121 results about "Key images" patented technology

The invention discloses a valve state monitoring and safety interlocking method based on machine vision, and belongs to the technical field of industrial process control and safety. The method comprises the following steps: 1, calibrating a camera and establishing a model; step 2, image acquisition and preprocessing; step 3, key image feature extraction and state decision; fourthly, opening degree calculation and state judgment are carried out; 5, safety interlocking is triggered; and 6, feeding back data. The valve state monitoring and safety interlocking method is achieved through the valve state monitoring and safety interlocking system. According to the invention, the millimeter-level or even submillimeter-level displacement and angle real-time accurate measurement of the opening degree of the valve can be realized, the state monitoring of the valve is realized, and the stability and reliability in the production process are improved; state judgment is carried out by comparing the real-time opening value with the instruction opening from the control system, a safety interlocking signal is generated and transmitted to the safety interlocking system, and finally physical safety action is triggered, so that a set of complete safety protection system is realized, and accidents are avoided.

A medical image processing apparatus includes: an image acquisition unit that acquires a plurality of medical images; an image classification unit that classifies the medical images into at least one of a plurality of categories; an image display unit that displays, on a screen, at least one of the plurality of medical images based on a result of the classification as an automatically selected image; an input receiving unit that receives an input from a user to select an image requiring reselection as a key image from among the displayed automatically selected images as a selected image; and a display control unit that switches and displays, as a switching image, a non-automatically selected image other than the automatically selected image among the plurality of medical images classified into each category, instead of displaying the selected image.

Disclosed herein are methods and systems for using artificial intelligence modeling techniques to generate a path for an ego. In an embodiment, a method comprises retrieving image data of a space around an ego, the image data captured by a camera of the ego; predicting by executing an artificial intelligence model, an occupancy attribute of a plurality of voxels corresponding to the space around the ego; generating a 3D model corresponding to the space around the ego and each voxel's occupancy attribute; upon receiving a destination, localizing, by the processor, the ego by identifying a current location of the ego using a key image feature within the image data corresponding to the 3D model without receiving a location of the ego from a location tracking sensor; and generating a path for the ego to travel from the current location to the destination.

The invention discloses a nasopharyngeal carcinoma necrosis prediction method and device based on feature fusion and a medium, and the method comprises the steps: obtaining multi-modal medical data of a target patient at a plurality of preset time points, and extracting medical features in different modals for each time point, so as to form a medical time sequence feature matrix in different modals; inputting the medical time sequence feature matrixes in the different modes into a gated twin-tower Transform network, and performing feature fusion on the medical time sequence feature matrixes in the different modes through a preset inter-channel-time step cross attention mechanism to obtain key image fusion features and nasopharyngeal carcinoma necrosis risk prediction results; and if the nasopharyngeal carcinoma necrosis risk prediction result is greater than a preset risk threshold, inputting the key image fusion features and the original three-dimensional image data into a 3D nnU-Net network to generate a nasopharyngeal carcinoma necrosis prediction map of the target patient. The prediction accuracy of nasopharyngeal carcinoma necrosis can be improved.

The invention relates to the technical field of satellite communication, and discloses an efficient encoding and decoding and distribution transmission method for satellite image data, which comprises the following steps of: acquiring satellite epoch parameters, estimating a downlink channel capacity sequence, decomposing the image data into a basic feature code stream and a background residual code stream, and distributing a transmission priority, and mapping the basic feature code stream to the most significant bit of the modulation symbol, mapping the background residual code stream to the least significant bit of the modulation symbol, and implementing nonlinear displacement stretching on the coordinate of the constellation point where the modulation symbol is located according to the carrier frequency shift change rate. A core semantic bit phase tolerance boundary is expanded by compressing a background data judgment space, high-fidelity transmission of key image features in a severe phase distortion environment is ensured, and the information transparent transmission capacity in a high-dynamic link is improved.

The application discloses a financial document multi-agent collaborative analysis method and device, a medium and an electronic equipment, and relates to the technical field of financial technology. The method comprises the following steps: preprocessing a financial document to be analyzed to obtain preprocessed image data; detecting the preprocessed image data by using a pre-trained target detection model to obtain a key image region; and performing multi-round collaborative analysis on the key image region by using multiple image recognition agents and conflict decision agents to obtain a collaborative analysis result. The method can improve the accuracy of financial document data analysis and obtain logically consistent and highly reliable structured data.

The invention discloses a three-dimensional cultural relic 3D scanning ultra-high-definition enhancement method, device and equipment and a medium, and relates to the technical field of three-dimensional modeling, and the method comprises the steps: obtaining a light supplement enhancement image set, carrying out the dense matching strategy of the light supplement enhancement image set, and generating image pairs with consistent structures; minimum coverage rectangular region extraction is carried out on the image pairs with the consistent structures, and redundant complementation region pairs are generated; performing texture completion processing on the receptor region image in the redundant completion region pair to obtain a texture completion image set; and key image pairs are extracted based on the final texture completion image set, a matching point pair set is obtained through scale invariant feature transformation, and a matching point three-dimensional coordinate set is obtained through triangulation of each matching point pair. According to the invention, pixel-by-pixel brightness compensation and color equalization are carried out by combining the time-weighted exposure consistency index, so that illumination equalization adjustment of the region with uneven local exposure is realized.

The application discloses a lightning target recognition model and method facing image input modal imbalance and high false alarm rate, and belongs to the technical field of image recognition and meteorological disaster monitoring. Firstly, Gaussian mixture density estimation and time-weighted fusion are performed on ground-based lightning location data to convert the image label with spatial probability distribution characteristics; then, Himawari-8 satellite multi-channel brightness temperature images and their derived feature maps and radar echo images are used as multi-source heterogeneous inputs. In view of the information imbalance between different observation modalities, an MDE-UNet deep learning model is constructed, a multi-scale feature fusion module, a MLP enhanced decoding unit based on a weighted sliding window and a radar image information enhancement module are innovatively designed, the key image features are enhanced and irrelevant noise is suppressed, and the feature imbalance problem in heterogeneous data fusion is solved. The model adopts an asymmetric weighted BCE-DICE loss function to strengthen the attention to lightning target pixels.

The invention relates to controller power-on test equipment, a test method and a manufacturing system. The controller power-on test equipment comprises one or more power-on test parts, a multi-axis displacement unit, an image acquisition part and an image test part, the power-on test part is used for placing, positioning and fixing a target test controller and powering on the target test controller through the test interface; the multi-axis displacement unit is used for executing multi-degree-of-freedom motion according to a preset or dynamically adjusted acquisition track; and the image acquisition part is used for acquiring test image data displayed by one or more image test parts according to preset shooting parameters in the motion process of the multi-axis displacement unit and transmitting the test image data to a target terminal in real time. According to the method and the device, the acquisition track can be dynamically adjusted according to the test image data, and the test stage of the target test controller is identified based on the continuous image feature change, so that key image acquisition and test judgment are completed in a proper stage, automation and adaptive detection of the power-on test process are realized, and the power-on test effect is improved.

The application discloses a brain function state analysis method and device, which comprises the following steps: collecting a brain image sequence combination of a target object, preprocessing, then performing feature extraction on the brain image to obtain image features, and selecting key image features; performing quantitative analysis on the key image features to generate a feature parameter set; inputting the feature parameter set into a brain function state prediction model to determine the brain function state of the target object; the brain function state prediction model is obtained by integrating and then training a semantic retrieval model and a fine-tuning generation model, the semantic retrieval model is trained based on a brain knowledge document library, and the fine-tuning generation model is trained based on reinforced learning parameters, brain information segments retrieved by the semantic retrieval model, and a labeled data set of a mapping relationship; statistical characteristic values corresponding to each brain function state are determined, and brain function physiological data of the target object is analyzed. The application can quantitatively and standardize overall analysis of the brain function state.

The invention relates to the technical field of image processing, and discloses a target detection method and device, equipment and a medium. The method comprises the following steps: acquiring a to-be-detected panoramic image; based on the key image features in the to-be-detected panoramic image, performing adaptive slicing on the to-be-detected panoramic image to obtain a plurality of original slice images; performing neighborhood information supplement on the original slice image to obtain a plurality of supplement slice images; performing target detection on the supplementary slice image to generate a corresponding target detection frame; and performing global consistency verification on the target detection frames to remove the target detection frames which do not meet a preset consistency condition, and obtaining a target detection result corresponding to the to-be-detected panoramic image. The embodiment of the invention can be suitable for various panoramic image slice scenes, and the panoramic image slice quality is improved.

The embodiment of the invention discloses a certificate anti-counterfeiting detection method, which comprises the following steps of: identifying a certificate type of a certificate corresponding to a certificate image to be detected, marking a key image area in the certificate image to be detected by adopting a mark mask matrix preset for the certificate type as a mark image; and inputting the marked image and detection prompt information preset for the certificate type into the VLM, so that the VML detects whether the certificate image to be detected is a forged certificate image or not according to the marked image under the guidance of the detection prompt information. According to the method, the detection prompt information comprises the first text information used for guiding the VLM to improve the attention weight of the marked key image area in the marked image, so that the attention of the VLM can be guided to the high-risk key image area, and the VLM is prevented from wasting computing power in other unimportant image areas; therefore, the counterfeit certificate image can be detected more accurately, and the security of online business based on eKYC is effectively improved.

The invention discloses a second-hand vehicle copywriting generation method, and the method comprises the steps: obtaining vehicle structured information, feature label information and a vehicle source image set after obtaining a vehicle source identifier; determining key images of description parts such as appearances, interiors and power from the image set; aiming at each description part, constructing part prompt information which contains task constraint, content constraint and length constraint and is injected with structured information, calling a visual understanding generation service based on the key image and the part prompt information to generate a part copywriting, and cleaning the part copywriting; and fusing the structured information, the characteristic label information and the part copywriting to generate a summary copywriting, and outputting a structured copywriting object comprising a summary field and a part field. Through the mode, the source of the copywriting content is limited, part-by-part constrained generation and structured output are realized, the image-text consistency and controllability are improved, assume and quality fluctuation are reduced, and the method is suitable for batch generation.

The embodiment of the invention provides an image recognition method and device, equipment, a medium and a product, and relates to the field of image processing, and the method comprises the steps: firstly segmenting an image into blocks, coding the blocks into embedded vectors, and calculating the contribution degree of each image block to recognition; and then only selecting the key image blocks with the highest contribution degree in the first preset number for weighting and inputting the key image blocks into the recognition model. According to the technical scheme, the model does not need to carry out indiscriminate dense calculation on all regions of the whole image, but can dynamically focus main calculation resources on the key region with rich information according to the image content, so that the calculation load and memory occupation during high-resolution image processing are reduced; by introducing a dynamic selection mechanism based on the importance scores of the image blocks, the problem of low image recognition efficiency in the prior art is solved, and the recognition efficiency is improved while the recognition precision is ensured.

The application discloses a natural water fish monitoring method and system based on image recognition, and relates to the technical field of image recognition, and comprises the following steps: determining a target monitoring fish, and configuring an attracting strategy; executing the attracting strategy, synchronously activating a sonar to execute regional scanning of a target area to establish a time sequence scanning data set, and starting an image acquisition device to perform positioning image acquisition, to establish a time sequence acquisition data set; simultaneously, updating sonar acquisition parameters to synchronously execute sonar update data acquisition, to establish an update data set; after extracting key image frames from the time sequence acquisition data set, performing image processing and recognition on the key frame images, to establish a first recognition result; performing positioning recognition on a time sequence window, to establish a second recognition result; after performing alignment and fusion of the first recognition result and the second recognition result, outputting a fusion recognition result. The application solves the technical problems that fish monitoring in the prior art is susceptible to environmental interference and has low recognition accuracy, and achieves the technical effect of improving fish monitoring accuracy and stability.

The invention discloses a navigation mark visual identification and collision early warning method and system based on deep learning, and relates to the technical field of intelligent shipping, and the method comprises the steps: collecting a navigation channel monitoring video stream in real time, and carrying out the preprocessing of a key image frame; constructing a navigation mark detection model based on an improved Officient Det network, and outputting the position and category of the navigation mark in the key image frame; the method comprises the following steps: constructing a ship track prediction model through ship historical track data, predicting a short-time track of a ship, calculating the relative position, speed and course angle of the ship and a navigation mark, and constructing a dynamic collision risk field; and fusing the visual detection result, the radar ranging data and the AIS information to obtain a ship collision risk, and triggering graded early warning based on a preset risk threshold. According to the method, the improved deep learning model and the real-time calculation framework are combined, the robustness of navigation mark identification is improved, an efficient collision early warning mechanism is established, and technical support is provided for intelligent shipping.

The application relates to a computer vision-based point cloud and finite element model generation system and method, the system comprising a video acquisition and processing module, a three-dimensional point cloud establishment module and a finite element model generation module, and the method comprising the following steps: acquiring video data of a target structure and processing the same to obtain key pictures and corresponding space-time information; based on the perspective difference between multiple key pictures and the corresponding space-time information, adopting a multi-perspective geometric reconstruction technology based on deep learning to perform three-dimensional point cloud reconstruction on the target object, and preprocessing the point cloud data; extracting effective structure space geometric information from the preprocessed point cloud data, and then adopting a tomographic scanning node mode to cut the point cloud into multiple thin layers, and layer-by-layer establishment is performed to obtain finite element components, and the finite element components are integrated to obtain a finite element model. Compared with the prior art, the application can efficiently and accurately convert video data into a finite element model required for structure analysis, and avoids the tedious operation of manual modeling.

The application provides a foreign matter detection method and system for safe operation of a railway track, relates to the technical field of image data processing, and periodically collects track image sequences and track point cloud sequences through a multi-source monitoring array; extracts key images from the track image sequences, generates track key image sequences, and maps to obtain track key point cloud sequences; uses the track key image sequences to perform point cloud fitting on the track key point cloud sequences, constructs a foreign matter simulation model distribution sequence, and performs foreign matter risk assessment based on the foreign matter simulation model distribution sequence to output a track foreign matter detection result. The application solves the technical problems in the prior art that due to single detection means, there are detection blind spots and poor environmental adaptability, resulting in insufficient detection precision and the inability to effectively evaluate the influence of foreign matter risks, achieves the technical effect of improving the accuracy of foreign matter detection, and further enhances the safety and reliability of track operation.

This invention relates to the field of cable inspection and maintenance technology, and discloses a robot for the operation and maintenance of power transmission lines in mountainous areas. The robot includes a main body, a sliding buffer assembly, a camera inspection assembly, a positioning assembly, and a drive assembly. The sliding buffer assembly decouples the camera mechanism from the robot body through a sliding groove and a slider, allowing the robot to move forward continuously at a constant speed while the camera mechanism remains stationary, eliminating motion blur and channel misalignment. Driven by the drive assembly, the positioning assembly first stretches the cable section to be tested axially and then pushes it axially, forcibly opening circumferential and axial micro-cracks to facilitate the visualization of early damage. Four sets of multispectral high-definition cameras in a ring array employ an intermittent static shooting strategy, requiring only one key image to be acquired for each detection section during the stretching and pushing states. This invention achieves high-precision, low-miss-rate detection of micro-cracks in cables in mountainous areas, simplifies the system structure, and improves the robot's portability and operational stability.

The invention relates to the field of defect identification, in particular to a spray gun coating defect identification system and method based on image identification. The method comprises the following steps: acquiring a surface initial image, analyzing the surface initial image, and constructing a coating substrate model; obtaining a surface coating image and coating time, and obtaining a coating blurred image according to the coating substrate model and the surface coating image; according to the surface initial bitmap and the surface coating bitmap, obtaining a coating frame image, and constructing a coating spraying event chain; analyzing the coating frame image through the coating spraying event chain to obtain a co-frequency spraying relation and a frequency hopping spraying relation, and constructing a key frame image and a key spraying event chain; analyzing the key frame image through the key spraying event chain to obtain a key image area and a spraying image area; and analyzing the key frame image through the key image area and the spraying image area to obtain a key defect event and an additional defect event. The defect identification precision can be improved.

The invention relates to the technical field of video generation, particularly provides a dynamic cartoon generation method, device and equipment based on a large model, and aims to solve the problem of low dynamic cartoon generation efficiency. The method comprises the following steps: acquiring an image material, an image description, a script and a scene description; according to the image material and the image description, training a preset image diffusion model to obtain a personalized image diffusion model; according to the script and the scene description, cue words are obtained; obtaining a key image according to the cue word and the personalized image diffusion model; obtaining a video clip according to the prompt word and the key image; and editing the video clip to obtain the dynamic cartoon. And the generation efficiency of the dynamic cartoon is improved.

This application relates to a cross-modal retrieval method and system based on image patches and dual-granularity alignment. The method includes: acquiring image data and text data; performing mapping and global pooling followed by coarse-grained global semantic alignment, calculating self-supervised semantic reconstruction loss, and selecting a set of key image patches while generating a set of redundant image patches; calculating a bidirectional alignment matching score; obtaining a fine-grained semantic alignment loss based on the bidirectional alignment matching score; constructing a retrieval model, iteratively optimizing the retrieval model until convergence is achieved, and outputting the final cross-modal retrieval result. This invention combines coarse-grained global contrastive learning with fine-grained interaction. Global alignment ensures overall semantic consistency, while fine-grained alignment captures rich local details. The two complement each other through joint optimization, effectively solving the problem of single-granularity alignment easily getting trapped in local optima or semantic bias, and significantly improving the accuracy of cross-modal retrieval.

The present application relates to the field of medical imaging technology, disclose a biliary tract cancer CT image prediction method and system based on momentum attention and large model verification, the method comprises: based on the pre-training visual-linguistic large model extracts the visual features of the patient's upper abdominal non-enhanced CT image, constructs a high-dimensional visual embedding vector, initializes the historical momentum attention map and the initial reasoning text sequence, and the first momentum attention map is obtained by guiding and updating through the momentum mechanism; again extract the key image patch set and the initial reasoning text sequence, and the first reasoning text is obtained by interlacing fusion of the image and the text; repeatedly execute attention update and image-text fusion operation until the reasoning thought chain and the preliminary diagnosis conclusion are generated; the correlation degree of reasoning and diagnosis is evaluated by a pure text logical verifier to obtain a logical consistency score, and the target intelligent auxiliary diagnosis report is generated by assembling the score, the present application can improve the efficiency of biliary tract cancer CT image prediction.

This invention relates to a container key recognition and pairing system and method. The system includes a recognition device fixing system, a light source correction system, a key image preprocessing system, a key image processing system, and a key recognition and pairing system. The recognition device fixing system is used to fix the positions of the light source and camera, and determine key parameters including the light source's deflection angle, the camera's height, and the tilt angle of the reference image acquisition. The light source correction system is used to acquire reference images from different lighting directions and correct the light source, calibrating its position. The key image preprocessing system is used to acquire key images from different lighting directions and select images within the region of interest for preprocessing. The key image processing system is used to calculate the surface normal vector, reflectivity, curvature, and surface height of the key in the key image. The key recognition and pairing system is used to extract the surface features of the key based on the calculation results and determine the pairing result of multiple keys. This invention has high recognition efficiency and strong applicability.

The present application belongs to the field of full-process automation and intellectualization of casting production, and provides a casting internal defect detection method based on deep learning, which comprises: image shooting for detection, HDR image enhancement processing, one-key image enhancement processing and mask image fusion, industrial X-ray AI network detection, post-processing operation and data statistics. The present application avoids the defects of relying on personnel experience and poor precision in conventional defect detection by constructing an industrial X-ray AI network. Through HDR image enhancement processing, one-key image enhancement processing and mask image fusion, the influence of the complex structure of the detection device, the defects of the shooting equipment itself and the noise points is reduced.

The invention discloses a method and system for transmitting a monitoring video of a chemical device, and the method comprises the steps: obtaining a YUV value of each frame of image in a to-be-transmitted video, generating a plurality of first matrixes which respectively correspond to a component Y, a component U and a component V and represent a YUV color space for each frame of image, converting each first matrix into a projection matrix, and transmitting the projection matrix to the video to be transmitted; obtaining a dimension reduction matrix for extracting key features of the image; each dimension reduction matrix is decomposed into a plurality of block matrixes with the same dimension, each block matrix is quantized, then for each dimension reduction matrix, first-time serialization sequencing is carried out on the block matrixes after single quantization, second-time serialization sequencing is carried out on all the quantized block matrixes after the first-time serialization sequencing is carried out, and each dimension reduction matrix is obtained; obtaining a YUV color space array of each frame of image; and performing coding compression on each frame of image by using the YUV color space array, thereby finishing video transmission by transmitting the coded and compressed image. According to the invention, efficient and accurate transmission of the monitoring video is realized.

The invention relates to an online store commodity query method and device, equipment and a medium, and the method comprises the steps: calling a graph-to-text model, and generating a commodity attribute word set which reflects related contents of commodities in a multi-dimensional image based on a key picture of each commodity in an online store; based on the commodity attribute word set of each commodity, the merchant description word set of the commodity and / or the promotion effect word set of the commodity, co-occurrence word pairs are extracted, and the co-occurrence word pairs and the corresponding commodities are stored in an associated mode; establishing a co-occurrence index library of corresponding commodities of the online store for the commodity attribute words appearing in the co-occurrence word pairs, and updating the co-occurrence word pairs to an anti-missing-detection word library; and responding to a missing detection processing event, calling a co-occurrence word having a co-occurrence relationship with a query word carried by the event from the missing detection prevention word library, and recalling a commodity list from the co-occurrence index library according to the co-occurrence word. The method and the device can prevent missing detection of commodities of online shops, and improve commodity query conversion rate.

The invention discloses a pixel structure, a display panel and a display device, and relates to the technical field of display, the pixel structure comprises a red sub-pixel, a green sub-pixel and a blue sub-pixel; the red sub-pixel is provided with a first main pixel region and a first sub-pixel region; each green sub-pixel is provided with a second main pixel region and a second sub-pixel region; the ratio of the area of the second main pixel region to the area of the second sub-pixel region is a second area ratio; the blue sub-pixel is provided with a third main pixel region and a third sub-pixel region; the ratio of the area of the third main pixel region to the area of the third sub-pixel region is a third area ratio, and the third area ratio is smaller than the second area ratio. According to the scheme, the area ratio of the blue sub-pixel with the minimum contribution to the brightness is set to be smaller than the area ratio of the green sub-pixel with the maximum contribution to the brightness, so that the visual angle characteristic of the sub-pixels under the low gray scale can be optimized, and the suppression of the color cast of the key picture is realized; and meanwhile, the aperture ratio loss caused by the area enlargement of the sub-pixel region can be controlled to the minimum limit.

The invention discloses an elevator operation management method. The method comprises the steps that 1, a relation table of corresponding relations among operation items, operation action processes, operation key images and first relative position relations is established in advance; 2, determining an operation action process, an operation key image and a first relative position relation of the operation item according to the relation table; 3, determining a second relative position relation and a third relative position relation; step 4, determining a first transition time for changing the relative position relation between the operation part and the operation object from the second relative position relation to the first relative position relation and a second transition time for changing the relative position relation between the operation part and the operation object from the first relative position relation to the third relative position relation; 5, determining a first moment according to the operation process and the operation key image; 6, determining a second moment and a third moment according to the first moment, the first transition time and the second transition time; and step 7, outputting the first switching instruction when the current moment is the second moment, and outputting the second switching instruction when the current moment is the third moment.

The present application provides a medical image processing device, a medical image processing device operation method, and a medical image processing device program, which can simply and appropriately select a key image. The medical image processing device includes a processor that performs the following processing: acquires a plurality of medical images including an object image, assigns first image information to the medical images by analyzing the medical images, performs designation of the first image information, performs control to display a medical image to which the designated first image information is assigned among the plurality of medical images, i.e., a selection image, in a first display area of a display screen in a selection image display manner, and performs control to display the designated first image information in a second display area.