108 results about "Depth mapping" patented technology

The invention discloses a naked-eye 3D display optimization method based on real-time eyeball tracking, and particularly relates to the technical field of naked-eye 3D display, and the method comprises the steps: capturing eyeball movement data in real time through a visual angle tracking sensor, and collecting illumination information in combination with an ambient light sensor; depth perception parameters are calculated based on pupil diameter variation and frequency, and a time sequence prediction type dynamic compensation coefficient is generated in combination with eyeball movement acceleration; acquiring an initial fixation point coordinate by using an improved spherical projection mapping model, and performing compensation coefficient correction to obtain a real-time coordinate; and finally, according to the real-time coordinates, dynamically adjusting the refractive index distribution of the nanostructure layer, the rotation angle of the polarizer, the focal length of the optical lens and other optical modulation parameters. The real-time distance is calculated through the binocular parallax algorithm, the depth mapping value is generated by combining the focal length of the camera and the baseline distance, the method can adapt to the illumination change and the user view angle, and the 3D display effect is optimized.

The invention belongs to the technical field of three-dimensional scene understanding, and particularly relates to a training method and device for a three-dimensional open vocabulary semantic segmentation model. The training method comprises the steps of obtaining multi-view RGB-D images of a target area, performing multi-stage reasoning on each image through a visual language model, generating a target vocabulary list, prompting a two-dimensional segmentation model to establish a pixel-level text label, performing depth mapping on the images to generate a first point cloud, and generating a second point cloud; mapping the text tag to the first point cloud to generate a point-by-point text tag; pre-training a neural network model with a sparse encoder-decoder structure by taking the point-by-point text label as a supervision signal, and generating a three-dimensional segmentation model on the first point cloud; and for the second point cloud of the complete scene of the target area, matching point feature embedding and text embedding with the highest similarity in the shared vision-language feature space, generating a credible point-text tag pair, and finely adjusting the three-dimensional segmentation model based on the credible point-text tag pair.

The invention provides a method for realizing carbon fiber defect detection by using image recognition, which comprises the following steps: acquiring a depth-diameter ratio of a hole according to an angle change correction coefficient matrix, and if the depth-diameter ratio is greater than a preset threshold, introducing an angle-related attenuation parameter to obtain a corrected light intensity depth mapping relationship; constructing three-dimensional shape data of the hole through the corrected light intensity depth mapping relation, identifying an internal space structure of the hole, performing voxelization discretization processing, obtaining a voxel unit and a light intensity response characteristic thereof, and determining an internal geometric boundary of the hole; and establishing a hole defect quality evaluation index according to a volume measurement result and measurement uncertainty evaluation, comprehensively evaluating the structural integrity of the carbon fiber composite material in combination with grid quality evaluation, and establishing a carbon fiber defect grade classification standard.

The invention relates to the technical field of image recognition, in particular to a construction site-oriented reinforcing steel bar binding normativity visual detection method, which comprises the following steps of: driving a camera lens to shoot a focus stack image sequence covering the depth of field before and after a single binding node, extracting a linear contour, and obtaining a multi-dimensional image set of a binding area. According to the method, the focus stack image sequence covering the depth of field before and after the single binding node is shot by the driving camera, and the mapping of the line segment coordinates and the depth of the reinforcing steel bar is established, so that the problems of mutual shielding and depth confusion caused by staggered reinforcing steel bars in the construction site are solved; on the basis, a steel bar binding topology network diagram describing the connection relation between steel bar intersection points is constructed, discrete steel bar line segments and intersection point information are integrated into an overall network with a global structure relation, detection is not limited to isolated nodes any more, and evaluation can be conducted from systematicness and continuity of a whole steel bar framework.

The invention discloses a multimedia content preloading method based on vehicle cloud collaboration, and particularly relates to the technical field of multimedia content preloading, and the method comprises the steps: carrying out the modeling through a multi-dimensional space-time driving path, and combining with historical road communication coverage features; a network availability prediction result and a user historical consumption mode are deeply mapped, and a weight fusion mechanism of content timeliness, capacity characteristics and scene correlation is introduced, so that a generated multi-level priority sequence better meets the instant requirements of vehicles at different time and different positions; through dynamic comparison of an initial plan, a real-time path, a network and user interaction data, a superposition out-of-control situation of prediction deviation can be rapidly identified, a scheduling priority offset mode under multiple tasks is identified in combination with vehicle calculation, caching and bandwidth occupation states, and then a pre-loading dislocation cooperative imbalance fault network is constructed through bidirectional correlation analysis. And visual diagnosis and quantitative evaluation of the imbalance state of the prediction layer and the execution layer are realized.

The invention relates to the technical field of artificial intelligence, provides a multi-language task execution method, device and equipment and a medium, is applied to financial and medical health care service scenes, and can acquire text data and visual action data of multiple languages according to an execution instruction and perform preprocessing to realize standardized processing of multi-modal data; constructing a cross-language word vector semantic space based on adversarial training according to the multi-language features to realize preliminary word vector alignment; performing multi-language semantic alignment on the multi-language features according to a cross-language word vector semantic space, further breaking language barriers, and realizing depth mapping and alignment among different language semantics; a multi-language culture knowledge graph is constructed, so that culture knowledge is introduced, and the culture perception ability is improved; fusion is carried out through a gating fusion mechanism, fusion features with language attributes and cultural attributes can be obtained, and therefore multi-language tasks can be executed more accurately by integrating multi-language information and cultural knowledge.

The invention discloses a fracture pressure prediction and internal friction angle joint inversion method for a heterogeneous stratum. The invention belongs to the technical field of petroleum engineering and artificial intelligence cross application, well depth is used as an independent variable, the minimum / maximum horizontal stress, pore pressure and cohesion are combined to construct well depth mixed input, only multi-scale periodic basis function transformation is carried out on the well depth, and a linear well depth channel is reserved; establishing a rupture pressure depth mapping model, introducing an augmented Lagrange physical constraint containing dual variables and a dynamic threshold, adaptively adjusting a physical weight according to a physical residual error and a threshold difference value, realizing dynamic balance of a rupture criterion and data fitting, and searching and optimizing a model topology by using a multi-target network structure; under a limited sample, a fracture pressure profile and a physically reasonable internal friction angle can be obtained at the same time, and the prediction precision and the physical consistency are improved.

The invention relates to an intelligent automobile environment sensing system, in particular to an intelligent automobile robust aerial view environment sensing system based on depth mapping and KAN-Mama fusion. Comprising an image modal semantic coding and hierarchical fusion module, a laser radar point cloud sparse coding and BEV feature construction module, a local self-attention depth feature enhancement module, a CMSM gating interaction module, a KAN high-order nonlinear fusion module, a sensor adaptive disturbance modeling module and a multi-task perception decoding module. The excessive dependence of the existing BEV environment sensing system on the ideal calibration condition of the sensor is avoided, the sensing performance can still be ensured when the sensor has a slight alignment error, and meanwhile, the defect of the self-adaptive modeling capability of the local area of the existing BEV environment sensing system is overcome; the non-ideal calibration robustness and the local area modeling robustness of the intelligent automobile BEV environment sensing system are remarkably improved, and a basic interaction architecture with high information coupling degree and strong feature expression is constructed for the intelligent automobile sensing system.

The invention discloses a naked-eye 3D augmented reality interactive display system, and relates to the technical field of naked-eye 3D display and reality interaction. According to the invention, the multi-source depth data acquisition module acquires and fuses a user limb depth image, the parallax physical depth mapping module completes depth and parallax mapping and visual angle correction, the occlusion relation determination module divides regions to determine the occlusion type, and the adaptive mask layer generation module generates a dynamic smooth mask layer. The shielding effect fusion display module is used for overlapping and synthesizing the shielding layer and the multi-view virtual image and then displaying the overlapped and synthesized shielding layer and the multi-view virtual image; the problems that in naked eye 3D interaction, the occlusion relation between the limbs of the user and the virtual object is not accurate, and the vision is abrupt are effectively solved, the space consistency and the vision fluency of interaction are remarkably improved, and the real and natural naked eye 3D augmented reality interaction experience is achieved.

The invention discloses a sit-and-reach measurement method and system based on machine vision, and the method comprises the steps: obtaining a dynamic image of a measurement object, correcting the dynamic image, and obtaining a corrected image; performing image optimization on the corrected image to obtain an optimized image, and inputting the optimized image into a pre-trained attitude estimation network to obtain a key point sequence; determining a time continuity parameter according to the key point sequence, the time sequence characteristics of the key point sequence, the inter-frame time interval and the confidence coefficient of the coordinates of the feature points; aiming at the key point sequence, determining key point track data containing a time continuity parameter; according to the key point trajectory data, obtaining a distance error from the pixel to an actual distance; if the distance error is smaller than a preset threshold value, motion trail data are obtained according to the depth mapping coefficient and the focal length parameter value; and based on the time sequence data of the historical trajectory data, calibrating the motion trajectory data, and determining a displacement measurement result. The method is high in precision and good in robustness.

The subject disclosure provides for systems and methods for generating a dense depth map. The method may include determining a first region of interest via input from a first sensor device configured to capture supplemental pixel images. The method may include determining a second region of interest via input from a second sensor device configured to capture supplemental pixel images. The method may include receiving depth measurements from a third sensor device configured to capture depth measurements. The method may include determining a plurality of depth maps from the pixel images from the first sensor device, the supplemental pixel images from the second sensor device, and the depth measurements from the third sensor device. The method may include generating from the plurality of depth maps a foveated region comprising a high-resolution image and a low-resolution image oriented on the periphery of the foveated region.

The invention belongs to the technical field of computer vision, and particularly relates to a depth estimation method and system for a pond monocular image, and the method comprises the steps: constructing a pond monocular image depth estimation model, inputting a pond image into the pond monocular image depth estimation model, and obtaining the depth mapping of a pond scene, distance information from each pixel point to the camera is obtained, a pond image depth map is generated based on the distance information, and depth estimation of the pond image is achieved. The system comprises a pond monocular image depth estimation model, wherein the pond monocular image depth estimation model comprises a U-MonoVIT encoder, a local plane guide layer and a jumping multi-scale extension self-attention mechanism. The method can effectively solve the problem that the depth estimation accuracy is reduced due to the problems of fuzzy pond images, low contrast ratio, overexposure or darkness and the like, can accurately and efficiently achieve monocular depth estimation of pond scenes, and provides key support for comprehensive exploration of pond environments and biomass estimation.

The invention discloses an industrial flexible object detection method based on machine vision, and relates to the technical field of industrial detection, and the method comprises the steps: combining a texture feature vector with an extracted geometric shape feature vector to generate a visual feature vector, extracting a global touch feature based on a pre-trained ViT-B / 16 model, and carrying out the detection of an industrial flexible object. The method comprises the following steps: extracting position and speed feature vectors through a CSPDarknet network architecture based on a radar mask pattern, and generating a comprehensive feature vector through a Transform model and a modal weight; a radar mask pattern is efficiently generated through a depth mapping conversion network, multi-modal features are extracted, a comprehensive feature vector is generated through model adaptive fusion, model parameters are dynamically adjusted through smooth features and online adaptive optimization, the calculation complexity and the marking cost are reduced, the robustness to illumination change and deformation scenes is improved, and the method is suitable for large-scale popularization and application. And the detection precision of the bounding box, the material classification and the deformation state probability of the flexible object is improved.

The invention relates to the technical field of surface engineering measurement, in particular to an image-driven super-hydrophobic coating uniformity evaluation method, which comprises the following steps of: firstly, acquiring a color image of a coating to be measured under one-way natural light and vertical imaging conditions, and graying; dividing the image into a plurality of image units according to a preset equidistant rectangular grid; linear gray-depth mapping is obtained through primary illumination calibration of the reference calibration plate, and the gray value of each unit is converted into a surface microstructure depth value; counting an average value and a standard deviation of all depth values, and calculating a variation coefficient to represent overall uniformity; further dispersing the variable coefficient according to a box line-quantile strategy and calculating the Shannon entropy, and verifying the fluctuation distribution balance; the method does not need multiple light sources and contact scanning, has the advantages of simplicity and convenience in operation, rapidness in measurement and full-breadth quantification, and can be used for monitoring the preparation process of the super-hydrophobic coating and inspecting the service quality.

The application discloses a power transmission line sag identification and early warning method and device combined with zoom ranging, relates to the technical field of power transmission line monitoring, and comprises the following steps: installing a binocular camera system in a target identification scene, adjusting variable focus lens parameters, calibrating the binocular camera system by using a checkerboard, and constructing a disparity-depth mapping relationship; collecting Q times of synchronous images by using the calibrated system, performing multi-point disparity analysis, obtaining disparity analysis results of a first suspension point, a second suspension point and a sag lowest point, and screening and determining target suspension point and sag lowest point analysis results; taking the results as indexes, searching the disparity-depth mapping relationship, and obtaining depth information; calculating target suspension point ranging results according to the depth information, calculating a sag value and judging whether the sag value exceeds a threshold value, and if not, triggering early warning, so as to solve the technical problem that a traditional ranging mode cannot feed back sag changes in real time, and thus cannot effectively early warn safety hazards caused by excessively large sag.

The invention relates to the technical field of accumulated snow depth downscaling, in particular to an accumulated snow depth downscaling method combining ResNet and an accumulated snow attenuation curve. Acquiring accumulated snow data of the target area and preprocessing the accumulated snow data to obtain a first accumulated snow depth image, a first auxiliary depth image and a second auxiliary depth image; inputting the first snow depth image and the first auxiliary depth image into a model for training, and obtaining a first snow prediction image and a second snow prediction image by using the trained model; obtaining a fine-scale residual value of the first snow depth image and the first snow prediction image, and correcting the second snow prediction image to obtain downscaling snow depth data; performing secondary correction on the downscaling accumulated snow depth data by using the accumulated snow attenuation curve to obtain optimal accumulated snow depth data; and performing accumulated snow depth mapping according to the optimal accumulated snow depth data. According to the invention, accurate snow depth downscaling can be carried out on snow depth data of a large-scale area, and efficient snow depth prediction is realized.

The invention discloses a winding imaging method and system based on SAFT and multi-transducer arc scanning, and belongs to the technical field of ultrasonic signal scanning imaging, and the method comprises the steps: installing a plurality of ultrasonic transducers on a three-dimensional mechanical track; synchronously driving a transducer to emit arc-shaped ultrasonic waves and receive echo signals, and generating a plurality of groups of original arc data sets; executing space intersection operation, and extracting three-dimensional coordinates of the surface of the winding at a single moment; performing envelope extraction and SAFT delay superposition focusing processing on the echo signal to generate a multi-moment enhanced signal; constructing a space-time fused three-dimensional data matrix; based on the coordinate deviation of a preset reference point on the surface of the winding, constructing a dynamic correction model to dynamically correct the three-dimensional data matrix; and executing spatial depth mapping on the corrected three-dimensional data matrix to generate a pseudo-color three-dimensional point cloud picture of the transformer winding. According to the invention, a multi-transducer arc scanning and SAFT signal focusing cooperation mechanism is adopted, and efficient three-dimensional imaging of winding deformation is realized.

A depth map generation method and a depth map generation device (10) are provided. The depth map generation method comprises: obtaining (S201) an RGB color image by a monocular camera provided in a robot system, obtaining (S202) a 3D point cloud by a LiDAR sensor provided in the robot system, generating (S203) a sparse depth map, which contains depth information only for some points in a given space, from the 3D point cloud, inputting (S204) the RGB color image and the sparse depth map into a pre-trained diffusion model, and generating (S205) a dense depth map with depth information for all points in the given space.

A depth camera assembly (DCA) determines distances between the DCA and objects in a local area within a field of view of the DCA. The DCA includes an illumination source that projects a known spatial pattern modulated with a temporal carrier signal into the local area. An imaging device capture the modulated partem projected into the local area. The imaging device includes a detector that comprises different pixel groups that are each activated to captured light at different times. Hence, different pixel groups capture different phases of the temporally modulated pattern from the local area. The DCA determines times for light from the illumination source to be reflected and captured by the imaging device from the phases captured by the different pixel groups and also determines distances between the DCA and objects in the local area based on deformation of the spatial partem captured by the imaging device.

A method for depth mapping includes providing a depth mapping device comprising a projector, which is configured to project a pattern of optical radiation onto a target area over a first field of view about a projection axis, and a camera, which is configured to capture images of the target area within a second field of view, narrower than the first field of view, about a camera axis, which is offset transversely relative to the projection axis. The projector projects the pattern onto first and second planes at first and second distances from the camera, and the camera captures first and second reference images containing first and second parts of the pattern on the first and second planes, respectively. The first and second reference images are combined to produce an extended reference image including both the first and second parts of the pattern.

The present application relates to a method for predicting pre-drilling stratigraphic profile based on multi-algorithm interpolation of well logging curves, comprising: predicting the top / bottom and thickness of each layer of the target well based on stratified data; aligning the normalized depth within the layer; predicting multiple well logging curves by using multi-algorithm (at least IDW and Kriging) interpolation; correcting the proximity and mapping the depth based on the well deviation trajectory; outputting the multi-depth reference predicted curve and the stratigraphic position QC result. Further comprising lithology profile prediction: obtaining the classification probability by using stratified supervised classification, and fusing the spatial prior probability obtained based on the lithology of adjacent wells and distance weight to output the lithology interval of the target well and the profile comparison chart. This method comprehensively utilizes the well logging curves of adjacent drilled wells, stratigraphic position stratified data, well location and well deviation trajectory information, and under the unified depth reference, performs multi-algorithm horizontal interpolation prediction on multiple well logging curves of the target well, and further combines supervised learning and spatial prior to realize lithology profile prediction, and outputs the stratigraphic profile results which can be used for pre-drilling decision and well location deployment.

This invention presents a heuristic recovery mechanism for malicious shortcut file parsing and threat detection. It employs a resilience-oriented parsing engine, utilizes a dynamic byte ordering mechanism to adapt to underlying data, leverages a sliding window algorithm to accurately locate variable-length field terminators to extract hidden strings, and employs a recursive parsing strategy to deeply traverse undocumented nested structures and extension blocks. Parsing anomalies are treated as threat features and archived during the parsing process, and an iterative extraction strategy with adaptive parameters ensures full data acquisition. Based on the comprehensive parsing results, a three-dimensional threat model encompassing deception, evasion, and execution is constructed, achieving a deep mapping of malicious shortcut files from binary structure to logical behavior. This invention not only significantly improves the parsing success rate of variant samples but also effectively extracts fingerprint features from automatically generated samples, providing key technical support for modern shortcut-based attack and defense countermeasures.

The invention provides an analysis method and system for electron re-collision energy in a non-uniform laser field, and the method comprises the steps: obtaining a field intensity peak value and a space gradient as input parameters, constructing an electron motion equation frame based on a classical re-collision theory, solving and calculating return kinetic energy through employing a perturbation method, calculating output characteristics based on energy-momentum conservation, and obtaining an electron re-collision energy model. The method comprises the steps of obtaining multiple groups of data by repeating more than multiple rounds of operation so as to construct a sample database, and constructing a physical heuristic deep mapping network based on the sample database so as to obtain corrected output characteristics. An efficient physical approximation model is constructed by introducing a perturbation method, and a physical heuristic deep mapping network of physical constraints is constructed according to the efficient physical approximation model, so that a novel analysis method with both physical consistency and calculation efficiency is constructed.

The invention discloses an active power failure monitoring method and system based on machine learning. The method comprises the steps of data acquisition, data processing, establishment of an active power failure monitoring model, adaptive optimization and active power failure monitoring. The invention relates to the technical field of power failure monitoring, in particular to a power failure active monitoring method and system based on machine learning, according to the scheme, noise data is removed by calculating a deviation item, constructing an entropy supplementing item and calculating noise intensity, and the data quality and the accuracy of a monitoring result are improved; the model is established by constructing a depth mapping function, designing a depth feature learning layer and constructing a loss function, so that the prediction performance, precision and stability of the model are enhanced; parameter search is carried out by designing an efficiency coordination factor, a direction transformation function and a noise reduction function, the search process is guided to evolve towards a more efficient direction, the search efficiency is improved, the diversity in the search process is increased, and jumping out of a local optimal solution and exploration of a global better solution are facilitated.

Described is a system for counting stacked items using image analysis. In one implementation, an image of an inventory location with stacked items is obtained and processed to determine the number of items stacked at the inventory location. In some instances, the item closest to the camera that obtains the image may be the only item viewable in the image. Using image analysis, such as depth mapping or Histogram of Oriented Gradients (HOG) algorithms, the distance of the item from the camera and the shelf of the inventory location can be determined. Using this information, and known dimension information for the item, a count of the number of items stacked at an inventory location may be determined.

The present invention discloses a security testing method and system for Face ID identity authentication, which belongs to the field of information security technology. The reference scattered image with different depth information generated by the original depth-sensing camera of the device under test is collected; the reference scattered image is preprocessed; according to the preprocessed reference scattered image with different depth information, a "depth information-infrared scattered displacement" mapping relationship is constructed based on the peak signal-to-noise ratio, and a depth mapping modulation function is obtained by polynomial fitting; the target depth image is modulated into the preprocessed reference scattered image with maximum depth information, and the pixel points in the infrared scattered image are reset to a preset size using a point spread function; the infrared scattered image to be tested is projected so that the original depth-sensing camera of the device under test obtains a false target depth image, thereby realizing security testing. The present invention can interfere with the imaging of the original depth-sensing camera, simulate non-existent depth information, and has high test accuracy.

Described herein are machine learning based systems, methods, and non-transitory computer readable media for determining three-dimensional (3D) information of structures of a patterned substrate. The 3D information can be determined using a neural network to convert a two-dimensional image to a 3D image. In a method, the neural network is trained by supplying varying parameters to simulated images of a sample to model a real manufactured IC structure. The trained network generates a depth map from a newly supplied image of the sample to predict defect locations. The depth map can be converted to a binary map to predict defect size and location, which can be used to guide inspection of the sample using an inspection tool.

The subject disclosure provides for systems and methods for generating a dense depth map. The method may include determining a first region of interest via input from a first sensor device configured to capture supplemental pixel images. The method may include determining a second region of interest via input from a second sensor device configured to capture supplemental pixel images. The method may include receiving depth measurements from a third sensor device configured to capture depth measurements. The method may include determining a plurality of depth maps from the pixel images from the first sensor device, the supplemental pixel images from the second sensor device, and the depth measurements from the third sensor device. The method may include generating from the plurality of depth maps a foveated region comprising a high-resolution image and a low-resolution image oriented on the periphery of the foveated region.

An embodiment includes encoding a first 32-bit floating point depth value, in a plurality of 32-bit floating point depth values, into a first signed 8-bit integer and a second signed 8-bit integer. An embodiment includes predicting, using a trained convolutional neural network (CNN), the first signed 8-bit integer, and the second signed 8-bit integer, a plurality of depth layer maps and a probability map. An embodiment includes generating, by selecting pixels from the plurality of depth layer maps according to the probability map, a depth map of a scene.