164 results about "Pixel image" patented technology

The invention relates to the technical field of light field image super-resolution reconstruction, in particular to a light field image spatial super-resolution method based on a multi-scale large kernel attention mechanism, which comprises the following steps: inputting a target light field image into a trained spatial super-resolution reconstruction model, and outputting a spatial super-resolution reconstruction image of the target light field image; in the spatial super-resolution reconstruction model, performing feature extraction on an input light field image through a shallow feature extraction network to generate a shallow feature map; respectively extracting angle features, spatial features and epipolar plane image features from the macro-pixel image of the shallow feature map, the sub-aperture image for generating the macro-pixel feature map and the epipolar plane image of the shallow feature map through a large kernel feature extraction network, and fusing to generate a 4D light field feature map; and performing up-sampling reconstruction on the 4D light field feature map through an up-sampling reconstruction network to generate a spatial super-resolution reconstruction image of the input light field image. According to the invention, the reconstruction quality of the spatial super-resolution of the light field image can be improved.

The invention relates to a laser radar point cloud densification method and device fusing image information, and a medium. The method comprises the following steps: fixing a visual camera and a laser radar on a rigid tool, and carrying out joint calibration and data alignment; a laser radar is used to collect three-dimensional point cloud data, and a visual camera is used to shoot a scene; performing super-pixel segmentation on the image by using an image brightness linear iterative clustering algorithm; mapping the three-dimensional point cloud data into a segmented two-dimensional image superpixel pattern spot region by using a jointly calibrated camera model conversion relationship to realize feature clustering of the original three-dimensional point cloud of the laser radar; performing curved surface fitting on a clustered result by using a random sampling consistency algorithm; and linear interpolation is carried out on the fitted curved surface, dense points which are not covered by the original point cloud of the laser radar are supplemented and generated, and densification of the point cloud of the laser radar is realized. According to the method, the point cloud densification precision and practicability are improved, and technical support is provided for automatic driving, robot navigation and three-dimensional reconstruction.

The invention provides a pathological image analysis method and device based on multi-instance learning, equipment and a storage medium, and the method comprises the steps: receiving a lung pathological section image, segmenting the lung pathological section image into a plurality of 256 * 256 pixel image blocks, and screening and reserving effective image blocks containing a tissue structure; extracting a high-dimensional feature vector of each effective image block by using a self-supervised pre-trained deep neural network on the large-scale dyeing pathological image; the features are input into a multi-branch attention module containing 15 parallel branches, self-attention, cross-attention and global attention mechanisms are executed respectively, and local spatial dependence, cross-image block similarity and global relevance are modeled; and carrying out weighted aggregation on the basis of the processed attention weight to generate a global feature of the slice image, outputting a pathological category through a classification network, and supporting heat map visualization and feature clustering analysis. According to the method, unlabeled region information is effectively utilized, and the pathological classification precision and the model generalization ability are improved.

The invention provides a multi-nozzle jet flow uniformity determination method based on vision, and relates to the technical field of thick plate hot rolling and computer vision. According to the method, key frame extraction and preprocessing are carried out on collected video data; jet region identification is carried out by using an image segmentation technology, and density reference values are calculated pixel by pixel for the identified region; designing an index for measuring the jet flow uniformity according to a density reference value and supporting thermodynamic diagram visualization and analysis; through a mode of combining visual image processing and linear calibration, efficient and accurate quantitative evaluation on the jet flow uniformity of the multiple nozzles is realized. Moreover, the method can be realized through laser jet visualization software developed on the basis of a Python language, necessary array acceleration operation is carried out in combination with a Cupy library, the calculation speed of a single 1080 * 720-pixel image is smaller than 800 ms, and the requirements of actual industrial deployment and real-time detection are met.

The utility model discloses a light beam offset device, an image resolution improving module and a near-to-eye display device, the light beam offset device comprises a light beam offset element and a driving device, the driving device is used for driving the light beam offset element to deflect N times in each period, and enabling the light beam offset element to sequentially go through an initial state and N-1 deflection states in each period, so that T low-pixel sub-images displayed by the image display device pass through the light beam offset element and the projection lens and then correspondingly output one non-offset sub-image and N-1 offset sub-images to human eyes in sequence, therefore, human eyes can feel a high-pixel image obtained by fusing and superposing one non-shifted sub-image and N-1 shifted sub-images; wherein N and T are integers, N > = 2, and 1 < = T < = N. According to the utility model, small volume can be realized and high-definition pictures can be provided.

Disclosed is a display apparatus including: a substrate including a plurality of pixels having a plurality of sub-pixels; a pattern portion disposed on the substrate and concavely formed in a non-light emission area between the plurality of the sub-pixels; and a reflective portion disposed on the pattern portion, wherein each of the plurality of subpixels includes a pixel electrode disposed to be spaced apart from the reflective portion, and the pixel electrode includes a main body and a plurality of protrusions protruding from the main body toward the reflective portion. This configuration enhances the reflective properties in the non-light emission regions, helping to improve overall display efficiency, brightness uniformity, and image clarity.

The invention provides a method for processing a curve in an image, and the method comprises the steps: obtaining the scale lines of a plurality of coordinate axes on a coordinate image and a pixel image of the coordinate image, and enabling the coordinate image to comprise a coordinate grid and a function curve; determining a pixel grid line corresponding to the coordinate grid and a pixel curve of the function curve in combination with the pixel map; determining a target axis pixel coordinate of each scale line on each coordinate axis according to the axis pixel coordinate of the pixel grid line on each coordinate axis; according to a scale value of an adjacent scale line corresponding to a first numerical value of a target variable and a target axis pixel coordinate, determining a pixel point corresponding to the first numerical value on the pixel curve; and determining a second value of the preset variable corresponding to the first value of the target variable on the function curve according to the scale value of the corresponding adjacent scale line of the pixel point on the coordinate axis of the preset variable and the pixel coordinate of the target axis.

The application relates to a dual-branch hyperspectral image classification method based on a graph convolutional neural network and an attention mechanism, which comprises the following steps: inputting a hyperspectral original image to data for pretreatment; inputting the data after dimension reduction into a CNN attention mechanism branch, and obtaining processed branch features through an attention mechanism module, a data processing module, an attention mechanism module and convolution operation; inputting the data after dimension reduction into a GCN branch, performing simple linear iterative clustering, segmenting into k superpixels, encoding the k superpixels, obtaining superpixel graph nodes, inputting the graph nodes into a GCN module for feature extraction, decoding the extracted superpixel node features into pixel-level branch features; performing feature fusion on the two branches, using a cross-entropy function as a loss function to train the model, and using a softmax function to obtain a class label probability. The application takes limited hyperspectral images as research objects, improves high classification precision, and ensures classification speed.

The application discloses a strip steel coiling time side guide plate spark detection method based on image recognition, which comprises the following steps: 1, an image collector (8) is arranged above the side guide plate, the image collector faces the strip steel (2) and the side guide plate, and is connected with a control system through a controller (7); 2, the image collector sends images to the controller, the controller converts the images into pixel images, and initializes a time domain median background model; 3, whether the image serial number is continuous is determined, if yes, step 4 is executed, and if not, the step 1 is returned; 4, a spark connected domain is extracted, and the actual width of the spark is calculated according to the spark connected domain; 5, whether the time domain median background model is updated is determined, if yes, step 6 is executed, and if not, step 7 is executed; 6, the time domain median background model is updated; 7, a spark recognition result is output; 8, whether the spark detection program is exited is determined, if yes, the detection is ended, and if not, the step 1 is returned. The application can recognize the spark condition between the side guide plate and the strip steel through image collection.

This invention belongs to the field of laser ultrasound imaging technology, specifically disclosing a laser ultrasound imaging method based on neural networks, comprising the following steps performed sequentially: S1, acquiring the A-scan signal; S2, decomposing and reconstructing the A-scan signal through empirical mode decomposition; S3, extracting the temporal features of the ultrasound signal using pooling; S4, establishing a neural network and training it multiple times to obtain multiple neural networks with different representation capabilities; S5, inputting the temporal features of the ultrasound signal of the sample to be tested into each trained neural network, and filling the recognition results into the corresponding positions according to the detection point order to form a pixel matrix; S6, superimposing all pixel matrices, setting a threshold to evaluate the state of each detection point, and generating the corresponding pixel image. The method of this invention is insensitive to noise and can be used to process low signal-to-noise ratio signals, improving the accuracy of material state identification. This invention is applicable to detecting the state of materials.

This invention relates to an image dehazing method, system, storage medium, and device based on energy domain transformation, comprising the following steps: Step 1: Acquire the hazy image to be processed and perform preprocessing; Step 2: Calculate the energy of each intensity in the RGB three channels of the preprocessed image, calculate threshold parameters for the energy, and limit the energy; Step 3: Convert from the energy domain back to the RGB color space and output the dehazed image. Based on the basic feature statistical information of image pixel image contrast, the energy of the image is obtained using local statistical information, and the energy of the image is limited, thereby improving the relative contrast of the energy. After conversion back to the RGB space, the local contrast in the RGB space is significantly enhanced, thus realizing complex spatial domain feature processing and achieving high-performance dehazing function.

The invention belongs to the technical field of mineral calculation and reservoir evaluation, and particularly discloses an electric imaging logging mineral component quantitative characterization method based on porosity spectrum correction, which comprises the following steps: preprocessing original electric imaging logging data, and generating a high-density pixel image; performing probability density statistical analysis on the pixel value of each depth point in the high-density pixel image to obtain a pixel waveform spectrum; calculating the apparent porosity of each pixel point in the high-density pixel image to generate a pore spectrum; performing pore fluid influence correction on the original electrical imaging data based on the pore spectrum, performing image enhancement to generate a pure mineral pixel waveform spectrum, and performing equalization processing to obtain an imaging mineral probability map; according to mineral content data obtained through core XRD total rock analysis, an imaging mineral spectrum is subjected to quantitative calibration, and distribution cut-off threshold values and pixel point value ranges corresponding to different minerals are recognized and deduced. The method can effectively solve the calculation error caused by the influence of the traditional imaging mineral calculation fluid on mineral calculation.

The application relates to the technical field of image processing, and discloses a light-weight single-pixel imaging method based on frequency domain-space domain modulation attention, which comprises a single-pixel image data generator and a calculation reconstruction generator; an initial inversion estimation is obtained through inverse mapping of a measurement matrix; then, the initial inversion estimation is input into a dual-domain collaborative attention deep reconstruction network; the network is composed of a plurality of dual-domain collaborative attention module cascades connected in series, a multi-head spectrum calibration module in the module realizes long-distance information reconstruction through dual-domain feature information fusion interaction; a multi-scale-depth dynamic convolution fusion module extracts features in a space domain, accurately processes high-frequency edges, and significantly enhances the reconstruction capability of texture details; and the features are corrected in combination with a residual error guiding update module. Through the dual-domain collaborative mechanism, the reconstruction precision and the detail recovery capability of single-pixel imaging are effectively improved while the network parameter quantity and the calculation complexity are greatly reduced, and the application is suitable for edge deployment.

The invention discloses a real ink mark driven facsimile authentic method based on submerged space diffusion, which comprises the following steps: firstly, collecting a target facsimile image and a large number of real ink mark sample images, carrying out preprocessing operation on the images, and constructing a standardized training set and a test set; secondly, extracting a contour diagram and a skeleton diagram from the target copy image, and mixing the contour diagram and the skeleton diagram in a linear mode to generate a continuous structural feature map; constructing an ink mark encoder, encoding the authentic mark sample, and extracting the pen and ink characteristics of the authentic ink mark sample; and finally, constructing a reinking network based on a latent space diffusion model, and finally converting the optimized data representation into a pixel image through a decoder to obtain a calligraphy image with natural ink marks and real strokes. The problems that in the prior art, copy images are inflexible in ink color and lack of natural ink color change are solved, meanwhile, stability of a font skeleton and refinement of the stroke of the to-be-restored area are guaranteed, and authentic transformation work is achieved.

The invention discloses an integrated circuit image processing method, electronic equipment and a storage medium, and the method comprises the steps: obtaining a graph data set of a graph layer from a graph data file of an integrated circuit, determining the total number of graphs according to the graph data set, dividing the graph data set into at least one graph data group corresponding to the graph layer, and converting each graphic data group into the first bitmap of the corresponding graphic layer, fusing all the first bitmaps corresponding to the same graphic layer to obtain the second bitmap, and converting the second bitmap into the pixel map of the corresponding graphic layer, so that the graphic data of the graphic layer can be processed on the memory level, and a new pixel map of the graphic layer can be quickly generated. In this way, it is not needed to traverse each data in all data sets for operation and draw each pixel point every time in the prior art, the operation time is shortened, and therefore large-batch graphic data of an integrated circuit can be loaded and displayed in time.

The application provides a reversible desensitization method for a visual rough visible sensitive image, and the method comprises the following steps: constructing a pixel matrix and dividing the matrix into blocks according to original image channel data and desensitization parameters; performing summation and encryption on the first two adjacent pixels of a first image block to obtain a ciphertext pixel group; then, the second pixel of the ciphertext pixel group and the next unencrypted pixel are combined into a new group, and the new group is iteratively encrypted by using a pseudo-random function and an encryption key until the pixels in the block are processed; a random number sequence is generated to rearrange the positions of the ciphertext block elements and complete permutation encryption; the current ciphertext block and the key are spliced and then hashed to derive the encryption key of the next image block, and all the blocks in the block and among the channels are processed in sequence; finally, the ciphertext blocks are reorganized according to the original channel order to obtain a desensitized image. The application forms a three-level encryption chain architecture which is mutually related and layer by layer transmitted; the architecture makes the pixels, image blocks and color channels of the image form strong encryption dependence at each level, and realizes the avalanche effect of the desensitization process.

The invention discloses a design network training method, a design method, a system, equipment, a medium and a product of a thin-plate type acoustic metamaterial, and relates to the technical field of acoustic metamaterial design, the method comprises the following steps: obtaining a plurality of thin-plate type acoustic metamaterial structures, mass blocks of each thin-plate type acoustic metamaterial structure being randomly distributed; preprocessing each sheet type acoustic metamaterial structure to obtain a sample data set; the sample data set comprises a mass distribution pixel image and a sound transmission loss curve corresponding to each thin plate type acoustic metamaterial structure; and training the bidirectional neural network model based on the sample data set to obtain a forward prediction network and a reverse design network. According to the method, the sound transmission loss curve can be quickly predicted, the matched mass block distribution structure can be efficiently generated, and the design efficiency is remarkably improved.

This application provides a wafer performance prediction method, apparatus, electronic device, medium, and product, relating to the field of semiconductor technology. The method includes: acquiring initial sensor data corresponding to at least one target process step of the wafer to be processed; filling the initial sensor data with data to obtain target sensor data; the initial sensor data includes process parameters collected at different sampling time points; converting the target sensor data to obtain multiple target pixel images; the wafer to be processed corresponds to at least two target pixel images for each target process step; inputting the multiple target pixel images into a trained machine learning model to obtain a target prediction region; the target prediction region is a region composed of electrical parameters affecting wafer performance; and determining the performance of the wafer to be processed based on the target prediction region. The above process, by combining sensor data, image conversion, and machine learning technology, achieves effective prediction of wafer performance and improves the accuracy of prediction.

This invention discloses a hierarchical supervised super-resolution reconstruction method, system, device, and medium for remote sensing images. The method includes acquiring a high-resolution original remote sensing image, preprocessing it to obtain multiple sets of sub-pixel remote sensing images; constructing a sub-pixel remote sensing image reconstruction network, and pre-training it using a low-resolution remote sensing image and a set of sub-pixel remote sensing images as input and ground truth, respectively; constructing an upsampling-free remote sensing image super-resolution network, and optimizing it using multiple sets of sub-pixel images; constructing a multi-level supervised upsampling-free super-resolution network, and training it using an intermediate-resolution remote sensing image and a high-resolution original remote sensing image as ground truth, respectively; and performing super-resolution reconstruction based on the remote sensing image super-resolution reconstruction model, inputting the remote sensing image whose resolution needs to be improved, to obtain a high-resolution remote sensing image. This invention, by employing a hierarchical supervised and upsampling-free super-resolution reconstruction deep network, can improve the quality and detail of super-resolution reconstruction of remote sensing images.

The disclosure provides a stereoscopic display device and a composite optical film material, and belongs to the technical field of stereoscopic display. The composite optical film material is used for being attached to the light emitting surface of a flat panel display panel, and comprises a taking image layer, a light transmitting medium layer and a stereoscopic grating layer which are sequentially stacked. The taking image layer comprises taking image columnar lenses which are sequentially arranged and correspond to pixel combinations of the flat panel display panel one by one. The taking image layer is used for imaging the pixel image displayed by each pixel combination on a real image plane to form a real image on the real image plane. The real image plane is located in the light transmitting medium layer. The stereoscopic grating layer is used for projecting the real image on the real image plane to an external space to realize stereoscopic display. The composite optical film material can improve the field of view width at a viewing distance.

The technology described herein relates to aspects of systems, methods, and computer-readable media for detecting motion in a scene. A first temporal pixel image is generated based on a first set of images of a scene over time, and a second temporal pixel image is generated based on a second set of images. One or more derived values are determined based on temporal pixel values in the first temporal pixel image, the second temporal pixel image, or both. Correspondence data is determined based on the first temporal pixel image and the second temporal pixel image to represent a set of correspondences between image points of the first set of images and image points of the second set of images. An indication is determined based on the one or more derived values and the correspondence data, the indication to indicate whether motion occurred in the scene.

The method of the embodiment of the application proposes a multi-modal data fusion postoperative infection early warning method and system. First, the postoperative wound image, clinical text, wearable vital signs and baseline features are quality evaluated and repaired. The repaired image is double-flow encoded and the coarse mask is obtained by prior segmentation. The superpixel graph is constructed and the topologically consistent fine mask is obtained by the graph neural network to quantify the erythema / effusion area. The text and vital signs are encoded and fused with the image features and the mask through cross-attention and self-attention, and the fused hidden vector and uncertainty are obtained through the quality weight gate. The individualized dynamic threshold is constructed by combining the baseline risk, the logit calibration and the sliding window trend, the shape-preserving prediction output confidence band is used to realize the early and reliable infection early warning.

A display device includes: a substrate including a plurality of pixels having a plurality of sub-pixels; a pattern portion disposed on the substrate and disposed between the plurality of sub-pixels in the non-emission area in a recessed manner; and a reflective portion disposed on the pattern portion, in which each of the plurality of sub-pixels includes a pixel electrode disposed to be spaced apart from the reflective portion, and the pixel electrode includes a main body and a plurality of protruding portions protruding from the main body toward the reflective portion.

The invention discloses a mine panoramic monitoring system based on array type hundred-million-level pixel image intelligent perception. The mine panoramic monitoring system comprises an array image acquisition module which acquires multiple paths of original video signals of a mine scene through a plurality of hundred-million-level pixel cameras which are spatially arranged; the edge preprocessing module receives multiple paths of original video signals and generates preprocessed video signals and preliminary target signals; the panoramic splicing module receives the preprocessed video signals and fuses multiple paths of video streams through a real-time image splicing algorithm to generate unified panoramic video signals; the intelligent sensing module receives the panoramic video signal and the preliminary target signal and generates an intelligent analysis result signal; and the early warning and interaction module receives an intelligent analysis result signal so as to realize alarm output and user operation feedback. According to the mine panoramic monitoring system based on array type hundred million-level pixel image intelligent perception, the problem that view, precision and real-time intelligent analysis cannot be considered in mine panoramic monitoring can be solved.

The invention discloses a portable line displacement measuring instrument and a measuring method, and the measuring instrument comprises a measuring camera which is used for collecting an image containing a coding point; the coding scale serves as a physical length reference, and a plurality of annular coding points are arranged on the coding scale; the computing terminal is in communication connection with the measuring camera and is used for receiving the image and operating displacement measuring software; wherein the displacement measurement software is configured to execute the following steps: identifying and decoding annular coding points in an image; feature matching among multiple images is carried out based on the uniqueness of decoding points; obtaining the three-dimensional coordinates of the coding points through relative orientation and a three-dimensional reconstruction technology; restoring the real scale of the three-dimensional coordinate by using the coded scale; and calculating the longitudinal displacement of the rail waist coding point based on the foundation pile coordinate system. According to the method, the sub-pixel image processing and high-precision three-dimensional reconstruction technology is adopted, scale normalization is carried out in combination with a coding scale, and the measurement precision of + / -1 mm is achieved.

A pixel driver arranged to drive a display device (e.g. phase modulator, LCOS SLM) to display a first and second colour pixel patterns on an array of pixels at the same time. The driver comprises n da

An organic light-emitting display device and a method for manufacturing the same are disclosed. The organic light-emitting display device includes a driving thin-film transistor on a substrate; an insulating film covering the driving thin-film transistor except for a portion thereof; a planarization layer disposed on the insulating film and having a pixel contact hole defined therein exposing a portion of a surface of an electrode; a first electrode disposed on the planarization layer and extending on an inner surface of the pixel contact hole so as to connect to the transistor via the electrode; a bank disposed on the planarization layer so as to define a light-emitting area of a pixel; a plurality of pixel patterns disposed on the first electrode; a protective pattern filling the pixel contact hole and having a trench-shaped space; and a photoresist plug pattern filling the trench-shaped space.

The invention is applicable to the technical field of display, and provides a display device and a display method of the display device, and the method comprises the following steps: obtaining to-be-displayed subtitles from a main thread; the subtitles to be displayed are subjected to rendering processing through a subtitle thread, subtitle images corresponding to the subtitles to be displayed are obtained, the subtitle images are stored in a set cache area, and the rendering processing is used for converting the subtitles to be displayed into two-dimensional pixel images; and obtaining the subtitle image from the cache region through the main thread, and performing drawing processing on the subtitle image so as to display the subtitle image on a display. The display effect and the interaction experience of the display equipment can be improved.