19 results about "Visual texture" patented technology

The application provides a display device and a subtitle language recognition method. The method can acquire a subtitle corresponding to a media video in response to a play instruction of the media video, extract visual texture features of the subtitle in a case where the subtitle is an image subtitle, perform language recognition according to the visual texture features to obtain a language recognition result, extract syntax topology features of the subtitle in a case where the subtitle is a text subtitle, perform language recognition according to the syntax topology features to obtain a language recognition result, write the language recognition result into metadata corresponding to the media video, control a display to play the media video and the subtitle, and display a language identifier corresponding to the subtitle according to the language recognition result in the metadata corresponding to the media video. The method recognizes a language by analyzing features of different languages in physical forms and syntax structures, so that the language identifier of the subtitle can be correctly displayed when the media video is played.

This invention discloses an adaptive end-to-end network learning-driven polarization descattering depth estimation method. It proposes a physics-driven end-to-end network to simultaneously acquire clear reconstructed images and high-precision depth images in scattering environments. The method first acquires linearly polarized images of the scene in four directions, extracting light intensity and linear polarization feature maps through decoupling calculations. Then, a dual-task network including a polarization-guided fusion module is constructed. Utilizing the physical sensitivity of DOLP to scattering particle concentration and optical path length, it transforms these into a spatial attention mask, physically constraining and enhancing the shallow texture of the intensity map in the feature space. Finally, dehazing and depth estimation branches are encoded and decoded in parallel. This invention overcomes the intensity-depth ambiguity of traditional RGB images, effectively suppressing the interference of non-uniform fog on geometric perception and visual texture, and achieving high-precision scene depth estimation and high-fidelity dehazing imaging under adverse weather conditions.

This application discloses a lane selection method, electronic device, and computer-readable storage medium, relating to the field of intelligent driving technology. The method involves acquiring the real-time smoothness score of the current lane while meeting safety and navigation requirements, and estimating the smoothness score of candidate lanes using observations of the movement of the preceding vehicle, historical data, or visual texture information. By comparing the differences between the two scores and based on a preset smoothness difference threshold, a decision is made on whether to switch lanes to a candidate lane with better smoothness. This invention solves the technical problems of related intelligent driving systems being unable to perceive smoothness differences between multiple lanes in the same direction and being unable to actively avoid bumpy lanes by introducing a cross-lane smoothness comparison perception and multi-source confidence weighted fusion mechanism. It achieves the technical effects of automatically selecting smooth lanes, improving driving comfort, reducing abnormal wear of vehicle suspension and tires, and suppressing ineffective lane changes under the premise of safety and compliance.

The application discloses a kind of unmanned aerial vehicle monocular vision inertial SLAM loop detection methods based on undergrowth trunk geometry, the method includes the following steps: step 1, trunk region segmentation and three-dimensional point cloud generation;Step 2, trunk clustering detection and position positioning;Step 3, triangular geometric feature construction and screening;Step 4, multidimensional similarity calculation;Step 5, geometric and visual joint decision and loop confirmation.This method extracts stable geometric information of trunk by driving depth data, constructs triangular features with invariance with trunk position as vertex to break through the dependence of visual texture, realizes double-path independent detection and joint decision by combining traditional visual bag-of-words model, balances detection accuracy and real-time through adaptive matching strategy, thereby effectively improves the accuracy, robustness and environmental adaptability of loop detection in undergrowth GNSS-free environment.

The application provides a kind of vehicle-mounted intelligent driving system obstacle identification function test method and device, related to automatic driving technical field, which comprises: adopting wave-transparent material to make three-dimensional pedestal, three-dimensional pedestal includes multiple surfaces;Establish the mapping relationship between aging degree and physical properties, determine the physical properties of different surfaces based on the mapping relationship, physical properties include radar cross section and contrast;According to physical properties, radar reflection layer with different radar cross sections and visual texture layer with different visual textures are respectively arranged on each surface of three-dimensional pedestal to generate obstacles with different aging degrees on different surfaces;Using obstacles, the obstacle identification function of vehicle-mounted intelligent driving system is tested. By using the above vehicle-mounted intelligent driving system obstacle identification function test method and device, the functional test efficiency and test scene coverage when testing the vehicle-mounted intelligent driving system are improved.

This invention discloses a velvet-textured stain-resistant ceramic tile and its preparation method, belonging to the field of ceramic tile manufacturing technology. It includes a body, a base glaze layer, a skeleton layer, a top glaze layer, and a protective glaze layer. This invention solves the problem of complex layering in the production of velvet-textured ceramic tiles by using the above materials and preparation method. By physically granulating the base glaze layer to form a skeleton with a micron-level protrusion array, it drives the top glaze layer to create undulations, directly shaping a velvet-like tactile and visual texture. The protective glaze layer achieves stain-resistant and slip-resistant functions. Simultaneously, this preparation method avoids the problem of insufficient precise control of microstructure and interlayer bonding strength caused by single firing or simple pre-firing. Through staged temperature-time control, it retains the micron-protrusion characteristics of the skeleton layer while ensuring the stability and functionality of the overall structure.

The application provides an automatic mapping method for complex terrain environment, and relates to the technical field of automatic mapping. The method comprises the following steps: firstly, acquiring multi-modal sensing data and performing pose calculation to construct an initial pose graph and a local mapping model; extracting spatial geometry and visual texture features to calculate an information matrix, and solving a spatial feature degradation index and a degradation type identifier; when the degradation index exceeds a preset threshold, determining a feature degradation area, calculating a dynamic observation weight matrix and extracting a local credibility topological boundary based on the feature degradation area; planning an active compensation survey trajectory based on the boundary, controlling a mapping platform to perform active compensation survey to acquire supplementary sensing data, and generating a compensation constraint edge by correlation; finally, injecting the weight matrix and the compensation constraint edge into the pose graph to perform global graph optimization solving, and outputting a globally consistent mapping product. The application solves the problems of divergence of pose solving caused by environmental feature degradation and lack of active compensation repair mechanism in the prior art.

This invention discloses a deep learning-based visual inspection method and system for micro-defects, belonging to the field of visual inspection technology. By establishing a correspondence between material temperature and visual images, and mining temperature gradients and visual texture features, the concealment of hidden defects is overcome. By determining the boundaries of abnormal temperature regions, the location distribution of hidden defects is pinpointed. Furthermore, through local vector extraction and vector similarity measurement methods, the positioning accuracy of hidden defects is improved, achieving precise capture of hidden defects. By filtering and matching sets, and then precisely fusing minute temperature difference signals to the pixel layer of the visual image, high-precision matching between the temperature information of hidden defects and the visual image is achieved. Finally, by enhancing the fused image, the morphological parameters of the hidden defects are calculated, achieving precise positioning of material hidden defects and significantly improving the accuracy and reliability of micro-defect detection.

This utility model belongs to the field of metal stamping technology and discloses a stamping die for bicycle tubing. The stamping die for bicycle tubing includes a punch assembly and a die assembly. Both the punch assembly and the die assembly include a pressing die and a retaining die. The retaining die is elastically disposed on the pressing die and protrudes relative to the pressing die. There is a cutting edge arc between the retaining die and the pressing die. The retaining die of the punch assembly abuts against the bottom wall and side wall of the tubing groove. The pressing die of the punch assembly abuts against one side of the tubing body, so that the cutting edge arc can reduce the R angle of the tubing groove. The retaining die of the die assembly abuts against the other side of the tubing body and the outer side wall of the tubing protrusion. The pressing die of the die assembly abuts against the top of the tubing protrusion, so that the cutting edge arc can reduce the R angle of the tubing protrusion. The convex or concave fonts or logos present sharp angles and clearer edges, meeting design requirements, improving visual texture, and are suitable for designs that pursue a tough style.

The application discloses a semantic segmentation method based on phase profile constraint and belongs to the technical field of computer vision. The method uses a calibrated camera-projector system to collect a two-dimensional texture image and a deformed fringe image of a to-be-measured scene under the same field of view; the deformed fringe image is solved through multi-step phase shifting and Gray code technology to obtain an absolute phase distribution and calculate a gradient field; according to a preset mutation condition, a physical edge profile reflecting the depth change of an object is extracted from the gradient field; the physical edge profile is mapped into the two-dimensional texture image based on the corresponding relationship of pixel coordinates to accurately define the semantic object area, the corresponding category label is given, and finally a semantic label mask image is generated. The application uses physical phase information to correct the uncertainty of visual texture edges, effectively solves the edge extraction problem under a complex light scene, realizes the automatic generation of semantic segmentation data, and improves the training precision and robustness of a semantic segmentation model.

This invention provides a method and system for dense 3D reconstruction by fusing imaging radar point cloud and visual image data, belonging to the field of 3D reconstruction technology. First, imaging radar point cloud data and visual image data are simultaneously acquired to form a dual-source data set. Next, a pixel-level depth correlation is established between the two, defining the pixel depth range. Then, visual texture detail information is extracted to fill in sparse areas of the imaging radar point cloud, forming intermediate dense point cloud data. Next, visual texture color information is mapped to the intermediate point cloud, and the spatial point depth positions are adjusted to generate dense point cloud data with coordinated depth and texture. Finally, all spatial point information is extracted to restore the target scene shape and texture, forming a color dense 3D reconstruction model. This invention effectively integrates the advantages of both types of data to achieve high-precision dense 3D reconstruction.

This utility model discloses a battery decorative cover, comprising: a fiberglass substrate; a gray base coat layer, screen-printed on the surface of the fiberglass substrate; a first color layer; a second color layer, used to supplement and strengthen the first color layer, thereby increasing color saturation and pattern complexity; a pearlescent ink layer, screen-printed on the second color layer; a first UV texture layer, transferred onto the pearlescent ink layer using a UV transfer device, providing a visual texture effect; a PVD coating layer, covering the surface of the first UV texture layer; a high-transmittance varnish layer, screen-printed on the PVD coating layer; and a second UV texture layer, transferred onto the high-transmittance varnish layer using a UV transfer device, providing a tactile texture effect. Through the above methods, this utility model can improve gloss and brightness, supporting the customization of complex decorative graphics.

The application discloses an image instance segmentation method, system, device and medium, and relates to the technical field of computer vision. The method comprises the following steps: acquiring an original image of a to-be-detected scene and a visual texture prompt image corresponding to a to-be-detected target object of the original image; performing preliminary detection and foreground extraction on the input image by using a target detection model guided by visual prompting; extracting multi-scale fusion features by using a multi-scale fusion texture similarity matching network based on DinoV2 local and global features; performing similarity matching on the extracted multi-scale fusion features and reference features; and performing category allocation on candidate target segmentation regions according to joint similarity scores, so as to screen out a final image instance segmentation result. The method effectively suppresses background clutter and fragmented error proposals, and realizes high recall rate while ensuring high accuracy.

The application provides a visual texture navigation AGV running track real-time monitoring method, comprising the following steps: obtaining real-time ground image frames and current driving mileage from an image acquisition module and a position solution module on an AGV chassis, combining pre-stored historical track coordinates to obtain initial positioning data; evaluating the similarity pattern repetition degree between adjacent image frames according to the initial positioning data and the height of the preset threshold, marking the positions higher than the threshold as texture confusion high-risk sections to determine potential misjudgment positions; for the candidate anchor points, combining the driving mileage cumulative error to dynamically adjust the filtering threshold value, retaining the anchor points with the threshold value lower than the average length of the texture period to obtain the optimized anchor points; from the optimized anchor points, selecting the texture section with the highest similarity of the texture features of each anchor point and the gray distribution of the current image frame as the matching reference, evaluating the size of the coordinate correction deviation and the allowed range, and determining the corrected track coordinates.

The application discloses a kind of dynamic path adjustment systems of automatic sampling car based on multi-objective optimization, it is related to the field of automatic driving, including the following modules: blind area boundary attribute identification module, for output blind area coordinate set;Non-visual texture feature matrix construction module, for calculating the direction vector and estimated detection depth of each virtual detection ray;Probability distribution map updating module, for output obstacle probability distribution map;Potential field oriented planning module, for constructing potential field based on probability distribution, generate heuristic path;Optimal heuristic path decision module, for screening optimal path;Dynamic re-planning trigger module, for constructing uncertainty evaluation index.The application overcomes the limitations of traditional methods, such as single representation of blind area, neglecting medium transmission characteristics and low exploration efficiency, and provides an efficient solution for autonomous operation of automatic sampling car in complex non-visual environments.

This invention discloses a remote monitoring method for multi-source sensing data fusion based on edge computing, belonging to the field of edge computing technology. This method operates on edge computing nodes. First, it utilizes the energy distribution characteristics of high-frequency one-dimensional signals to perform physical weighted interpolation reconstruction of missing frames in low-frequency images at transient impact moments. Second, it extracts the visual texture entropy of the reconstructed frames and the physical spectral entropy of the corresponding signals, and calculates the matching confidence of the data through a consistency check function to eliminate environmental interference. Finally, differential data encoding and uploading are performed only when the data is reliable and the Euclidean distance between the current feature and the locally maintained reference state vector exceeds a threshold. This invention solves the spatiotemporal semantic uncertainty problem of heterogeneous sensing data at the edge by synergistically combining temporal reconstruction, spatial verification, and semantic differential transmission, achieving high-fidelity monitoring under low bandwidth conditions.

A computer implemented method of creating a simulated realistic virtual model of a geographical area for training an autonomous driving system, comprising obtaining geographic map data of a geographical area, obtaining visual imagery data of the geographical area, classifying static objects identified in the visual imagery data to corresponding labels to designate labeled objects, superimposing the labeled objects over the geographic map data, generating a virtual 3D realistic model emulating the geographical area by synthesizing a corresponding visual texture for each of the labeled objects and injecting synthetic 3D imaging feed of the realistic model to imaging sensor(s) input(s) of the autonomous driving system controlling movement of an emulated vehicle in the realistic model where the synthetic 3D imaging feed is generated to depict the realistic model from a point of view of emulated imaging sensor(s) mounted on the emulated vehicle.

This invention discloses a synthetic aperture radar target recognition method based on the fusion of physical perception and vision. This method effectively addresses the problems of insufficient utilization of physical topological information and difficulties in feature semantic alignment in existing methods by constructing a GAT-Former dual-stream heterogeneous network. This invention models the dual-stream input space of the complex domain physical and visual flows, explicitly solves the scattering center, and constructs a dynamic topological graph. During the recognition process, a physical-guided cross-attention fusion mechanism and a heterogeneous integration strategy are used. Geometric topological features are extracted using deep unfolding networks and graph attention networks, visual texture features are extracted using ConvNeXt, and the final decision is optimized through weighted fusion with ResNet-50. Simulation results show that, compared with other benchmark methods, the proposed method significantly improves fine-grained classification accuracy and noise robustness on the ATRNet-STAR dataset, achieving deep complementarity between physical mechanisms and visual texture.

This invention discloses an artificial intelligence constraint generation method based on physical truth, comprising: acquiring three-dimensional geometric data and visual texture data of a target object in real time via a data acquisition terminal; performing local precision verification on the three-dimensional geometric data; dynamically generating guidance information if the verification fails; and digitally signing the acquired data to obtain digital signature information if the verification passes. A controlled three-dimensional data structure is generated based on the verified three-dimensional geometric data, compressed visual texture data, and digital signature information. A cloud server verifies the legality of the digital signature of the controlled three-dimensional data structure, parses the three-dimensional geometric data in the controlled three-dimensional data structure, converts the parsed three-dimensional geometric data into spatial constraints, and sends the spatial constraints to a generative artificial intelligence model. The generative artificial intelligence model generates content that conforms to physical truth and has a reliable source based on the spatial constraints. This method solves the problems of geometric illusion, verification lag, and lack of data trust.