42 results about "Visual distortion" patented technology

The invention discloses a robust watermarking method for stable diffusion generated images. The robust watermarking method comprises the following steps: step 1, watermark embedding; step 2, generating a watermark image; step 3, watermark extraction; step 4, watermark detection and source tracing; according to the method, the watermark information is encrypted and mapped into Gaussian noise approximately conforming to normal distribution, and is embedded into a potential space instead of directly modifying pixels, so that visual distortion of an image is avoided, native logic of a diffusion model is fit, model parameters do not need to be adjusted or network fine adjustment is not needed, and the original generation performance of the model can be fully guaranteed; the watermark is expanded to the potential space full dimension, the watermark covers the whole semantic level of the image, a voting mechanism is introduced in the extraction stage to carry out consistency verification on the multi-copy watermark, various common attacks can be effectively resisted, the generated image can still keep the complete content structure and semantic consistency even under the high watermark capacity, and the image quality is improved. And the balance of model performance, image quality and robustness is realized.

The invention relates to a fan blade defect detection method and system based on visual feature enhancement, and belongs to the technical field of computer vision and wind turbine operation and maintenance. The method comprises the following steps: establishing a perspective transformation matrix, and analyzing an image distortion rate of a local curvature radius and a normal inclination angle of a blade to obtain an image distortion correction parameter; a blade target detection model is constructed, the perspective transformation matrix is analyzed through a double-branch feature fusion network, and blade defect detection is carried out; performing three-dimensional reconstruction on the fan blade image, and obtaining blade defect detection information in combination with binocular vision point cloud data; and performing cross validation on an image recognition result, synthesizing and constructing a blade defect sensing space based on visual distortion recognition, and generating a defect feature detection report.

This application discloses a rendering method, apparatus, device, and storage medium, relating to the field of intelligent driving technology. The method includes: receiving a user's theme switching command and loading a theme configuration file corresponding to the command to obtain corresponding theme rendering parameters; acquiring environmental perception data of the vehicle and determining the object to be displayed on the vehicle's in-vehicle interface based on the environmental perception data, the object representing obstacles within a preset range of the vehicle; generating a geometric line drawing of the object to be displayed based on the environmental perception data, the geometric line drawing representing the outline of the obstacles; and rendering the object based on the theme rendering parameters using the geometric line drawing as a constraint to obtain the theme object corresponding to the in-vehicle interface. This application constrains the theme rendering process by combining the geometric line drawing generated from the environmental perception data, achieving theme switching while avoiding visual distortion or obstruction of core safety information, thus improving driving safety.

The invention discloses a vehicle type recognition method and system based on multi-modal perception and dynamic feature matching, particularly relates to the technical field of vehicle type recognition, is used for solving the problems of feature drift and recognition errors caused by multi-source signal interference, and constructs a multi-source time-sharing tensor by synchronously fusing vision, millimeter wave and wheel speed data through a self-adaptive time mark; real-time tracking of cross-domain features is realized by means of a pseudo-topology tracking grid and an elastic anchor point, drift signals of visual distortion and dynamic abrupt change are sensitively captured by flash stripe lag and speed feedback turn-back amplitude parameters, a drift cut element index is generated in real time through a time-error cut-off mapping operator, self-adaptive correction is performed on different-source features, and complex dynamic disturbance is identified; high-risk drift entries are retrieved in a matching memory library, a correction matrix is constructed, directional back-compensation is carried out on the time-sharing tensor, and multi-dimensional alignment of vision, echo and dynamic characteristics is achieved; and finally, fusing the correction feature and the power mode mapping distance, efficiently outputting a vehicle type label, and pushing the vehicle type label to a charging and safety node.

A method includes determining a viewpoint modification of a first viewpoint from which an input image represents a scene, and determining, based on the input image and the viewpoint modification, a reoriented image that (i) represents the scene from a second viewpoint that differs from the first viewpoint and (ii) includes a visual distortion of the scene. The visual distortion may be associated with the viewpoint modification. The method also includes processing the input image and the reoriented image using an image correction model configured to remove visual distortions associated with viewpoint modifications, and generating, using the image correction model and based on processing the input image and the reoriented image, an output image that includes a correction of at least part of the visual distortion in the reoriented image. The output image may represent the scene from the second viewpoint. The method additionally includes outputting the output image.

A virtual eye test for measuring and adjusting visual distortions caused by astigmatism can be conducted in a virtual reality (VR) environment. The test employs an electronic device featuring a head-mounted display (HMD) and a camera. The device generates a VR user interface corresponding to a three-dimensional virtual environment and renders it on the HMD. A plurality of visual scenarios is simulated within the VR interface. Using the camera, the device tracks user interactions and responses to visual stimuli presented in these scenarios. Visual distortions are measured and adjusted for based on the tracked user interactions and responses, providing a comprehensive assessment and correction of astigmatism-related visual distortions in a controlled, immersive environment.

The invention discloses a visual guidance control method for an underwater hull cleaning robot, and relates to the technical field of visual guidance, and the method comprises the steps: obtaining underwater original images of the vertical and inclined surfaces of a broadside hull, and carrying out the longitudinal partitioning of the images along the broadside height based on the gravity direction; in combination with partition brightness distribution prior, adaptive exposure compensation is performed on each partition, brightness mutation of adjacent partitions is eliminated through longitudinal consistency constraint, and a broadside image with continuous brightness is generated; on this basis, hull weld joint and reinforcing rib structure feature lines are extracted, a visual reference coordinate system consistent with the hull structure is constructed, then real-time deviation correction control is conducted on the cleaning robot according to the visual offset, and the cleaning robot is guided to execute a stable and continuous broadside cleaning path in the direction of the structure feature lines; the problems of visual distortion, unstable structural feature recognition and easy deviation of a cleaning path caused by non-uniform illumination and posture change in an underwater broadside cleaning scene are solved.

The invention relates to the technical field of data visualization, in particular to a dynamic water ball chart real-time generation method based on Berlin noise, which comprises the following steps of: acquiring a UV matrix, extracting a Y value in the matrix and mapping the Y value to a range of 0-1.875, generating an offset vector offset UV matrix by combining the current frame time, generating a Berlin noise matrix in a range of 0-0.875 on the offset matrix, and superposing data values, according to the dynamic water ball graph real-time generation method based on the Berlin noise, a color matrix is obtained through binaryzation and color mapping, then the alpha value of the color matrix is adjusted based on the Euclidean distance of the center coordinate of the matrix, and finally the color is assigned to the corresponding vertex and the fragment element, water ripples can have C2 continuity and non-repeatability, the edge is free of sawteeth, and the dynamic water ball graph real-time generation method based on the Berlin noise can be used for generating the dynamic water ball graph. The dynamic effect and the real-time data are accurately associated, visual distortion is avoided, execution is efficient, and the visualization quality of the water polo chart is effectively improved.

The invention discloses a high dynamic range video tone mapping method, device and equipment, and the method comprises the steps: carrying out the display preprocessing of a target high dynamic range video frame image, and obtaining a linear light image; processing the linear light image by using a Retinex algorithm to obtain an illumination image; estimating a reflection map according to the illumination map by using an illumination reflection theory; converting the illumination image to a frequency domain, and performing spectral analysis to obtain a first radial average power spectrum; converting the reflectogram into a frequency domain, and performing spectral analysis to obtain a second radial average power spectrum; determining homomorphic filtering parameters according to the first radial average power spectrum and the second radial average power spectrum; processing the linear light image by using an illumination reflection model, and converting an obtained processed time domain signal into a frequency domain to obtain a processed frequency domain signal; and performing filtering processing on the processed frequency domain signal according to the homomorphic filtering parameter, and performing tone mapping on an obtained filtering result. According to the invention, detail loss and visual distortion are avoided.

The invention discloses a pet three-dimensional reconstruction and personalized interaction behavior generation system and method based on monocular video driving. According to the system, skeleton and acoustic features in a video are extracted through a multi-modal feature decoupling module, and high-fidelity appearance reconstruction and physical collision feedback are realized by using an explicit-implicit hybrid rendering technology based on discrete volume primitives (such as 3D Gaussian). Meanwhile, a style-constrained imitation learning network (GAIL) and a user feedback fine tuning mechanism (RLHF) are introduced, and the personalized exercise expression of the pet is copied from the video. According to the method, the problems of visual distortion, lack of physical feedback of interaction and rigid behavior mode of the virtual pet in the prior art are effectively solved, immersive digital twin experience of visible, touch and personality is realized, and the method has perfect end-side performance adaptation and privacy protection capabilities.

The application discloses a kind of robust watermarking methods for stable diffusion generation image, including the following steps: step one, watermark embedding;Step two, watermark image generation;Step three, watermark extraction;Step four, watermark detection and tracing;The present application is by watermark information encryption mapping as approximately conforming normal distribution Gaussian noise, embedding latent space rather than directly modifying pixel, both avoid visual distortion of image, and also fit diffusion model original logic, without adjusting model parameters or network fine-tuning, can fully guarantee model original generation performance;By extending watermark to latent space full dimension, make watermark cover the whole semantic level of image, introduce voting mechanism to check consistency for multiple copies of watermark in extraction stage, can effectively resist various common attacks, and even under high watermark capacity, generated image can keep complete content structure and semantic consistency, realizes the balance of model performance, image quality and robustness.

The astigmatic lens comprises a central area and a peripheral area, the peripheral area is arranged around the central area, and the radial size of the central area is larger than that of the peripheral area; the central area is provided with an astigmatism cylindrical lens, the peripheral area is provided with a plurality of micro cylindrical lenses in array arrangement, the diopters of the micro cylindrical lenses gradually change along different meridian directions, and the difference between the diopters of the micro cylindrical lenses and the diopter of the central area is gradually increased in the direction far away from the central area. And the diopter of each micro-cylindrical lens is smaller than that of the central area. Furthermore, the micro-cylindrical lenses are arranged in the peripheral area of the astigmatic lens, sudden change of the diopter of the original astigmatic lens is relieved through gradual change of the diopter of the micro-cylindrical lenses, and the magnification difference generated by the astigmatic lens at the edge is reduced, so that visual distortion is reduced, and when a wearer wears the astigmatic lens, the astigmatic lens is more comfortable to wear. The dizziness feeling of a wearer can be reduced, and the experience of the wearer is improved.

The present invention relates to the technical field of curved microcrystalline panels, and more specifically, to a curved microcrystalline glass plate, its preparation method, and its application. The roughness of the inner and outer surfaces of the curved microcrystalline glass plate is both ≤0.06μm; the radial unevenness A of the curved microcrystalline glass plate is ≤0.4%, where A = h / L×100%; and the axial unevenness B of the curved microcrystalline glass plate is ≤0.4%, where B = d / L×100%. The curved microcrystalline glass plate provided by the present invention has a flat and smooth surface, can avoid optical scattering, and meets visibility requirements. It also has good flatness, low radial and axial unevenness, and no "warping" phenomenon or obvious visual distortion.

The invention provides a polyamide composition reinforced fiber, a CPI material and a preparation method of the polyamide composition reinforced fiber and the CPI material. The polyamide composition reinforced fiber is prepared from the following components in percentage by weight: 50 to 75 percent of semi-aromatic polyamide precursor, 10 to 25 percent of epoxy grafted cellulose nanofiber, 0.5 to 1 percent of epsilon-caprolactam sodium salt and 0.3 to 0.5 percent of N-acetyl caprolactam, wherein the polyamide composition forms a core-shell structure reinforced fiber through an in-situ fiber forming technology. Through multi-scale structural design, the comprehensive mechanical property of the CPI thin film is remarkably improved, the tensile strength and the elastic modulus of the material are improved by 50%-80% and 30%-50% respectively compared with pure CPI, meanwhile, high optical performance is kept, and visual distortion and brightness loss are avoided; and the folding durability of the material is improved through the core-shell fiber network-polyamide transverse crystal structure-polyamide phase multi-mechanism synergistic effect, and collaborative optimization of mechanical properties and optical properties is realized.

The application discloses a kind of astigmatic lenses and glasses, wherein, astigmatic lenses, including: center area and surrounding area, surrounding area is arranged around center area, the radial dimension of center area is greater than the radial dimension of surrounding area;Center area is arranged with astigmatic cylinder, surrounding area is arranged with multiple array arrangement microcylinders, the diopter of each microcylinder is gradually changed along different meridian direction, in the direction away from center area, the difference between the diopter of each microcylinder and the diopter of center area gradually becomes large, and the diopter of each microcylinder is less than the diopter of center area.Further, by setting microcylinders in the surrounding area of astigmatic lenses, the diopter of microcylinders is gradually changed to relieve the abrupt change of the diopter of original astigmatic lenses, reduce the magnification difference generated at the edge of astigmatic lenses, to reduce visual distortion, so that the wearer can reduce the dizziness of the wearer when wearing the astigmatic lenses, improve the experience of the wearer.

The invention discloses a gear production quality detection method and system based on machine vision, and relates to the technical field of gear intelligent production, and the method comprises the steps of gear image acquisition, image adaptation enhancement, grading defect detection and production quality feedback. According to the invention, image adaptation enhancement is carried out by adopting an image adaptation enhancement method combined with visual distortion perception, surface details and structural features are enhanced in a targeted manner through automatic identification of an illumination distortion area in a gear image, the phenomena of reflective interference and texture fracture are effectively inhibited, and the accuracy of the image adaptation enhancement is improved. Therefore, the recognizable degree of the defect area and the stability of overall detection are obviously improved. A grading defect detection method combined with polar coordinate convolution enhancement is adopted to carry out grading defect detection, through fine modeling and periodic analysis of a gear rotational symmetry structure, the sensitivity and judgment capability of gear structural defects are effectively improved, and more stable and reliable data support is provided for gear quality evaluation.

The invention relates to the field of intelligent driving testing, and discloses a darkroom lane departure early warning electromagnetic anti-interference evaluation method and system, and the method comprises the steps: debugging a projection device, prefabricating an animation for simulating lane departure, including equivalent visual distortion and the like when a lane line deviates leftwards and rightwards, and setting the four actions of leftward movement, return, rightward movement and return to be executed in sequence in a reciprocating manner; a tested vehicle is fixed in a darkroom, the vehicle is fixed on a rotating hub through a bandage, the rotating hub is set to be in a non-loading mode, a pneumatic robot is installed, an instruction is issued in a control room through an optical fiber, the pneumatic robot is made to adjust the depth of a vehicle accelerator, and the vehicle is made to run on the rotating hub according to the required speed. According to the invention, the anti-interference capability of the lane departure early warning system can be accurately tested.

The present application belongs to the technical field of length and area measurement, and relates to a worm and gear meshing contact area measurement method based on image recognition. Three-dimensional grid models of tooth surfaces are constructed by obtaining worm and gear design parameters, and a two-dimensional image pixel and three-dimensional grid model reverse mapping relationship is established by calibrating an image acquisition device. Two-dimensional meshing mark images are collected, and are inversely projected onto three-dimensional grid model development maps to eliminate perspective distortion. Edge pixel gray gradient features of the development maps are extracted, a mapping function of the gray gradient features and tooth surface elastic deformation is constructed based on a Hertz contact mechanics model, and visual distortion areas caused by elastic recovery are corrected. The present application solves the problem of contact area measurement distortion caused by curved surface distortion and mark distortion.

The embodiments of the present application provide an image processing method, apparatus, and electronic device. The method includes: first, obtaining a reconstructed image; then, performing image enhancement on the reconstructed image to obtain an intermediate image; then, determining the texture complexity weight corresponding to each region in the intermediate image, where the texture complexity weight is a number between 0 and 1; and then, based on the texture complexity weight corresponding to each region in the intermediate image, attenuating the texture intensity corresponding to each region in the intermediate image to obtain an enhanced image corresponding to the reconstructed image. In this way, the texture of the textured regions in the intermediate image can be preserved while the texture of the non-textured regions can be attenuated. This enhances the texture of the textured regions in the reconstructed image while avoiding the generation of false texture in the non-textured regions, thereby reducing the visual distortion of the reconstructed image, increasing the realism of the image, and improving the subjective quality of the image.

The invention discloses a quality evaluation method and system for a generated video, electronic equipment and a storage medium, and belongs to the technical field of video quality evaluation. In one aspect, through the combination of learnable text vectors and quality level embedded representations, a pre-trained text encoder is enabled to adapt to a quality assessment task in a parametric efficient manner without modifying its core weight, significantly reducing trainable parameters, and maintaining the semantic understanding ability of pre-training. And on the other hand, the to-be-trained LoRA module is injected into the pre-trained image encoder, and the LoRA module enables the image encoder to capture a specific visual distortion mode of the AI generated video in a low-rank updating mode, so that the problems of parameter expansion and disastrous forgetting caused by full-amount fine tuning of the pre-trained image encoder are avoided. In addition, the generated video output by the text-to-video generation model is used for updating the sample generated video, so that the accuracy of AI generated video evaluation is ensured.

The invention relates to the technical field of video rendering, in particular to an AI aerial video intelligent rendering method based on a scenic area unmanned aerial vehicle, and the method specifically comprises the steps: obtaining aerial video stream data, holder stability augmentation log data and shooting scene metadata of the unmanned aerial vehicle; evaluating the disturbance degree of each frame of picture based on the video stream data and the stability augmentation log data; scene repair strength evaluation is carried out based on the shooting scene metadata; and combining and analyzing the picture disturbance degree evaluation result and the scene repair strength evaluation result to obtain a repair strategy of AI intelligent rendering of the current video segment. According to the method and the device, the problems of single evaluation dimension, scene perception deficiency and restoration strategy stiffness when coupling visual distortion caused by pan-tilt stability augmentation, natural medium motion and a complex depth-of-field structure is processed in the prior art are solved.

The present disclosure relates to the technical field of arc-shaped microcrystalline panels, and particularly relates to an arc-shaped microcrystalline glass plate, and a preparation method therefor and the use thereof. The roughness of the inner and outer surfaces of the arc-shaped microcrystalline glass plate is less than or equal to 0.06 μm; the radial unevenness A of the arc-shaped microcrystalline glass plate is less than or equal to 0.4%, and A=h / L×100%; and the axial unevenness B of the arc-shaped microcrystalline glass plate is less than or equal to 0.4%, and B=d / L×100%. The arc-shaped microcrystalline glass plate provided in the present disclosure has a flat and smooth surface, can avoid optical scattering, satisfies the visibility requirement, has good flatness, low radial unevenness and axial unevenness, and has no "warping angle" phenomenon or obvious visual distortion.

The present application relates to the technical field of foundation pit engineering image monitoring, and in particular to a method and system for automatic identification and measurement of cracks in a foundation pit support structure, the method comprising the following steps: obtaining a reference image recording the surface state of the foundation pit support structure before applying a cover layer to the surface of the foundation pit support structure; obtaining a calibration image recording the surface state of the foundation pit support structure after the cover layer has solidified; for an existing crack in the foundation pit support structure that exists in the reference image, determining the visual distortion parameter of the cover layer on the existing crack by comparing the apparent differences of the existing crack in the reference image and the calibration image; and determining the source of a new linear feature by comparing the new linear feature with the reference image, thereby effectively reducing the influence of the cover layer on crack monitoring, and having the advantages of being able to reduce the influence of the cover layer on the measurement accuracy of existing cracks and improving measurement accuracy.

The invention belongs to the technical field of computer graphics, and discloses a large-scale terrain variable-rate coloring method based on scene texture and motion information. Depending on the support of NVIDIATuring architecture on variable rate coloring characteristics, the method comprises the following steps: firstly, quickly calculating a motion vector of a segment corresponding to each pixel in a screen space at an application program end according to relative motion of a virtual camera and a terrain object; secondly, approximating a current rendering frame by using a previous frame of rendering image, and calculating information entropy of each pixel block for analyzing the richness of texture features of the pixel block; and finally, respectively mapping the pixel block motion vector and the texture information entropy to an optional coloring rate space, and determining the final coloring rate of the pixel block according to the maximum value of the pixel block motion vector and the texture information entropy and an adjacent frame constraint. According to the method, the coloring rate can be adaptively adjusted by effectively utilizing the characteristics of related applications in the aspects of motion states and texture information, and the calculated amount of the coloring process is reduced at the cost of relatively small rendering image visual distortion.

The present application relates to the field of chip production, and mainly relates to a visual system compensation method and device under high-temperature environment and a storage medium, the method comprises the following steps: S1: acquiring a visual system input image under high-temperature environment; S2: detecting the image through an image processing algorithm; S3: compensating the detected visual distortion; S4: post-processing the compensated image and outputting a final image. The present application detects the visual distortion caused by high-temperature effect in the image through the image processing algorithm, and references a compensation model and trains the model based on the visual characteristic data under high-temperature environment, so as to compensate the thermal deformation and optical distortion of the optical system caused by high temperature, and the device offset and alignment error, so that the visual system can realize the effective method of image quality compensation, optical deformation correction and alignment error under high-temperature environment.

Automatic creation of product photoshoots that seamlessly combine real-life image portions and Artificial Intelligence (AI) generated portions. An input photograph of a product is received, and a background-removed version is generated. The image is fed into a Vision and Language Model (VLM) that generates textual attributes that pertain to the product. The textual attributes are fed into a Large Language Model (LLM), that generates a proposed textual prompt that will command an Image Diffusion unit to generate via Generative AI an image of a scenery that would be appropriate for showcasing that product. The AI-generated image is fed into an AI-based unit that detects and cures visual abnormalities or visual distortions, with AI-based blending and refinement. The method generates a high-definition abnormality-free and distortion-free output image that depicts the real-world product blended seamlessly within the Generative-AI scenery. A similar process performs virtual staging of a room or a house or other venue.

The invention discloses a scenic spot AR guide system supporting dynamic loading of a 3D model and a data processing method. According to the technical scheme of the invention, through the innovative design of a dynamic resource scheduling algorithm, a multi-level cache architecture and a hybrid rendering pipeline, the long-standing key technical problems of poor equipment adaptability, high network dependency, serious visual distortion and the like in the AR guide field are systematically solved. Practical application data shows that in typical scenes such as scenic spots, libraries and scientific halls, system performance indexes comprehensively exceed existing schemes, and reliable technical support is provided for large-scale commercialization.

The present application relates to the technical field of data mapping, in particular to a cross-platform geological vector attribute mapping method based on geometric data conversion, comprising separating the geometric sequence, original symbol index and attribute item of source platform data; generating conversion coordinate sequence by using the Boolean seven-parameter matrix and elastic grid deformation algorithm; normalizing the attribute and binding the unique symbol identifier with the coordinate sequence to generate geometric element object; extracting multi-level graphic element logic protocol by using semantic translation operator and constructing multi-dimensional index mapping matrix by using manifold learning; matching target platform entity handle, calculating expression difference compensation factor and performing adaptive alignment; and performing integrated packaging according to the target platform standard. The present application effectively solves the problems of symbol loss, topological distortion and visual distortion in cross-platform migration by decoupling the underlying logic and semantic projection, and eliminates the need for manual secondary redrawing.

This invention provides a method and related apparatus for detecting potential hazards in power distribution network equipment based on large field-of-view stitched video data. This method significantly expands the detection field of view by merging video data from different perspectives, and solves the problem of repeated shooting of multiple power distribution network devices at the same location through a large field of view. By combining synchronous timestamps to precisely align the spatial and temporal parameters of multiple video data streams, the continuity and consistency of the stitched video in time and space are ensured, improving the reliability and effectiveness of the detection results. Through precise extraction and matching of image features from multiple video frames, visual distortion and seam marks during the stitching process are reduced, generating high-quality fused video data, providing a clear and continuous visual foundation for subsequent hazard detection. The use of a pre-trained hazard detection model for automatic detection of the fused video data greatly improves detection speed and efficiency, and enhances the objectivity and standardization of the detection process.

The invention discloses an audio and video processing method and device and electronic equipment, and relates to the technical field of audio and video coding and decoding, and the method comprises the steps: determining a dynamic texture region and a static background region of an audio and video based on the time-space domain features corresponding to the audio and video to be transmitted; performing dynamic block coding on the dynamic texture region, and performing compression coding on the static background region to obtain layered coding data corresponding to the audio and video; generating an initial synchronous code stream corresponding to the audio and video based on the layered coding data; performing alignment operation on the initial synchronous code stream to obtain a target synchronous code stream corresponding to the audio and video; and after the target synchronous code stream is transmitted from the coding end to the decoding end, decoding the target synchronous code stream. According to the method and the device, the technical problems of visual distortion and asynchronous lip sound in the audio and video encoding and decoding process caused by processing the audio and the video based on a fixed encoding strategy in the prior art are solved.