92 results about "Low contrast" patented technology

The application discloses a kind of based on jigsaw puzzle's flat large-size printing plate detection method, it is related to the technical field of printed matter layout machine vision detection, the method first single-step detection image is spliced into complete product graph, using grid division and NCC template matching extract adjacent two graph high reliability point pair, and combined with K-means clustering, remove abnormal point pair, solve homography matrix, using homography matrix carries out accurate perspective transformation and image splicing, again carries out rotation scaling transformation and makes template graph and product graph alignment, and difference configuration parameter is set in detection area, using the dual comparison dimension detection logic of chromatic aberration and edge detection combination accurately determines defect position;The application can effectively eliminate the false matching caused by complex background and similar texture, solves the problem of splicing misalignment;Real defect is accurately locked by double-dimension detection, effectively suppresses false alarm, significantly improves the capture ability of low contrast small defect and geometric deformation defect, ensures the reliability of detection under complex environment.

The present application belongs to the technical field of endoscope and image processing, and in particular to an endoscope image anti-shake method based on motion vector compensation, comprising the following steps: collecting a sequence of continuous video frames, performing stability screening on the initial feature point set, eliminating unstable feature points located in the edge region, high light reflection region and low contrast region of the image to form a reliable feature point set, performing robust estimation processing on the original motion vector set based on the matched feature points in the reliable feature point set to generate a stabilized output frame, wherein the remapping adopts a bilinear interpolation algorithm; by constructing a closed-loop anti-shake system taking local feature motion vectors as input, spatio-temporal domain joint filtering as core and sub-pixel remapping as output, high-precision separation and compensation of high-frequency micro-shaking in endoscope imaging are realized without introducing additional mechanical structures.

This invention provides a method and apparatus for detecting surface defects in steel based on an improved YOLOv11. The method includes: integrating an EA module into the backbone network of YOLOv11, combining the Sobel operator edge detection idea with an attention mechanism, significantly enhancing the network's ability to perceive the geometric edges of defects, effectively improving the detection performance of low-contrast defect-background boundaries, and enhancing the ability to identify key defects such as cracks; by integrating three MSA modules into the neck network, the model's ability to capture multi-scale information in the feature space is significantly improved; the special position design of the EA module enhances the network's attention to horizontal and vertical gradient features during the final fusion of multi-scale feature representations, avoiding excessive interference with original features or highly abstract features; and achieving end-to-end processing from the original image to the defect detection result, significantly improving the detection performance of low-contrast defects and cross-scale defects on the surface of steel equipment.

The application discloses a wafer defect detection method and system based on edge features and gradient boosting trees, and belongs to the technical field of semiconductor manufacturing. The method first carries out denoising and gray scale normalization preprocessing on a wafer image, then extracts gradient, direction consistency, texture and edge density features through multi-scale edge detection to represent the difference characteristics between low-contrast linear defects and point defects and construct a pixel-level feature vector; gradient boosting tree model is used to realize pixel-level defect classification, and the final result is output in combination with morphological processing and connected domain geometry checking. The application can represent the difference characteristics between low-contrast linear defects and point defects, and is suitable for on-line defect detection of semiconductor wafers.

The application discloses a Down syndrome facial deformity classification method based on structure perception heat map and focused selective convolution. The technical scheme comprises the following steps: a structure perception attribute (SPA) module generates a structure mask by using facial key points and uses the structure mask as a semantic prompt to compensate for the lack of spatial perception of traditional CNN and CLIP; an abnormality detection region alignment (ADRA) module introduces an adapter on multi-level features and combines a zero sample / less sample double branch to strengthen the feature alignment of weak supervision and low contrast regions; and a selective feature extraction and classification (SFEC) module performs channel normalization and threshold clipping on the heat map, and uses fast / slow branch selective convolution to improve the discrimination and reduce redundant calculation. The method still maintains stable accuracy under changes in shooting distance and posture, significantly improves the accuracy and F1 of the existing method on the Down syndrome and facial paralysis data sets, and is suitable for early screening, objective evaluation and follow-up monitoring in clinical practice, especially for automatic diagnosis assistance in resource-limited scenarios.

The application discloses a polyp segmentation method based on SAM boundary enhancement and uncertainty guidance, aiming at the problems of high pixel-level labeling cost, blurred polyp boundary, difficult cross-scale feature alignment and poor effect of SAM direct migration to a medical scene in the polyp segmentation task under enteroscopy, fusing the hierarchical feature extraction ability of Res2Net-50, the general segmentation ability of SAM, combining core strategies such as boundary enhancement, uncertainty guidance and dynamic prompt generation, and realizing high-precision polyp weakly supervised segmentation under extremely sparse scribble labeling. The application significantly improves the spatial calibration accuracy of the model in a low-contrast environment, eliminates noise interference, and effectively solves the problem of supervision imbalance caused by the large size difference of lesions in an endoscope image.

This invention discloses a deep learning-based intelligent detection and classification method for wafer defects, belonging to the field of computer vision technology. First, a multi-scale pyramid decomposition is performed on the original wafer grayscale image. Then, a Fast Fourier Transform is performed on each layer of the image in the frequency domain. An adaptive sliding window detector and Hough transform are used to accurately identify and locate frequency peaks representing periodic background. Next, an adaptive notch filter is constructed and applied to suppress these periodic background frequencies, and non-periodic defect signals are explicitly enhanced by calculating frequency domain residuals. Finally, the multi-scale processed defect enhancement residual maps are fused to generate the final defect enhancement map, which serves as the input to the deep learning detection network. This invention solves the problem in existing technologies where low-contrast micro-defects are easily obscured by the strong periodic background texture of the wafer, leading to low detection accuracy and poor model generalization ability.

The application discloses an engineering inspection image disease automatic detection system and method, and relates to the technical field of image processing. The system comprises a user interface layer, a control logic layer, an image processing layer, a model inference layer and a result output layer. The method comprises the following steps: performing system initialization and parameter configuration; classifying engineering inspection images into small-size images and large-size images; using the model inference layer and the image processing layer to perform image intelligent processing, fine detection, result merging and block artifact elimination on the large-size images to obtain final detection results; using the image processing layer and the model inference layer to perform image enhancement and detection on the small-size images to obtain final detection results; and using the result output layer to perform visual display. Through the coarse-to-fine strategy and the frequency domain enhancement technology, the application effectively reduces calculation redundancy, improves detection accuracy and engineering practicability for low-contrast and irregular diseases, and eliminates the detection discontinuity problem caused by block artifacts.

The application discloses a photovoltaic surface defect detection system and method based on image recognition processing. The system comprises a feature extraction network, a hybrid encoder and a decoder. The feature extraction network comprises a feature enhancement module adopting a convolution and additive self-attention hybrid structure, which is used for extracting multi-scale features. The hybrid encoder is composed of an intra-scale feature interaction module and a cross-scale feature fusion module. The intra-scale feature interaction module performs global modeling and noise suppression on high-level features through a self-attention unit and a frequency domain discriminative feedforward unit. The cross-scale feature fusion module adopts a fusion strategy based on a gating convolution unit to realize dynamic screening and fusion of multi-scale features. The method corresponds to the operation process of the system. The application is aimed at the characteristics of low contrast, multiple noises and defect superposition of EL images. Through the cooperative optimization of the above modules, the weak defect capture ability, the anti-interference ability and the multi-defect distinguishing ability are improved, and high-precision and high-real-time defect detection is realized.

The application discloses an online screening method and system for optical fiber active connectors based on deep learning, belongs to the technical field of optical fiber active connector detection, and aims at solving the problems that the traditional detection mode is prone to missing detection and false detection of submicron, low-contrast and edge-distributed tiny defects, and cannot determine defect positions and quantify defect information.The online screening system for optical fiber active connectors based on deep learning comprises a coaxial imaging acquisition module, an image preprocessing module, a deep learning detection module, an explainability judgment module, an execution control module, an online optimization module and a data interaction module.The application realizes high-sensitivity detection of micrometer and submicron tiny defects by combining deep learning deep feature extraction and attention mechanism with accurate parameters of an imaging and preprocessing unit, and the detection precision is far superior to that of traditional artificial and machine vision schemes.

This invention discloses a method and system for surface defect detection in the integrated circuit packaging process. The method includes the following steps: acquiring digital images of the surface of the integrated circuit packaging process; performing image preprocessing on the acquired digital images; extracting features from the preprocessed surface images of the integrated circuit packaging process; classifying the extracted images according to defect features; applying a corresponding defect detection method to each type of defect to effectively separate the normal background from the defect area; determining the defect area for each defect type; determining whether the defects meet industrial production requirements; and outputting the surface defect detection results. This invention preprocesses the surface defect images of the integrated circuit packaging process before performing defect detection, avoiding interference from noise, low contrast, and other factors on subsequent defect detection, thus improving the accuracy of defect detection.

The invention provides a multi-branch enhanced retinal vessel segmentation network method based on stimulation guidance, and belongs to the technical field of medical image intelligent diagnosis. The technical problems that small blood vessels are difficult to accurately identify and boundaries are not clear in retinal blood vessel segmentation are solved. According to the technical scheme, the method comprises the following steps: S1, collecting fundus color image data to be segmented; s2, constructing a boundary enhancement module; s3, constructing a multi-scale feature aggregation module; s4, constructing a stimulation guide gating fusion module; and S5, after model training is completed, for each to-be-segmented test image. According to the method, boundary gradient enhancement, multi-scale interaction and stimulation guide gating cooperate to improve boundary continuity and detail fidelity, thin blood vessel distinguishability is improved, and robustness under a complex background and weak contrast is enhanced.

This invention discloses a method, system, method, and manufacturing method for flexible components, belonging to the technical field of flexible component processing. It is applied to flexible component laminates with encapsulated electrical connections, and includes the following steps: S1: Irradiating the flexible component laminate with an excitation light source of a specific wavelength to induce photoluminescence in the cell array; S2: Acquiring the photoluminescence image formed after the cell array is excited using an image acquisition module; S3: Processing the photoluminescence image to identify the boundary between the photoluminescent and non-photoluminescent areas to determine the edge of the cell array. This application solves the technical problem of inaccurate and unreliable edge positioning of cell arrays in low-contrast or special front panel structures, achieving high-precision, non-contact edge positioning.

The application discloses a semiconductor manufacturing defect detection method, an electronic device and a storage medium, the method comprising: using a trained defect detection model to detect a scanning electron microscope (SEM) image to obtain a detection result; wherein the defect detection model is based on YOLOv8, and the overall loss function is a mixed loss function combining a classification loss and a regression loss; when the defect detection model performs feature extraction, frequency domain decoupling feature extraction is performed and spatial pixel-level feature fusion is performed, and a specially designed mixed loss function is combined to solve the problems of class imbalance and accurate positioning. Compared with the detection method of the prior art, the application has higher detection accuracy, stronger model robustness, and meets real-time requirements. While ensuring accuracy improvement, the inference speed that meets the production line beat is maintained, and finally, more robust and accurate automated detection of defects in semiconductor manufacturing, especially subtle and low-contrast defects, is realized.

The application discloses a code nail stamping defect detection method and system based on machine vision and belongs to the technical field of industrial automation quality control. The method comprises the following steps: acquiring multi-modal visual data such as two-dimensional bright field images, two-dimensional dark field images and three-dimensional profile data of code nails to be detected; pre-processing the multi-modal visual data to obtain bright field, dark field and three-dimensional image data suitable for a neural network model; inputting the three kinds of image data into a preset neural network model for defect recognition processing, outputting defect segmentation masks and defect category information; finally, quantitatively analyzing defects based on the output defect information, and determining whether the product is qualified according to engineering specification threshold values. The application can effectively solve the problem of insufficient detection capability for micro, low contrast and three-dimensional geometric defects in the prior art by fusing multi-dimensional and complementary visual information and using a specially designed neural network model for analysis.

The application relates to the technical field of computer vision and medical image processing, and particularly discloses a medical image segmentation method based on boundary enhancement and double decoders, a boundary enhancement network model is constructed and initialized: the network model as a whole adopts a double decoder architecture, and comprises a shared encoder, a segmentation decoder and a boundary decoder; a hybrid loss function is defined to train the model: in order to jointly optimize segmentation and boundary prediction, a hybrid loss function is used to supervise the two decoding branches; model training and parameter configuration: a three-dimensional medical image is acquired; medical image prediction and segmentation output: a three-dimensional medical image to be segmented is input into the trained boundary enhancement network model, and a final three-dimensional medical image segmentation result image is output; the application effectively strengthens the collaborative modeling between boundary and segmentation features, realizes more accurate contour positioning, reduces the boundary leakage phenomenon in an organ boundary fuzzy or low-contrast region, and improves the quality of multi-scale feature representation.

This invention proposes an image reconstruction method for physical degradation sensing of planar diffraction lens arrays. It utilizes a deep learning network integrating an optical degradation sensing mechanism to optimize sub-aperture feature fusion and weight allocation strategies, improving the imaging contrast and detail recovery accuracy of planar diffraction lens arrays. The method includes: acquiring a sequence of monochromatic sub-aperture images of the planar diffraction lens array carrying spatially non-stationary blur and diffraction degradation features; implicitly aligning the feature responses of each sub-aperture using an edge enhancement feature alignment module; constructing a pseudo-array feature representation; introducing a physically prior-guided weight modulation strategy during adaptive grouping upsampling, learning the distribution law of spatially heterogeneous point spread functions through degradation sensing units, and spatially adaptively modulating the feature contribution; and finally reconstructing a high-quality grayscale image. This invention significantly suppresses the diffraction blur and low contrast phenomena unique to planar diffraction lenses, better sensing optical degradation distribution and preserving high-frequency details in the image. It also provides a new approach for subsequent joint reconstruction research of arrayed lightweight imaging systems.

The application relates to the field of computer vision, and particularly discloses an image segmentation method and system, wherein a boundary distribution map is predicted by a multi-scale boundary distribution generation module as prior knowledge, a space-frequency domain joint supervision strategy is adopted, the boundary distribution map simultaneously approximates a real boundary in a spatial position and a spectral structure, false edge responses caused by local high-frequency interference are effectively inhibited, and the continuity and stability of the boundary are improved; an effective receptive field is expanded by a state space semantic enhancement module, and global semantic consistency is enhanced; a boundary-guided cross-scale decoder is combined with a boundary feature enhancement module, boundary-related responses are strengthened in multi-resolution fusion, and collaborative optimization of region positioning and boundary refinement is realized, so that accurate and complete polyp segmentation results are obtained under complex imaging conditions such as low contrast, various morphologies, and interference such as reflection and wrinkles, and reasonable calculation overhead is maintained.

The application discloses a polaroid defective product identification method and system based on machine vision, and relates to the fields of polaroid detection and machine vision technology; the application collects polaroid multi-band images through multi-spectral imaging and corrects registration, completes background separation through gray scale normalization, filtering, morphological operation and inflation corrosion reconstruction, adopts a partition adaptive contrast enhancement strategy to strengthen low-contrast defects and fuse multi-band images, extracts multi-dimensional feature vectors in multiple scales, matches and grades defects according to a preset defect type feature library, and generates a defective product identification report containing defect information; the application effectively solves the problem of low recognition rate of polaroid morphological changes and low-contrast defects, and significantly improves the accuracy, stability and detection efficiency of defect identification.

The application discloses a geosynchronous orbit SAR sea surface wave imaging quality evaluation method and belongs to the technical field of geosynchronous orbit space correlation SAR sea surface wave imaging quality evaluation. The application obtains a sea surface wave echo signal, performs distance direction matching filtering and space correlation processing on the sea surface wave echo signal, generates an imaging result by using a back projection algorithm, calculates image intensity and a sea wave spectrum main energy direction slope, and finally quantifies imaging quality by using a space correlation evaluation factor. The evaluation factor can effectively represent the space coherence of sea surface wave imaging, and the larger the value is, the better the imaging quality is. The application solves the problem of insufficient applicability of existing evaluation methods in sea surface wave imaging due to low contrast and weak texture features, significantly improves the accuracy of geosynchronous orbit SAR sea surface wave imaging quality evaluation, and provides reliable technical support for marine environment monitoring.

The application discloses a return scattering ionization map F area automatic segmentation method based on Haar wavelet downsampling, to solve the problems of fuzzy F area boundary, low contrast and high frequency information loss in the prior art. The method comprises the following steps: step 1, data acquisition, labeling and pretreatment; step 2, constructing a SegNext segmentation network based on Haar wavelet downsampling; step 3, decoding fusion and segmentation prediction output; step 4, training optimization based on edge weighted focus joint loss. The method retains the high frequency characteristics of the ionization map by using the wavelet downsampling module, and combines the focus loss and the edge weighting mechanism, so that the segmentation result is more accurate and stable in the boundary area.

Those features of the air fryer component depicted in the representations that are new and distinctive are the shape and / or configuration and / or pattern and / or ornamentation. The features shown in broken or dotted lines are not part of the design and are to be disregarded when considering the overall impression of the design. Any features shown in low contrast shading are not part of the design and are to be disregarded when considering the overall impression of the design. Shading indicates a grey brushed metal appearance that forms part of the design.

There is described a computer implemented method of image segmentation for identifying one or more vehicles in an image comprising a seascape. The method comprises inputting the image into a first neural network trained to perform image segmentation to generate a first segmentation result, inputting the image into a second neural network trained to perform image segmentation on low contrast images to generate a second segmentation result, wherein at least one of the first segmentation result and second segmentation result comprises a segmentation mask corresponding to a vehicle in the input image, and generating an output image in which the one or more vehicles are identified by corresponding segmentation masks by combining the first and second segmentation results. Using two differently trained neural networks allows for accurate segmentation in both foreground and background of the image.

The application provides a histogram equalization method for protecting details, comprising the following steps: S1, dividing an input luminance image into blocks; S2, counting a histogram of the image; calculating an average gray level of a current block; S3, histogram redistribution, comprising the following steps: calculating a clipping threshold; histogram clipping and redistribution; S4, calculating a mapping function; S5, adjusting the mapping function; and S6, enhancing the input image. The histogram equalization method solves the problem of over-enhancement in bright areas and dark areas, and ensures that low-contrast images will not lose the details of bright areas and dark areas after being enhanced.

This invention relates to the field of UAV pose map optimization technology, and particularly to a method for optimizing the pose map of a UAV monocular camera. The method includes: blurring different image frame data, constructing a Gaussian pyramid, and downsampling; determining candidate feature points by subtracting Gaussian blurred images of adjacent scales; accurately determining the positions of candidate feature points on the image frame; removing low-contrast candidate feature points and edge response feature points to obtain the final feature points; and using the LM algorithm to adjust the UAV attitude parameters and optimize the pose map for discontinuous image frame data. The process of Gaussian difference point interest detection on continuously acquired image frame data suppresses noise in the image frames, effectively smoothing noise in the image. The resulting pose map is smoother with smaller gradient changes, improving the robustness of attitude estimation. The pose map optimization process expands the scope of pose map optimization, allowing optimization even for discontinuous images.

The application discloses an underwater image enhancement method and device based on adaptive color-brightness embedding and content perception. The method first converts the input image to YCbCr space for feature decoupling using a color-brightness embedding learning module, and extracts the brightness prior feature and the color prior feature respectively; then the original image features are input into the content perception backbone network composed of a shallow detail branch and a deep content branch, and the local texture and global contrast of the image are recovered through dynamic filtering and Fourier domain modulation; finally, through an explicit multi-prior fusion module, the learned physical prior is injected into the enhanced features for collaborative refinement, and the color deviation is corrected and a clear image is reconstructed. The PSNR and SSIM indicators of the application on multiple data sets are better than those of the existing advanced algorithms, which can effectively solve the color deviation, low contrast and detail blur problems of underwater images, and have very high fidelity and robustness.

The application discloses a tobacco safety foreign matter identification method based on an image segmentation algorithm, comprising the following steps: S1, collecting an original image of tobacco to be detected; S2, preprocessing the original image to generate a standardized image; S3, extracting multi-scale features by using a ConvNeXt model to generate a feature image set; S4, inputting the feature image set into an improved Mask2Former model, outputting an instance segmentation mask through a scale perception query and a local texture attention mechanism; S5, calculating foreign matter categories and coordinates according to the mask; and S6, packing to generate a visual identification result data packet. Through the fusion of the powerful feature extraction capability of ConvNeXt and the accurate segmentation mechanism of the improved Mask2Former, the identification precision and robustness of small, low-contrast and texture-similar foreign matters are improved, the missing detection and false alarm problems of traditional methods in a complex industrial background are solved, and a high-precision detection scheme is provided for tobacco storage safety.

This invention relates to the interdisciplinary field of artificial intelligence and industrial vision inspection, specifically disclosing a nonwoven fabric defect detection system based on a deep learning convolutional neural network. The system includes a polarization image acquisition device, a Stokes parameter calculation unit, a physical perception feature extraction network, a dual-stream fusion decision module, and a defect localization output unit. By simultaneously acquiring images in multiple polarization directions and generating polarization degree and polarization angle images, combined with a physical perception network embedding optical priors to enhance the response to weak defects, and then dynamically integrating light intensity and polarization features through dual-stream fusion and a cross-channel attention mechanism, pixel-level defect classification and localization are ultimately achieved. Through the above technical solution, this invention improves the detection rate of low-contrast defects such as oil stains and transparent adhesive dots, reduces false alarms and missed detections, and possesses online adaptive calibration capabilities, ensuring the stability and reliability of long-term industrial field inspections.

The application discloses a coronary microvessel angiogram intelligent segmentation method and relates to the technical field of image processing. The method comprises the following steps: acquiring a plurality of first regions based on each frame of gray difference value image; acquiring a microvessel probability corresponding to each first region based on each first region; acquiring a plurality of second regions based on each microvessel probability; superimposing each second region in each frame of gray difference value image to acquire an overlapping heat map; and segmenting and labeling coronary microvessels in each frame of time sequence angiogram based on the overlapping heat map. The application can effectively amplify the weak signal of microvessels with thin diameter and low contrast agent concentration by acquiring a plurality of time sequence angiograms and calculating gray difference value images to capture the dynamic gray change of microvessels in the contrast agent perfusion process, thereby breaking through the limitation of traditional methods that rely on static single-frame images and ignore time sequence information.