36 results about "Image separation" patented technology

This invention discloses a laser image signal-to-noise ratio (SNR) enhancement method based on single-camera dual-channel imaging, relating to the fields of optical microscopy, digital image processing, and machine learning. The method includes: acquiring the original image from a single camera; separating the original image into a green channel image and a red channel image; constructing a denoising model based on the linear relationship between the green and red channel images, using the minimization of background noise in the green channel image as the objective loss function of the denoising model, and solving for an optimized parameter set; processing the green and red channel images according to the optimized parameter set to obtain a high SNR laser image after removing fluorescence background noise. This invention can remove fluorescence background noise from the original image acquired by a single camera without increasing hardware costs, effectively eliminating spectral crosstalk in the laser channel and significantly improving the laser image signal-to-noise ratio.

The application relates to the technical field of surface defect intelligent detection, and discloses a surface defect detection method and device based on multi-view synthetic imaging. The method comprises the following steps: acquiring a multi-view synthetic image without dead angles of a detected target after quality evaluation; performing highlight suppression and shadow enhancement on the multi-view synthetic image to obtain an enhanced multi-view synthetic image; separating the enhanced multi-view synthetic image into corresponding multiple single-view images, respectively performing defect detection on each single-view image to obtain corresponding single-view defect detection results, and generating a final overall detection result based on the single-view defect detection results. Through multi-view synthetic imaging, image quality control, high-reflective surface enhancement processing and multi-view joint defect detection, the application can effectively solve the problems of light interference and defect missing detection in high-reflective surface detection, and can realize high-precision surface defect detection with no dead angle coverage in a single detection.

This invention discloses an integrated broadband communication and stereo remote sensing method based on multi-satellite cooperation. For a ground target area, a central satellite and several neighboring auxiliary satellites are selected from a low-Earth orbit (LEO) satellite constellation. Each satellite collaboratively transmits radio frames towards the joint beam coverage area. Ground users receive the shared pilot signal embedded in the radio frames to obtain channel state information and recover data. Simultaneously, the central satellite receives and stores the ground echo signal corresponding to the shared signal and performs channel estimation. Combining the Doppler center differences between echoes from different cooperating satellites, multi-view range image separation is achieved. Furthermore, a stereo synthetic aperture radar (SAR) imaging method is used to calculate the three-dimensional geographic coordinates of reflectors in the target area. This method achieves integrated collaboration of communication and remote sensing functions within a LEO satellite constellation, contributing to improved constellation resource utilization efficiency and three-dimensional observation capabilities of the target area.

The invention relates to the technical field of PCB production detection, and discloses a method and a system for PCB multilayer board alignment deviation identification. According to the method, the multi-layer target theoretical projection image is generated through physical modeling, frequency domain analysis is carried out in combination with the X-ray detection image and the multi-layer target theoretical projection image, and quantitative evaluation of the reliability of different areas of the image is achieved through the obtained pixel-level credibility weight matrix. Further, through a method of deeply combining credibility evaluation and image separation, noise in the image is effectively suppressed, structural information of a high-credibility region is reserved, and accurate and reliable interlayer alignment deviation data can be provided for alignment deviation recognition through the obtained single-layer target separation image and target design parameters; the problem that the alignment recognition result is inaccurate due to aliasing distortion when an X-ray detection image is directly analyzed by a traditional method is solved, and the recognition accuracy of the alignment deviation condition of the PCB is remarkably improved.

In order to solve the technical problems that the existing droplet observation device cannot monitor the multi-nozzle printing process and the overall observation precision is low, the present application provides a multi-nozzle single droplet real-time observation device, an observation method and a 3D printing system.The observation device of the present application can make the observation camera capture the current working printing nozzle and automatically focus on it in real time through the cooperation of the first steering engine and the displacement platform, and complete the real-time monitoring of the printing state of the printing nozzle; through the high-frequency multiple flashing of the LED flash, the observation camera can capture the images of the same droplet at multiple times in one shot, at the same time, the high-frequency deflection of the piezoelectric deflection mirror and the telecentric lens are matched with each other, the overlapping images of the single droplet at adjacent moments are separated, so that a single photo can have enough droplet images to show the forming process, the overall observation precision is improved, the quality monitoring and fault diagnosis problems during additive manufacturing are improved, and the waste rate of electronic 3D printing is reduced.

The invention provides an image processing method and device, equipment and a medium, and relates to the technical field of artificial intelligence, in particular to the technical field of computer vision, deep learning and the like. The method comprises the steps of identifying a main body object corresponding to a target area on an original image in response to a detected input target operation for the original image; performing expansion operation on the main body object to obtain an adjacent area of the main body object; inputting a region identification graph for identifying the positions of the main body object and the adjacent region in the original image and the original image into an image matting model to obtain a foreground mask; separating the original image into a foreground image and a background image based on the foreground mask; performing blurring operation on the background image to obtain a blurred background image; combining the fuzzy background image and the foreground image into a preview image; and outputting the preview image.

The application discloses a multimodal medical image heterogeneous signal synchronous fusion method and device and system, comprising the following steps: S1, acquiring a DSA video stream and an IVUS video stream; S2, performing cross-modal medical image registration based on a space-time Transformer according to the DSA video stream and the IVUS video stream, and realizing medical image heterogeneous signal synchronous fusion. The technical scheme of the application solves the problems of image separation, inaccurate positioning, system delay and the like in the prior art, and improves the precision and safety of interventional surgery.

The invention relates to a structure-semantic double-graph collaborative SAR image quality evaluation method, which comprises the following steps of separating an SAR image, and constructing a double-channel quality measurement set for independent modeling; indexes are obtained, correlation calculation is carried out, an index map is constructed, and each node corresponds to one index; on the basis of the index map, the dependency relationship between indexes is calculated from the structure dimension and the semantic dimension through self-supervised learning, the dependency relationship of the two dimensions is integrated, and an image level score based on the integration result is output; and constructing a measurement index shared with system analysis and a measurement index special for information retrieval, and carrying out quality measurement evaluation: after extracting a structure sub-graph, integrating an IR special index which is not contained in SAR prior into a global graph. The method is superior to a reference model in the aspects of recognition accuracy, recall rate, precision and efficiency, and a solution with high universality and good interpretability is provided for cross-modal image quality evaluation.

The invention discloses a single-camera partitioned optical microscopic imaging device which comprises a bottom plate, a support is mounted at the top of the rear side of the bottom plate, and a mounting plate is mounted on the front side of the support; the sample table is mounted on one side of the top of the bottom plate and is used for clamping a sample; the transmission illumination assembly is mounted on the other side of the top of the bottom plate and is used for providing uniform illumination for the sample; and the microscopic imaging assembly is mounted at the bottom of the front side of the mounting plate. According to the invention, through the arrangement of the image separation assembly, the intermediate image of the sample can be separated into microscopic images of a plurality of channels, and the microscopic images can be imaged on the target surface of the camera in real time after being transmitted through different paths, so that a single camera can carry out multi-channel imaging without considering whether a plurality of cameras are aligned or not; according to the optical microscopic imaging device, the problem of difference of time resolutions of a plurality of cameras does not exist, the optical path of the optical microscopic imaging device is relatively simple in the propagation process, some very rapid dynamic events can be synchronously captured, and the use effect is good.

The application provides a multimodal fidelity hallucination detection method and system based on closed-loop multi-dimensional verification, belonging to the technical field of large visual language models, comprising: S1: decomposing the user instruction input into the LLM and the answer of the LLM into structured units; S2: through the collaborative mode of the LLM and the visual tool, the consistency score of the assertion is calculated from three dimensions of text-instruction consistency, text-image consistency and logical consistency, all verification results are bound to a unique evidence ID, and the assertion without evidence support is determined as unknown; S3: performing type-specific calibration on the consistency score, and then generating a unified hallucination score H through weighted fusion; S4: providing traceable evidence support for the detection result, and triggering the abstention mechanism when the evidence is insufficient, so as to ensure the credibility and practical value of the detection result. The method realizes accurate detection of three types of hallucinations, i.e., instruction inconsistency, text-image separation and logical inconsistency.

This application provides an image processing method, electronic device, and program product, relating to the field of astrophotography technology. The image processing method includes: performing background target removal processing on an acquired image containing a high-speed moving target using a depth model to obtain a background target image and a moving target image; overlaying multiple moving target images to obtain a first overlay image; overlaying multiple background target images to obtain a second overlay image; and fusing the first and second overlay images to obtain a target image. This application can separate an original acquired image containing a high-speed moving target into a background target image and a moving target image, perform overlay processing on each separately, and then fuse the two overlay images to obtain the final target image, effectively improving the quality of the target image and meeting the user's needs for capturing and imaging high-speed moving targets.

The application relates to the technical field of image processing, and discloses a platform for separating and tracing comprehensive identification and analysis of image and text elements, which comprises the following modules: an image preprocessing decoupling module, which is used for separating a digital document matrix to obtain a pure text layer and calculating a physical degradation index; a semantic extraction topology module, which is used for extracting text from the layer to construct an interlingual translation directed graph; a closed-loop reconstruction module, which is used for performing structure verification and inversely deducing parameters to reconstruct image separation parameters; a degradation weight fusion module, which is used for encoding the physical degradation index into the directed graph to generate a joint tracing topology graph; and a tracing identification module, which is used for calculating a graph edit distance based on the topology graph and outputting a quantitative numerical value to draw a conclusion. The application accurately distinguishes image conversion errors and local tampering operations by guiding the repair of image and text element separation parameters through a semantic closed loop and integrating physical states into a topology structure, and reliable tracing comprehensive identification and analysis is realized.

The invention relates to the technical field of portrait acquisition and comparison, and discloses a portrait acquisition and comparison method and system for frontier inspection, and the method comprises the steps: obtaining an original intensity image in a preset polarization direction, and obtaining an original intensity image based on a Stokes vector principle and a two-color reflection model; separating the collected image into a specular reflection component representing surface texture and a diffuse reflection component representing subcutaneous tissue characteristics; spatial frequency domain energy distribution of the two components is calculated, Boltzmann distribution is used for weighting and determining an overlapping characteristic value of a frequency domain of a diffusion mirror surface, and a light scattering scale parameter of the current skin is inversed according to the overlapping characteristic value; constructing a tissue optical characteristic filter, modulating the diffusion component in a frequency domain so as to map the diffusion component to a preset standard light scattering scale, and combining the mirror component subjected to energy whitening equalization to synthesize a normalized face image; and generating a binary final verification result by calculating the feature vector cosine similarity of the normalized image and the certificate reference image and combining an abnormal deviation penalty term.

The invention provides a multi-modal loyalty illusion detection method and system based on closed-loop multi-dimensional verification, and belongs to the technical field of large visual language models.The method comprises the steps that S1, a user instruction input into an LLM and an answer of the LLM are decomposed into structured units; s2, through a collaborative mode of LLM and a visual tool, calculating consistency scores of assertions from three dimensions of text-instruction consistency, text-image consistency and logic consistency, binding all verification results with a unique evidence ID, and determining that the assertions without evidence support are unknown; s3, executing type exclusive calibration on the consistency score, and generating a unified illusion score H through weighted fusion; and S4, providing a traceable evidence support for the detection result, and triggering a discarding mechanism when the evidence is insufficient, so as to ensure the credibility and practical value of the detection result. According to the method, accurate detection of three types of illusions of instruction inconsistency, text-image separation and logic inconsistency is realized.

Provided is a method for generating separated images, said method comprising: a selection step for selecting intended image separation information from a storage unit that stores a plurality of pieces of image separation information for obtaining separated fluorescent images, which are fluorescent images separated for different wavelengths; an acquisition step for irradiating a sample with each of a plurality of wavelengths of excitation light, and acquiring, for each of the plurality of wavelengths of excitation light, an observation fluorescent image related to fluorescence generated from the sample; a first conversion step for, on the basis of the image separation information selected in the selection step, converting the plurality of observation fluorescent images acquired in the acquisition step into a plurality of separated fluorescent images; and a generation step for performing an update process for the plurality of separated fluorescent images obtained in the first conversion step to have increased independences, respectively, and generating updated separation information.

The application provides a superpixel-based resolution improvement method and a scanning controller, relates to the technical field of image display, and can improve the resolution without increasing the size of a display screen, multiplexes information of adjacent pixels, makes one light-emitting diode participate in imaging of multiple adjacent pixels, and meets the requirements of ultrahigh resolution and refresh rate of a micro display; the method comprises the following steps: S1, performing image processing on an original image; S2, determining the number of frames for separating the original image according to the spatial resolution of the original image; S3, separating the original image into a plurality of frame sub-images according to the number of frames determined in S2 and the row and column positions of each pixel in the original image, and meanwhile adjusting the gray intensity of each pixel color channel according to the separation ratio; S4, expanding the pixels in the sub-image by dividing sub-pixels; and S5, superimposing and merging the plurality of expanded frame sub-images to obtain a high-resolution image.

A display method includes: creating a global container and a virtual display; determining a display-hidden state of a target pixel; performing image splitting processing on a sub-image whose display-hidden state is a display state in the image to be displayed based on a naked-eye 3D film parameter of the virtual display, to obtain left-right eye views; in a case that the display-hidden state of the target pixel is the display state, determining a current refresh position of the target pixel; performing image interleaving processing on the left-right eye views to obtain a first vector texture; updating the first vector texture according to the current refresh position to obtain a second vector texture after rotation; and pushing the second vector texture to the global container, and outputting the second vector texture to the virtual display based on the global container, to obtain a target stereoscopic image.

The application discloses an interpretable harsh environment image enhancement method, system, device and medium, and relates to the technical field of image processing.The method comprises the following steps: inputting a harsh environment image; separating the harsh environment image into a background layer and a noise layer; denoising the noise layer to obtain a denoised layer; and obtaining an enhanced harsh environment image according to the background layer and the denoised layer.The application introduces a neural network unfolding technology, and unfolds a traditional optimization iterative algorithm into a deep neural network, wherein each layer of the neural network can be regarded as a solving process of the optimization iterative algorithm, so that a deep neural network operator with certain interpretability is constructed, and an interpretable harsh environment image enhancement algorithm is realized through combination of the interpretability operator.

This invention discloses a method for detecting defects in concrete dams, belonging to the technical field of dam crack detection. The method includes: acquiring an image of the concrete dam; performing environmental adaptive compensation adjustment on the concrete dam image; preprocessing the concrete dam image; separating the concrete dam image into crack regions and background regions; extracting the contours of the crack regions; geometrically filling the contours of the crack regions to obtain key feature values ​​of the cracks; and evaluating the crack hazard based on the key feature values. This invention not only reduces the impact of environment, lighting, and weather on the captured images but also suppresses transmission noise, effectively improving image capture and transmission quality. Furthermore, the geometric filling of the crack region contours enables rapid crack hazard evaluation. The entire method is computationally simple, fast, has low computational cost, low resource consumption, and is adaptable to processing various types of dam image data, exhibiting strong generalization ability.

This application discloses a method and apparatus for evaluating and processing retrieved presentation documents. The method includes: parsing the presentation document to obtain its text information, visual elements, and structural information, wherein the text information includes presentation text and / or presentation notes; performing multi-dimensional evaluation processing based on the text information, visual elements, and structural information to obtain at least two of the following: a text-image separation evaluation result, a context integrity evaluation result, a structured degree evaluation result, and an information density to noise ratio evaluation result; and weighting at least two of the following: a text-image separation evaluation result, a context integrity evaluation result, a structured degree evaluation result, and an information density to noise ratio evaluation result, to obtain a comprehensive evaluation result for the presentation document.

A method for classifying plants for agricultural purposes. The method includes: detecting a field section having plants using an optical detection unit inclined toward the field section to obtain image information having a rear image edge having an image width and a front image edge having a larger image width; classifying the plants in the image information into first and second plant classes using an algorithm to treat the plants of the second plant class with a liquid spray agent, in accordance with the classification. The step of classifying comprises a step of reclassifying, in which classified plants are reclassified from the second plant class to the first plant class using the algorithm if they extend in a rear image section of the image information defined by a defined image separation line from a lateral image edge of the image information maximally up to a defined reclassification boundary line.

The invention relates to the field of cloth detection, and discloses a fabric inspection system for identifying defects of waste fabrics, comprising: an acquisition module for constructing a multi-modal data matrix; the processing module is used for performing enhanced denoising processing on the multi-modal data matrix; the decomposition module is used for performing local texture complexity analysis on the feature image; the enhancement module is used for outputting an enhanced feature image; the separation module is used for constructing a high-dimensional feature matrix and outputting a pure defect feature set; the detection module is used for detecting an image corresponding to the pure defect feature set; and the optimization module is used for constructing a Bayesian posterior probability model by taking the detection result as an observation value and carrying out iterative optimization. A multi-modal data matrix is constructed based on multispectral polarization acquisition, high-frequency defect features are adaptively screened, a pure defect feature set is extracted, identification and positioning are performed, and different working conditions are adapted through Bayesian model iterative optimization, so that the feature difference between tiny defects and a background is enhanced, and accurate detection of the defects of the waste fabric is realized.

The invention relates to the technical field of intelligent detection of surface defects, and discloses a surface defect detection method and device based on multi-view synthetic imaging. The method comprises the following steps: acquiring a dead-corner-free multi-view synthetic image of a detected target after quality evaluation; performing highlight suppression and shadow enhancement on the multi-view synthetic image to obtain an enhanced multi-view synthetic image; and separating the enhanced multi-view-angle composite image into a plurality of corresponding single-view-angle images, performing defect detection on each single-view-angle image to obtain a corresponding single-view-angle defect detection result, and generating a final overall detection result based on each single-view-angle defect detection result. Through multi-view synthetic imaging, image quality control, high-reflective surface enhancement processing and multi-view joint defect detection, the problems of illumination interference and defect missing detection in high-reflective surface detection can be effectively solved, and high-precision surface defect detection without dead angle coverage can be completed through single detection.

This invention discloses a method, system, device, and medium for adversarial augmentation of distribution network inspection data. The method includes: separating target equipment and defective components under the target equipment from a real-world distribution network inspection image to obtain a real component image; performing defect simulation on the real image of the target equipment to obtain a rendered component image; performing style transfer on the real component image and the content image to obtain a rendered composite image; training a generative adversarial network (GAN) based on the rendered composite image and the real component image; generating a defect local image containing defective equipment and defective components under the defective equipment corresponding to a given defect type based on the trained GAN; and replacing the defect local image with the corresponding equipment location in a normal distribution network inspection image to obtain an augmented sample of the distribution network inspection image containing defects. This achieves rapid and high-quality augmentation of distribution network line defect samples.

This invention discloses a method, apparatus, device, and medium for detecting color moiré patterns in micro-coded images. The method includes: acquiring a target micro-coded image to be detected and separating the target micro-coded image to obtain single-channel feature maps; generating an extreme value difference feature map based on the difference between the maximum and minimum pixel values ​​at the same pixel position in each single-channel feature map; and determining whether moiré patterns exist in the target micro-coded image based on the extreme value difference feature map. By constructing feature maps using the extreme value differences of each channel at each pixel position in the image, the inter-channel differences caused by moiré patterns are quantized into a single, high-contrast image, effectively amplifying the moiré signal and suppressing irrelevant background. This provides a more direct and sensitive feature for moiré detection, thereby enabling efficient and accurate automatic determination of the existence of moiré patterns in micro-coded images and expanding the application scenarios of moiré detection technology.

This invention relates to the fields of image processing and industrial inspection, and provides a quantum-level image reconstruction and dynamic spectral restoration method, comprising: acquiring an original row scan image obtained from a single camera scan, containing information on various exposure pixels arranged in an interleaved column pattern; separating the original row scan image into various sparse exposure images based on a preset pixel arrangement rule; using the original row scan image or a preliminary interpolated image as a guide image, performing guided filtering on the various sparse exposure images to fill in missing pixels, generating various spatially aligned and uniformly reconstructed images with different exposures; and performing weighted fusion on the various reconstructed images with different exposures to generate a fused image, wherein the weights are adaptively determined based on the brightness information of the pixel positions. This invention avoids temporal misalignment and registration errors from multi-frame acquisition, improves acquisition efficiency, and effectively suppresses artifacts and enhances details in dark areas through guided filtering and adaptive fusion.

The embodiment of the invention discloses a forest fire smoke image generation model and a forest fire smoke image generation method, and the model comprises an image separation module, a smoke concentration fine adjustment module and an image synthesis module. The image separation module is used for separating a background image and a pure forest fire smoke image from a to-be-processed image; the smoke concentration fine adjustment module is used for performing fine adjustment on the smoke concentration of the input pure forest fire smoke image to generate a pure forest fire smoke image containing the target smoke concentration; and the image synthesis module is used for synthesizing a forest fire smoke image by means of the background image and the pure forest fire smoke image, the forest fire smoke image with the controllable smoke concentration can be generated, high fidelity and definition are achieved, and the model training stability is guaranteed.

The application relates to an intelligent paper marking method, a printed data processing method and device, and an image forming apparatus. The intelligent paper marking method comprises the following steps: obtaining a first image of a paper answered by a user; obtaining a paper description list corresponding to the paper, the paper description list being generated in a process of printing an un-answered paper, and the paper description list being used to record a corresponding relationship between a test question content and a position of the paper; separating the answering content of the user from the first image according to the paper description list; and marking the paper based on the paper description list and the answering content of the user. The technical scheme of the application utilizes the printing process of the paper to improve the efficiency of automatic paper marking.

The embodiment of the invention provides an image speckle noise suppression method, device and equipment for industrial detection. The method is applied to the technical field of image processing and computer vision, and comprises the following steps: acquiring an industrial detection image containing multiplicative speckle noise, and performing logarithmic transformation on the acquired industrial detection image to obtain an image after logarithmic transformation; separating the image after logarithmic transformation into an illumination component and a reflection component based on a Retinex theory; calculating a gradient field of the illumination component, and performing bilateral filtering processing on the illumination component in a gradient domain; performing differential operation on the image subjected to logarithmic transformation and the filtered illumination component to obtain an enhanced reflection component; and executing an exponential inverse transformation operation on the enhanced reflection component to obtain an output enhanced image. According to the method, speckle noise can be effectively removed in a complex industrial detection scene, the natural authenticity of workpiece textures and structures is kept, and the image definition and contrast are remarkably improved.