an execution mechanism, records batch switching parameters, solves the problem of false and missing judgment caused by batch texture fluctuation, and improves the detection accuracy. And the quality inspection precision is improved.

The utility model relates to a kind of drilling device for super-long graphite rod processing, including tool tip and the tool holder connected with tool tip. Tool tip is set to three-prism pyramid shape. Tool tip includes: first cutting edge, at least two first cutting edges are symmetrically arranged, and the edge of at least two first cutting edges forms first angle;Second cutting edge, at least four second cutting edges are axially symmetrically arranged, and the plane of at least two second cutting edges forms second angle with the plane of opposite at least two second cutting edges. Chip removal platform formed due to first angle and second angle not being identical is arranged between tool tip and tool holder. The utility model is set to three-prism pyramid shape tool tip, and adopts multi-angle cutting edge (with symmetrically arranged first cutting edge and axially symmetrically arranged second cutting edge), so that tool is more uniform in cutting process Stress, can effectively reduce cutting force and cutting torque, can further reduce cutting resistance and vibration, improve cutting efficiency and machining precision.

The utility model discloses a triangular-cone-shaped high-temperature-resistant and hydrogen-sulfide-resistant screw pump sealing impact device, the main body structure of the triangular-cone-shaped high-temperature-resistant and hydrogen-sulfide-resistant screw pump sealing impact device is a lower sealing occlusion ring and an upper sealing occlusion ring which are occluded up and down, and the triangular-cone-shaped high-temperature-resistant and hydrogen-sulfide-resistant screw pump sealing impact device comprises the lower sealing occlusion ring and the upper sealing occlusion ring, the bottom face of the upper sealing meshing ring is provided with a triangular cone-shaped meshing groove, and the material has the advantages of being good in high temperature resistance, hydrogen sulfide resistance and abrasion resistance, high in strength and good in toughness. Through the arrangement of the lower sealing occlusion ring and the upper sealing occlusion ring, close fit of the lower sealing occlusion ring and the upper sealing occlusion ring can be achieved, the occlusion face is of the regular triangular cone type three-dimensional structure distributed in the radial direction, the structure has the good impact resistance and anti-scouring characteristics, the overall structure is simple in process, convenient to install and long in service life, and the service life is long. The stable sealing performance of the sealing device in a high-temperature and hydrogen sulfide environment is ensured, and medium leakage is prevented.

The invention discloses an intelligent wearable watch case precision detection method based on visual inspection, and relates to the technical field of machine vision, and the method comprises the steps: collecting a watch surface point cloud and a watch case image, carrying out the point cloud denoising and image distortion correction, and obtaining registration multi-modal data; performing multi-layer pyramid convolution on the registered multi-modal data, and outputting geometric-texture features of the watchcase; curvature gradient calculation and spatial relation modeling are carried out on the geometric-texture features of the watchcase, and curvature distribution parameters and a spatial relation adjacency matrix are obtained; the curvature distribution parameters are used as nodes of the topological structure, the spatial relation adjacent matrix is used as edges of the topological structure, and a case geometrical relation graph is constructed. According to the method, the watch surface point cloud and the watchcase image are fused and subjected to multi-layer pyramid convolution, the comprehensiveness and positioning precision of watchcase defect recognition are improved, and meanwhile, the engineering practicability of the detection process is effectively improved by utilizing a graph attention defect prediction model and an SPC statistical process control algorithm.

The invention relates to the technical field of construction engineering and artificial intelligence, in particular to a high-strain test tower state monitoring method, which comprises the following steps: acquiring a high-strain test tower image; based on the high-strain test tower image, performing instance segmentation on the high-strain test tower structure by adopting deep learning to obtain an image after instance segmentation; based on the image obtained after instance segmentation, on a tower structure, a brightness normalization method based on a long-short-term memory mechanism is adopted to normalize the surface brightness of the tower, and an image obtained after brightness normalization is obtained; and acquiring an image sequence after brightness normalization, performing optical flow feature extraction by using a pyramid optical flow method based on the image sequence after brightness normalization, and monitoring the state of the high-strain test tower based on the extracted optical flow features. According to the invention, the external environment can be greatly reduced, and meanwhile, high-strain test tower state monitoring can be accurately realized.

The present application relates to the technical field of stacking system, specifically to a multi-variety bending finished product flexible stacking system, comprising: a visual recognition module configured to scan the plate to be stacked by a 3D laser sensor, generate point cloud data and extract the following parameters: the size set of the flat plate, wherein L is the length, W is the width, and H is the thickness; the curvature distribution and the gravity center coordinates of the bending piece; a dynamic stacking type planning module connected to the visual recognition module and configured to: calculate the interlayer offset based on the pyramid stacking constraint, wherein H is the current height, and the safety inclination angle is ≤15°. The plate combination is calculated in real time based on the greedy backtracking algorithm, the space utilization rate is improved to 92% (the traditional system ≤75%), the small plate filling response time is ≤0.5 seconds through the temporary storage area scheduling mechanism, and the gap waste is reduced.

A screed plate for a screed assembly to be towed by a mobile paver has a lower textured surface including a plurality of protruding elements arranged in a gradation pattern. The gradation pattern is characterized by an increasing number or density of protruding elements between a forward leading edge and a rearward trailing edge of the screed plate. The gradation pattern is also characterized by a decreasing size of the protruding elements between the forward leading edge and the rearward trailing edge. The protruding elements may be shaped as oblong rhombic pyramids defining channeling grooves there between.

The present application relates to the technical field of structural safety monitoring, and specifically relates to a multi-antenna Beidou monitoring device and a monitoring method, wherein the monitoring device comprises a mounting base, a monitoring pile, a mounting cabinet, a first positioning antenna, a second positioning antenna and a third positioning antenna, the mounting base is used for being fixed to a measured structure, the bottom end of the monitoring pile is fixed to the mounting base, the connecting line of the first positioning antenna, the second positioning antenna and the third positioning antenna forms a triangle, the connecting line of the centroid of the first positioning antenna, the second positioning antenna, the third positioning antenna and the mounting base forms a triangular pyramid, and the mounting cabinet is mounted on the monitoring pile. The present application solves the technical problem that the monitoring pile in the prior art causes abnormal error of Beidou positioning under the condition of self-space displacement interference, thereby causing an error monitoring result.

The invention belongs to the technical field of three-dimensional modeling, and particularly relates to a novel multi-detail-level terrain rendering method, which comprises the following steps: S1, inputting terrain data; s2, calculating a vertex, a normal and a texture coordinate array, and transmitting into a rendering pipeline; s3, calculating an LOD maximum value, a resolution ratio scale factor and a sight distance switching factor; s4, setting basic attributes of root nodes of the quadtree, and completing construction of terrain geometry; s5, rendering terrain geometry in real time; the method has the advantages that by setting basic attributes of root nodes of the quadtree, a topographic data structure with the quadtree as an organization form is constructed; a topographic organization mode of'single-source geometric data + multi-level index rules' is established, and a mode of'repeatedly storing complete geometric data in each level of LOD 'in a traditional tile pyramid is converted into a mode of only storing a vertex, a normal and a texture coordinate array; the memory pressure in a high-precision terrain scene can be reduced, and the terrain real-time rendering efficiency can be improved.

The application discloses a multi-dimensional demonstration method and device based on time domain technology and a storage medium, and relates to the technical field of data interaction. The application first acquires a time domain file containing multiple perspectives and / or multiple time points, and encapsulates multi-dimensional index information containing a space index and a time index. Then, a demonstration narrative stream including a key frame sequence of the demonstration and a camera motion path of the key frame is constructed based on the multi-dimensional index information. In the process of playing and rendering, the current camera state is acquired in real time, the prediction information of the camera motion path is combined, and the target display area in the current view cone is calculated. Finally, the corresponding target pyramid tile data is called from the multi-dimensional index information, and progressive decoding and rendering are performed in the order from low to high resolution, so as to realize the non-susceptible loading of the demonstration scene from the overall overview to the local micro details. In this way, the loading efficiency of the multi-dimensional demonstration scene is improved, and the resource consumption is reduced.

System, methods, and other embodiments described herein relate to single-shot multi-object three-dimensional (3D) shape reconstruction and categorical six-dimensional (6D) pose and size estimation. In one embodiment, a method includes inferring a heatmap based upon a feature pyramid, where the feature pyramid is generated based upon a red green blue depth (RGB-D) image that includes objects. The method further includes sampling a 3D parameter map at locations corresponding to peaks in the heatmap, where the 3D parameter map is inferred based upon the feature pyramid, and where the locations include latent shape codes, 6D poses, and one-dimensional (1D) scales. The method further includes generating point clouds based upon the latent shape codes, the 6D poses, and the 1D scales.

The invention discloses a multi-scale parallax calculation module, method and device based on Haar wavelet transform and a medium, and the method comprises the steps: firstly constructing a feature pyramid through the Haar wavelet transform, extracting high-frequency and low-frequency details, and then carrying out the parallax calculation through employing a Coarse-to-Fine strategy. According to the method, the problem that random initialization is slow in convergence in a weak texture area or falls into local optimum can be effectively solved, and meanwhile, the wavelet detail component can provide richer texture and edge information for cost calculation. The method has the advantages of being high in precision, high in robustness and high in speed in dealing with parallax calculation tasks of weak-texture, non-texture or repeated-texture scenes.

The invention discloses an indoor scene three-dimensional reconstruction method and system based on feature pyramid fusion AdaVGGT and CNN features. Obtaining an indoor scene three-dimensional image, and processing the image to obtain a plurality of local segments; inputting each local segment into a CNN branch and an AdaVGGT branch to obtain a multi-scale local feature and a global feature, and fusing the multi-scale local feature and the global feature to generate a multi-scale fusion feature; the multi-scale fusion features are respectively projected into three-dimensional voxel grids, a GRU network is used for fusion, and then MLP prediction is used to obtain three-dimensional reconstruction local segments; and generating a complete dense three-dimensional surface model according to each dense three-dimensional surface model local segment, and taking the complete dense three-dimensional surface model as a reconstructed three-dimensional scene. According to the method, by introducing an adapter and an extended feature fusion mechanism, the reconstruction consistency of a large-scale scene structure and a shielding region is improved, and the recovery quality of geometric boundaries and texture details in a complex scene is improved.

The invention discloses a ship fluid computational domain full-automatic non-structural hybrid grid generation method, which comprises the following steps: acquiring a ship geometric model, adding a computational domain bounding box, and generating an initial discrete geometric model consisting of triangular surface grids; acquiring user input control parameters, and generating a size field in combination with the initial discrete geometric model; embedding a symmetric plane and a horizontal plane according to the initial discrete geometric model, generating an initial triangular surface grid with consistent boundaries, and performing re-gridding under the control of a size field to generate a model boundary surface grid; according to a horizontal plane, dividing the boundary surface grid into an upper surface grid and a lower surface grid, and respectively generating an upper boundary layer grid and a lower boundary layer grid which are composed of anisotropic triangular prism units and pyramid units; on this basis, surface grids of the boundary of the computational domain are regenerated at the boundary of the bounding box, tetrahedral grid units are generated under the control of the size field, and therefore the ship fluid computational domain full-automatic non-structural hybrid grid is formed through merging.

The invention discloses a full-slice image cancer prediction and subtype classification method, system and device, and relates to the technical field of image processing and medical artificial intelligence. Comprising the following steps: preprocessing a full-slice image, cutting the full-slice image into image blocks with position coordinates, and extracting features; reconstructing the feature sequence into a two-dimensional feature map which retains the original spatial topology through a spatial recovery module; scanning and fusing along eight directions including a horizontal direction, a vertical direction and a plurality of diagonal lines by using a hyper-cross scanning module so as to capture multi-direction local space correlation; multi-scale global features are extracted and fused by adopting convolution layers with different expansion rates through a pyramid module; and finally, outputting a prediction result and a subtype label through a customized classifier, and generating a focus attention heat map. Through the architecture of spatial reconstruction-multidirectional scanning-multi-scale fusion, while the linear calculation complexity of O (n) is kept, the small focus recognition capability and classification precision are remarkably improved, and an efficient and reliable technical scheme is provided for digital pathological diagnosis.

A method and system for non-invasive archaeological investigation of pyramid structures are disclosed. The method combines high-resolution three-dimensional seismic acquisition with Marchenko-based wavefield reconstruction and multiple-reflection imaging to detect and characterize subsurface features such as voids, tunnels, and hidden chambers beneath and around pyramids. Electromagnetic vibrators, generate compressional and shear waves without ground penetration, and three-component (3C) receiver nodes, optionally equipped with over-under muon detectors, record seismic and muon flux data simultaneously. The Marchenko method enables reconstruction of virtual internal sources and receivers, while multiple-reflection imaging provides information from deeper layers below the pyramid base. The invention offers a fully non-invasive, high-resolution, and reversible approach suitable for cultural heritage preservation.

The invention discloses a step-by-step marginalized point-line visual inertia SLAM (Simultaneous Localization and Mapping) method and system. Firstly, an environment image and IMU data are obtained and preprocessed; point-line features in the image are extracted, line features with the length larger than the minimum line segment length threshold value in all layers of an image pyramid are reserved, and the direction included angle and the center point Euclidean distance of any two reserved line features are calculated; if the direction included angle of the two line features is smaller than or equal to an angle threshold value and the center point Euclidean distance is smaller than or equal to a distance threshold value, combining the two line features into a new line feature; then, aligning the visual inertial coordinate system, and estimating an initial state; and finally, performing local visual inertia optimization and mapping by using a sliding window based on a step-by-step marginalization strategy to obtain a prior error. And meanwhile, loopback detection is carried out, and loopback errors and prior errors are added into the next round of sliding window optimization to carry out tight coupling state estimation. The method has relatively high robustness and consistency, and meanwhile, the real-time performance is ensured.