352 results about "Image type" patented technology

Systems and techniques are provided for automatically analyzing and processing domain-specific image artifacts and document images. A process can include obtaining a plurality of document images comprising visual representations of structured text. An OCR-free machine learning model can be trained to automatically extract text data values from different types or classes of document image, based on using a corresponding region of interest (ROI) template corresponding to the structure of the document image type for at least initial rounds of annotations and training. The extracted information included in an inference prediction of the trained OCR-free machine learning model can be reviewed and validated or corrected correspondingly before being written to a database for use by one or more downstream analytical tasks.

The invention discloses a multi-source remote sensing image zero sample change detection method, and relates to the technical field of remote sensing image processing, and the method comprises the following steps: obtaining remote sensing images collected by two or more remote sensing sensors at different time points in the same geographic area, the image types including optical images and radar images; preprocessing each source image, unifying the spatial resolution and the registration precision, and denoising and standardizing the image; according to the method, the cross-modal shared semantic embedding space is constructed and unsupervised comparative learning is introduced, so that the semantic consistency of the multi-source remote sensing image is effectively improved, and the change recognition capability of the model under the zero sample condition is enhanced; and meanwhile, a difference fusion calculation and structure consistency constraint module is adopted, so that the boundary judgment precision of a change region and the overall structure consistency are improved, and the accuracy and stability of a detection result are remarkably improved.

The invention provides a concrete bridge damage detection and residual life prediction method based on machine vision. The method comprises the following steps: acquiring a concrete bridge surface image and a corresponding structure coordinate thereof; the structure coordinates are explicitly embedded into the concrete bridge surface image, and a coordinate surface image is obtained; the coordinate surface image is input into an image recognition model, image types and corresponding structure coordinates are recognized, the image types comprise an extrinsic damage type image and a mark type image, and the mark type image is a mark image generated by manual testing; determining each first life influence parameter according to the external display damage image and the corresponding structure coordinate; determining each second life influence parameter according to each mark type image and the corresponding structure coordinate; and estimating the remaining life of the concrete bridge according to each first life influence parameter and each second life influence parameter. By implementing the method, the comprehensiveness of concrete bridge damage detection and the accuracy of residual life prediction can be improved.

The invention discloses a multi-modal document structured processing method and device, equipment and a medium. For other different data types extracted from the to-be-processed document, the data of the other data types can be subjected to structured processing according to the preset structured processing steps corresponding to the other data types, so that the information carried by the data of the other data types is deeply mined, and the processing efficiency of the to-be-processed document is improved. And the information carried by the data of other data types in the to-be-processed document is presented in a unified and structured form through effective integration and utilization of the multi-modal data. For the image data, the character image type and the non-character image type are distinguished, and different processing modes are adopted, so that the information carried in the image data is extracted more accurately. For the table data, the table data is converted into an information set with rich semantics from a simple numerical matrix by obtaining a header name and a line name and generating a descriptive text for each data item.

The invention provides a PDF (Portable Document Format) document structured loading method based on image recognition, which relates to the technical field of image data processing, and comprises the following steps: setting a predetermined extraction scale of a document image based on margin information of a PDF document; obtaining an initial image of the PDF document with a predetermined coarse scale in predetermined extraction scales; introducing a preprocessing strategy to process the initial image to obtain a target image; a target feature parameter set of the target image is collected in a multi-dimensional mode, the loading engine classifier is activated to analyze the target feature parameter set, and a target engine category is determined; and carrying out structured loading on the initial image through the target engine category. The technical problem that in the prior art, due to the fact that a loading engine cannot be adaptively selected according to semantic features and layout forms of images in a PDF document, the image type PDF structure restoration effect is poor is solved, and the technical effect of improving complex PDF document structure reconstruction accuracy and data loading quality is achieved.

Systems and techniques are provided for automatically analyzing and processing domain-specific image artifacts and document images. A process can include obtaining a plurality of document images comprising visual representations of structured text. An OCR-free machine learning model can be trained to automatically extract text data values from different types or classes of document image, based on using a corresponding region of interest (ROI) template corresponding to the structure of the document image type for at least initial rounds of annotations and training. The extracted information included in an inference prediction of the trained OCR-free machine learning model can be reviewed and validated or corrected correspondingly before being written to a database for use by one or more downstream analytical tasks.

Described herein is a computer implemented method for generating a plurality of synthesised designs based on a selected design that includes a first element. The method includes accessing design generation data that includes source data and field association data. The source data defines a first field and a first design generation record that defines a first content item that is an image-type content item and that is associated with the first field. The field association data defines a first association between the first field and the first element. The method further includes generating a first synthesised design that is based on the selected design and that is generated such that the first element of the selected design is updated in the first synthesised design using the first content item.

The invention relates to a cross-modal heart image segmentation method and system based on space-time causal. The method comprises the following steps: acquiring a clinical heart image; the clinical heart image comprises a magnetic resonance heart image, a computed tomography heart image and an ultrasonic heart image; identifying the image type of the clinical heart image, and obtaining annotation information corresponding to the image type; the image type comprises a time sequence image, a 3D image and a 3D time sequence image; inputting the clinical heart image and the annotation information into the space-time causal diffusion network model to obtain a heart image segmentation result; the space-time causal diffusion network model is used for performing image segmentation on the clinical heart image according to a set mode and a corresponding image type to obtain a heart image segmentation result; the heart image segmentation result comprises a heart segmentation image and annotation data. By adopting the method, the dynamic change characteristics of the heart in different time and space can be captured according to different types of clinical heart images, and a heart image segmentation result is obtained.

Systems and methods are described for creating three dimensional models of building objects by creating a point cloud from a plurality of input images, defining edges of the building object's surfaces represented by the point cloud, creating simplified geometries of the building object's surfaces and constructing a building model based on the simplified geometries. Input images may include ground, orthographic, or oblique images. The resultant model may be scaled according to correlation with select image types and textured.

The embodiment of the invention provides an information retrieval method and device, electronic equipment and a storage medium, belongs to the technical field of artificial intelligence, and is suitable for financial science and technology and medical science and technology scenes. The method comprises the steps that image conversion is carried out on an original document to obtain an initial document, and the document content type of the original document comprises at least one of the following types: an image class and a table; determining a target information extraction model based on the document content type; performing content extraction on the initial document through a target information extraction model to obtain target document content features; performing information query on the target document content features based on the target query problem to obtain preliminary query content; screening from the original document based on the preliminary query content to obtain a hit document position; generating retrieval basis information based on the hit document position; and performing content optimization based on the historical query context information, the preliminary query content and the retrieval basis information to obtain target retrieval content. According to the embodiment of the invention, the accuracy of information retrieval can be improved.

The invention relates to the technical field of medical report generation, and discloses a medical image automatic annotation and report generation system which comprises a preprocessing module, a hierarchical feature extraction module, a semantic annotation module, a guide generation module, a diagnosis reasoning module and a man-machine cooperation interface. According to the method, through standardized preprocessing, hierarchical feature extraction and automatic generation of the annotation and the report, the processing efficiency is greatly improved, and the manual load is reduced; secondly, on the basis of hierarchical feature extraction and semantic and visual embedding alignment, feature support and semantic mapping precision of annotation and reporting are enhanced, and result accuracy and cross-scene consistency are improved; meanwhile, multi-modal alignment processing and multi-dimensional information fusion enhance the adaptability of the system to different image types, generate a structured report conforming to the clinical process and improve the clinical applicability; in addition, a man-machine cooperation feedback mechanism forms a parameter optimization closed loop, so that the system dynamically adapts to new scenes and specifications, and continuous improvement of generalization ability is realized.

The invention discloses a transformer substation hidden danger comprehensive monitoring method, system and equipment integrating multiple sensors and a medium, and belongs to the technical field of transformer substation hidden danger monitoring. Cleaning and fusing the collected data, and constructing a unified data set; inputting the unified data set into an identification structure constructed by combining a convolutional neural network and a recurrent neural network, respectively processing image type and time sequence type data, and completing hidden danger type identification according to a preset training set proportion and training parameters; carrying out risk value calculation and grade judgment in combination with an identification result and hidden danger related characteristics; and early warning information is generated based on a judgment result, classified pushing and display are performed, early warning processing feedback information is received, and state updating and recording and archiving are completed. Fusion sensing and recognition of multi-source monitoring data are achieved, hidden danger early warning capacity and good environmental adaptability are achieved, and the operation safety and management efficiency of the transformer substation are improved.

The invention provides a breast lump ultrasonic image classification method and system based on deep learning, relates to the technical field of ultrasonic image classification, realizes integrated processing of various ultrasonic image types through a unified feature extraction model, improves the diagnosis efficiency, and improves the diagnosis accuracy. Meanwhile, the influence of uneven probe pressure, equipment noise interference and motion artifacts on the imaging quality is quantified based on an image quality score generated based on the physical basic characteristics, and the knowledge enhancement intensity is dynamically adjusted according to the influence, so that the pathological sensitive characteristics are accurately compensated under the clamping of a medical knowledge graph; a hierarchical decision-making mechanism is established through a two-stage cascade classification network, subtype deep analysis is started for malignant lesions on the premise that rapid diagnosis is ensured, and the recognition rate is improved; and a two-factor dynamic calibration system is formed by combining model cognitive factors based on decision logic self-consistency to provide support for efficient and accurate classification of breast lump ultrasonic images.

Methods, systems, and apparatus for selecting an input mode are described. In one aspect, a method includes receiving request data specifying a request to launch a virtual assistant application from a lock screen of a mobile device. In response to receiving the request data, input signals are obtained. A selection of an input mode for the virtual assistant application is made, from candidate input modes, based on the input signals. Each candidate input mode is of an input type different from each other input type of each other candidate input mode. The input types include an image type and an audio type. The input mode of the image type receives pixel data for input to the virtual assistant application. The input mode of the audio type receives audio input for the virtual assistant application. The virtual assistant application presents content selected based on input signals received using the selected input mode.

The invention discloses an anti-halo interference fisheye image effective area detection method, which performs modeling simulation on halo interference possibly existing in a fisheye image shooting process aiming at halo intensity, fuzzy degree, color, position and size factors, and constructs a fisheye image effective area detection model based on a deep neural network. The input of the model is a halo interference fisheye image, the output is the center coordinate and size of an effective area, and the anti-halo interference fisheye image effective area detection model comprises a fisheye image type identification module, an effective area identification module and an effective area parameter extraction module. The fisheye image effective area detection method provided by the invention has stronger robustness and higher detection precision for various halo interferences, can effectively solve the problems existing in the traditional method, and further improves the distortion parameter estimation precision and the distortion correction effect of the fisheye image.

The invention relates to the technical field of medical image analysis and processing, and discloses a complex aortic lesion intelligent analysis method and system and a storage medium. The method comprises the following steps: collecting a plurality of aorta images of multiple image types containing different lesion types, collecting lesion information as label information, and labeling the lesion aorta images based on the label information; extracting a lesion area and a key area, generating first training data and training a first recognition model; acquiring to-be-analyzed aortic images of multiple image types of a patient, performing enhancement processing, using a trained first identification model to identify each key region in the to-be-analyzed aortic images, acquiring a lesion score of each lesion type, calculating a first lesion score of each lesion type of the to-be-analyzed aortic images, and performing image processing on the to-be-analyzed aortic images according to the first lesion score; and determining a corresponding lesion type based on the first lesion score. According to the method, various image types can be effectively analyzed, and the recognition accuracy of aortic lesion types is improved.

The present invention relates to the technical field of liquid crystal display drive control, and discloses a driving method and system for a high refresh rate liquid crystal display screen. The method includes: a motion amount detection module calculates the motion intensity of the dynamic area through inter-frame comparison; a voltage excitation adjustment module applies voltages differentially according to the dynamic / static areas, and the dynamic area uses asymmetric voltages to improve the response speed; a gray scale adjustment module dynamically optimizes the voltage amplitude based on the local contrast to smooth the gray scale; an adaptive refresh rate module fuses the motion intensity and the image type to switch the refresh rate gear; a charge balance management module intelligently triggers sub-region reverse voltage pulses in a static scene to eliminate afterimages. The system eliminates afterimages through the cooperation of multiple modules, dynamically manages charge balance, intelligently switches the refresh rate gear based on the motion intensity, and cooperates with voltage regulation to reduce static power consumption, adopts sub-region gray scale compensation to optimize the detail display, and combines asymmetric high-voltage drive to improve the response speed of liquid crystal molecules.

The invention discloses a droplet uniformity on-line detection and alarm method and application. The method comprises the following steps: acquiring an original image through an industrial camera, carrying out image type conversion on the original image to obtain a droplet grayscale image, setting different corrosion values according to different background pollution conditions to carry out morphological operation on a solder ball image, and carrying out online detection on the droplet uniformity according to a preset connectivity value. Eliminating a part smaller than a connected value to obtain a binary image with only the microdroplet, extracting contour coordinate information of the microdroplet by using a Canny algorithm, performing first correction on the contour coordinate information by using open operation, and then performing second coordinate information correction by using a least square ellipse fitting algorithm to obtain a binary image with the microdroplet; and finally, the particle size and the roundness of the microdroplet are calculated according to the contour coordinate information of the microdroplet, and when the software detects that the threshold values of the particle size and the roundness of the microdroplet are lower than a preset alarm threshold value, the alarm gives an alarm to remind an operator to adjust parameters of equipment for preparing the microdroplet in time, so that the final product quality and the production efficiency are improved.

The invention relates to the technical field of computers, and discloses a document format conversion method and device for large model training, and the method comprises the steps: carrying out the classification of a PDF document based on a plurality of classification indexes, and obtaining a document type; when the document type is an image type, performing image conversion and preprocessing to obtain a preprocessed page image; analyzing the preprocessed page image to obtain multi-modal content, and processing the multi-modal content to obtain corresponding processing results; performing content reconstruction and optimization on the PDF document to obtain a first intermediate document; and performing content rearrangement on the first intermediate document to obtain a Markdown document. According to the method, the PDF documents are accurately classified by integrating multiple dimensions, different types of documents are subjected to differentiated format conversion, the format conversion efficiency and accuracy are improved, the generated Markdown document conforms to the original PDF document, semantic coherence and format standardization are achieved, the large model input requirement is met, and therefore the model training effect can be improved.

The invention relates to a document content customized extraction method and system based on a large model. The method comprises the following steps: inputting a document, and dividing the document into a text document or an image document; formulating a structured cue word according to the document and user requirements; calling a large model to process the document, and performing content extraction based on the structured cue word to generate a preliminary extraction result; comparing the preliminary extraction result with the extraction target, updating the structured cue word for iterative optimization, and recalling the large model until the extraction result meets a precision threshold value; and confirming that the extraction result meeting the precision threshold meets an extraction target and storing the extraction result as a final extraction result, and storing the final extraction result as an output format. According to the method, efficient extraction of multi-modal document content information is achieved through large model calling, customized extraction is achieved through structured cue words, the output precision is improved through an interactive process, a large amount of time cost is saved for a user, and remarkable achievement benefits are brought.

The invention provides a rumor detection method and system based on multi-modal evidence fusion, and relates to the technical field of data processing, and the method comprises the steps: obtaining N initial evidences and M associated declarations based on a to-be-detected declaration; calculating the quality score of the initial evidence according to the information density of the text type evidence and the image type evidence contained in each piece of initial evidence; according to the similarity between each piece of initial evidence and the to-be-tested statement and the correlation statement, calculating to obtain a similarity score, and according to the similarity score and the quality score, calculating to obtain a weight corresponding to each piece of initial evidence; calculating according to the initial evidence and the corresponding weight to obtain a target evidence; and obtaining a detection result according to the target evidence and the to-be-detected statement. According to the method, the evidence information density serves as an index for evaluating the evidence quality, the evidence weight is adjusted by combining the evidence quality, the similarity between the evidence and the to-be-detected declaration and the similarity between the evidence and the associated declaration, it is guaranteed that the high-quality evidence is obtained, and rumor detection is conducted according to the high-quality evidence.

For improving the efficiency of malware detection, a method is proposed that can handle computer files (1) of varying types and sizes and at the same time maintain a high detection performance by classifying a number of different types (A, B, C, D, E) of images (4), each calculated or derived from a particular computer file under test (1) (CFUT), using artificial intelligence methods such as machine learning, in particular deep learning, for example as provided by neuronal networks (24) or supervised deep learning algorithms. The different image types (A, B, C, D, E) are generated using different image conversion techniques and a number of approaches are presented for computing images (4) of uniform size Si that contain relevant information for classifying the CFUT (1).

The invention discloses a complex document recognition method based on layout analysis and OCR (optical character recognition) and a medium. Obtaining a to-be-recognized document image in real time, and performing preprocessing operation through a preprocessing module to obtain a to-be-recognized standard document image; processing the to-be-identified standard document image through a layout analysis module, and obtaining at least one target area image, a target area image type and an area vision matrix feature in combination with a layout analysis strategy method; performing text analysis on each target area image through a multi-mode OCR engine module according to the type of each target area image to obtain each text semantic analysis result and an area semantic matrix feature; obtaining a current visual text fusion feature through a multi-modal attention fusion module; and generating and feeding back a complex document identification output result according to the current visual text fusion feature. The problem that the accuracy is poor due to the fact that irregular layouts cannot be processed is solved, and the accuracy and flexibility of complex layout recognition are improved.

The invention discloses a deep transfer learning-based medical image classification model training method and system, and relates to the technical field of medical classification. Comprising the following steps: acquiring an intestinal medical image with known feature information and environmental parameters thereof, and performing image preprocessing to generate a training data set; the feature information comprises texture entropy, edge contour, color contrast, shape area and circularity. Normalization and sliding window filtering are performed on the environmental parameters to construct humidity, illumination and vibration influence functions. Thirdly, a first classification model and a second classification model are established, pre-training is carried out through deep transfer learning, and the learning rate is dynamically adjusted based on the environmental influence parameters; and finally, inputting the to-be-identified intestinal medical image and the environmental parameters thereof into the training model to obtain the corresponding image type and focus type, thereby improving the automation level of medical image analysis, and helping to assist doctors in more scientific and accurate diagnosis.

The invention provides a multi-modal large model-based screen content detection method and system and a storage medium, and the method comprises the steps: carrying out the image preprocessing of a screen content image, and obtaining a to-be-detected image and a to-be-detected text; inputting a to-be-detected image into the pre-trained visual detection large model for visual risk detection to obtain a visual risk value, and inputting a to-be-detected text into the pre-trained text detection large model for text risk detection to obtain a text risk value; performing weighted operation on the visual risk value and the text risk value according to the image type of the screen content image to obtain a target risk value; and if the target risk value is greater than the first risk threshold, controlling the target screen to start the virtual private network, and controlling the target screen to display preset content according to the virtual private network. According to the embodiment of the invention, a multi-modal content detection mode combining visual risk detection and text risk detection is adopted, so that the screen content detection accuracy is improved.

The invention relates to the technical field of graph generation, in particular to a self-adaptive topological plant station wiring diagram generation method, which comprises the following steps of: acquiring a wiring diagram image, extracting edge features, identifying an image type, positioning a line structure boundary, identifying text elements, classifying corresponding graph areas, and constructing a connection path to generate a topological wiring diagram. According to the method, clear recognition of the structure contour is realized through extraction of the edge features of the line structure of the image, the spatial position information of the graphic area is acquired in combination with a multi-scale primitive positioning technology, and then identity extraction and classification judgment are performed on the text elements contained in the image in a character recognition mode. And establishing a corresponding relationship between the equipment identifier and the graph region, constructing a complete path by combining the graph arrangement direction and the starting and ending coordinates of the connecting line, and determining a connection sequence and a relative position relationship between the equipment. Coordination and unification of spatial arrangement are realized through application of an attraction and repulsive force model between nodes, and clearness and coherence of a wiring diagram in the aspect of topological structure expression are improved.

The invention discloses a literature diffraction pattern quality evaluation and intelligent repair system and method, and belongs to the technical field of computers. The method comprises the following steps of: 1, inputting and processing data; step 2, image-text consistency alignment and consistency inspection are carried out; step 3, calculating quality parameters, constructing a quality evaluation system, and performing grading and report output on atlas data; step 4, repairing and warehousing the atlas; the system adopts a modular architecture design, is composed of an image processing module, a text analysis module, a signal analysis module, a quality evaluation module, a report generation and output module, a grading feedback module, an atlas restoration module and the like, and can flexibly adapt to different types of diffraction atlas quality evaluation tasks. According to the technical scheme, aiming at the limitation of literature image type XRD atlas quality evaluation, the evaluation system and method with intelligent repair, knowledge reasoning and self-learning capabilities are constructed, and the availability, credibility and structured processing efficiency of literature XRD atlas data are remarkably improved.

To provide a gaming machine that enhances the enjoyment of the game by utilizing presentations that suggest reliability based on the player's choices. [Solution] In the pre-performance, one of several candidate images is selected as the pre-selected image, and in the post-performance, one of several candidate images is selected as the post-selected image. When the combined performance is executed in state A, the reliability is higher when the pre-selected image and post-selected image are of the same type than when they are of different types. When the performance is executed in state B, which is different from state A, the reliability is higher when the pre-selected image and post-selected image are of different types than when they are of the same type.

The invention discloses an ultrasonic image automatic quantitative evaluation method and system based on artificial intelligence, and relates to the technical field of medical image processing. And respectively acquiring a modal category of the ultrasonic image, a detection frame of each target structure, a structure type and / or a sampling line position and a detection frame of each cardiac cycle through the depth image classification model and the target detection model. And the section type of the ultrasonic image is accurately judged according to the data. According to the method, in combination with the modal category and section type of the ultrasonic image and the image type contained in the ultrasonic image, an appropriate quantitative analysis mode can be accurately provided for different ultrasonic sections, and a quantitative analysis result is visualized, so that a doctor can observe the result conveniently. According to the invention, comprehensive analysis of the multi-modal section can be realized, and more comprehensive and deep analysis and evaluation of the heart structure or function can be realized.

System and methods for extracting data from and authenticating digital images of documents. The system may receive first digital image data. The system may determine a first image type of the first digital image data. The system may determine a first version identifier based on the first digital image data. The system may determine a first ruleset. The system may determine a first set of pixel coordinate ranges for authenticating digital images. The system may determine a first authentication status for the first digital image. The system may determine a second portion of the first digital image at a second set of pixel coordinate ranges. The system may compare a first portion to the second portion. The system may determine a second ruleset. The system may update the image ruleset database to include the second ruleset.