113 results about "Candidate image" patented technology

This application discloses an image detection method, apparatus, computer device, storage medium, and program product, relating to the field of artificial intelligence. The method includes: extracting visual features from an input image; obtaining category text features corresponding to at least two candidate image categories, where the category text features are features used to characterize the candidate image categories from a textual perspective; obtaining category visual features corresponding to at least two candidate image categories, where the category visual features are features used to characterize the candidate image categories from an image perspective; and determining the image category to which the input image belongs from the at least two candidate image categories based on the visual features, the category text features corresponding to the at least two candidate image categories, and the category visual features. Using the method of this application can improve the accuracy of determining the image category to which the input image belongs.

This invention relates to the field of medicinal material identification technology, specifically to a method and system for identifying the authenticity of medicinal materials based on AI vision. The method includes the following steps: iteratively generating candidate images and comparing them with images of genuine medicinal material slices, adjusting the generation logic based on the comparison results. This invention constructs a latent vector mapping relationship consistent with the visual features of images of genuine medicinal material slices, combines image reconstruction errors for anomaly extraction, and enhances the perception of adulteration locations. The pixel-by-pixel reconstruction error method can locate areas with slight deviations in shape or color, and based on this, extracts the thermal decay curve of the isotopic infrared image sequence. Through time constant fitting calculation, it completes a quantitative modeling of the thermal conductivity characteristics in space, and further calculates the spatial gradient of the thermal decay time constant, realizing the quantification of the boundary strength of the material's thermal diffusion behavior, forming a thermal anomaly diffusion gradient map, providing physical property basis for subsequent multi-spectral fusion.

This invention provides a visual positioning method and system, belonging to the field of visual positioning technology. The method includes: extracting query feature vectors from a query image to be located; obtaining candidate images and their corresponding candidate geographic coordinates based on the similarity between the query feature vectors and reference feature vectors in a reference database; calculating the spatial geographic distance between the candidate geographic coordinates; performing density spatial clustering on the candidate geographic coordinates based on the spatial geographic distance to obtain spatial clusters; determining the target spatial cluster containing the most candidate geographic coordinates; and calculating the geometric center of the candidate geographic coordinates within the target spatial cluster, using the center coordinates as the positioning coordinates of the query image. This invention introduces a physical scale to cluster and vote on the retrieval results, removing mismatch noise caused by homogeneous appearances in different locations, overcoming the jump error caused by relying solely on single visual similarity, and achieving high-precision UAV visual positioning with spatial consistency.

The present disclosure relates to a recognition processing method and device and a non-transient computer readable storage medium. The recognition processing method comprises: obtaining first thermal distribution data of an input image, and, according to the first thermal distribution data, calculating first color feature data of the input image; obtaining second thermal distribution data of each candidate image among multiple candidate images, and, according to each item of second thermal distribution data, respectively calculating second color feature data of a corresponding candidate image; calculating the color similarity distance between the first color feature data and each item of second color feature data respectively, and, according to each color similarity distance, determining the candidate image, among the multiple candidate images, that matches the color of the input image as an output image.

An edge device performs automated visual inspection on a production line. A sequence of item images is received, and patch-level feature embeddings are extracted for each image. During live operation, a memory bank modeling normal behavior is built by computing distances between patch embeddings and existing entries and appending embeddings whose distance exceeds a dynamic add-threshold derived from a running mean and standard deviation of prior distances. For a candidate image, nearest-neighbor distances from its patches to the memory bank are computed and aggregated into an image-level anomaly score. A prediction threshold is calibrated by synthesizing defects from normal images, scoring both perturbation-augmented and normal images to obtain score distributions, and selecting a threshold that discriminates between them. The candidate image is classified as defective or normal by comparing its anomaly score to the calibrated threshold. All computation executes on the edge device without reliance on cloud resources during operation.

The application discloses a picture screening processing method and system based on big data collection and relates to the technical field of image data processing.The picture screening processing method and system based on big data collection comprise the following steps: S1, acquiring picture behavior data and performing time axis alignment, field unification and normalization processing on the picture behavior data; S2, constructing a picture fingerprint, generating a candidate picture set and performing credibility evaluation on same-picture merging; S3, screening a candidate repeated operation sequence and calculating a fixation revisit index of a picture track; and S4, extracting picture behavior characteristic quantities and evaluating the disposal level of a current picture track.The application effectively improves the screening processing capability, behavior track tracing capability and key picture identification capability of picture data, and solves the problems of difficult same-picture merging, difficult repeated operation identification and lack of structured basis for picture screening.

To improve the accuracy of security identification using a simple method. [Solution] The securities processing device 10 includes an identification unit 154 that determines whether or not to accept a gift certificate GT by identifying the gift certificate GT based on an image IIMG of the gift certificate GT. The identification unit 154 performs an identification process that includes a gift certificate type identification process that identifies the gift certificate type of the gift certificate GT based on a plurality of pre-registered reference images RIMG and an image IIMG of the gift certificate GT, and identifies the gift certificate GT based on the result of the identification process. The gift certificate type identification process includes a first process that uses AI to calculate the similarity between the image IIMG and the reference image RIMG to search for candidate images from a plurality of reference images RIMG that have a similarity to the image IIMG of the gift certificate GT that is equal to or greater than a standard value, and a second process that identifies the gift certificate type of the gift certificate GT based on the degree of agreement of the candidate image with the image IIMG of one or more predetermined feature parts FT.

This invention provides a furniture product recognition method and system based on multimodal heterogeneous data processing, relating to the field of image recognition technology. The method includes: acquiring a target product image; preprocessing the target product image to obtain global text features, target text features, and visual features; processing the global text features, target text features, and visual features using a modal fusion model to obtain a classification result for the target product image; searching for a candidate image set in a pre-constructed database based on the classification result of the target product image; inputting the target product image and the candidate image set into a knowledge graph model, and using the knowledge graph model to perform entity alignment between the target product image and the candidate images; and outputting the recognition result for the product to be identified in the target product image based on the entity alignment result. This application can achieve accurate identification and rapid matching of furniture products.

The application provides a self-evolution presentation document generation method and system based on multi-modal and knowledge graph, wherein the self-evolution presentation document generation method comprises the following steps: constructing an enterprise-level picture meta-database containing a vector index; parsing a source document to obtain structured content and generating a presentation document outline; searching for matching candidate pictures in the enterprise-level picture meta-database based on the target page core content in the outline; querying a design decision knowledge graph to generate layout and color matching planning; assembling a presentation document according to the planning and synchronously generating design decision metadata; driving knowledge graph updating by displaying the design decision metadata and obtaining user modification feedback on the design; and the self-evolution presentation document generation system comprises functional modules for realizing the above method. The application solves the technical problems of poor picture quality, inaccurate image-text matching, rigid design, high copyright risk and difficulty in continuous learning and evolution from use in the existing automatic presentation document generation technology.

Embodiments of the present application provide a video generation method and device, electronic equipment and storage medium, and relate to the technical field of artificial intelligence. The method comprises: performing image generation processing on a three-dimensional color feature vector, a motion instruction feature vector and an original pose feature vector by a preset image generator to obtain an initial candidate image; performing image discrimination processing on the initial candidate image and a reference image by a preset image discriminator to obtain an image discrimination label of the initial candidate image; performing frame discrimination processing on the initial candidate frame by a preset video discriminator to obtain a frame discrimination label of the initial candidate frame; filtering a target frame label from the frame discrimination label according to a preset frame label, filtering a target candidate frame from the initial candidate frame according to the target frame label, and performing merging processing on the target candidate frame to obtain a target generated three-dimensional video. The embodiments improve the continuity of the video.

This application provides a geospatial location positioning method, apparatus, device, and product, relating to the field of map positioning technology. The method includes: receiving multimodal data, including initial coordinates, image data, and text data; inputting the multimodal data into a multimodal large language model to obtain visual fingerprint features; determining the corresponding target road based on the initial coordinates; acquiring candidate images corresponding to multiple candidate points on the target road, and performing semantic consistency matching between the semantic features of the candidate images and the visual fingerprint features to select the target point from the multiple candidate points; and determining the target location based on the target point. By fusing multimodal data and utilizing a multimodal large language model to extract visual fingerprint features, combined with road constraints and semantic consistency matching, highly adaptable, efficient, and accurate geospatial location positioning is achieved in complex real-world environments.

The application discloses an image recognition-based fan abnormality monitoring method and device, equipment and a medium, relates to the technical field of image recognition, and is applied to a wind turbine nacelle, and comprises the following steps: collecting a raw image sequence of a first candidate image area inside the wind turbine nacelle in real time, eliminating environmental interference factors in the raw image sequence of the first candidate image area, and obtaining a target image sequence; extracting visual features in the target image sequence of the first candidate image area, identifying visual abnormalities in the target image sequence of the first candidate image area based on the visual features of the first candidate image area; and in the case where the visual abnormalities are identified in the target image sequence of the first candidate image area, outputting alarm information containing the visual abnormalities of the first candidate image area. The application solves the technical problem of poor monitoring effect in the current fan monitoring scheme.

The present application relates to a kind of methods for extracting image frame in video analysis, belong to image processing field.The present application includes extracting candidate image frame step, extract a candidate image frame sequence from the image frame sequence of a video, including color image frame conversion gray image frame, calculate the average gray of each image frame, calculate the variance of each image frame, calculate average variance, extract candidate image frame according to the variance of image frame etc.Sub-step.The present application also includes extracting final image frame step, extract final image frame sequence from candidate image frame sequence, including calculating the average absolute difference of adjacent two frames, calculate the average value of average absolute difference, calculate the variance of average absolute difference, extract final image frame according to average absolute difference etc.Sub-step.The present application extracts clear, the image frame sequence of larger difference using the characteristics of image frame in video clip, it is advantageous to subsequent image frame based on the processing of depth neural neural network etc.method for extracting.

This disclosure provides a method, apparatus, device, and storage medium for training a text-dense image recognition model, relating to the field of image processing technology, particularly computer vision, and applicable to scenarios such as information flow recommendation. The specific scheme is as follows: Text features are extracted from each original image in the original image set to obtain text statistical features and text region geometric features; based on the text statistical features and text region geometric features, a candidate image set is selected from the original image set according to a preset text density determination rule; for each candidate image set, a prompt word corresponding to each candidate image is constructed based on the text region geometric features and text density determination rule corresponding to each candidate image; using the candidate image set and prompt words, the initial multimodal large model is fine-tuned and trained to obtain a text-dense image recognition model. This scheme can reduce data annotation costs and the impact of sample imbalance, improving robustness and consistency.

Embodiments of the present disclosure provide a method for determining an image processing model, an electronic device and a storage medium, and relate to the technical field of artificial intelligence. The method comprises: obtaining first size information of a to-be-processed image; determining a search space of target model parameters of the image processing model, generating a plurality of first candidate image processing models; determining an algorithm requirement index corresponding to the first candidate image processing model according to model parameters of the first candidate image processing model, and screening a second candidate image processing model from the first candidate image processing model; determining a performance evaluation index corresponding to the second candidate image processing model based on the Gaussian entropy upper bound theorem according to the first size information and the model parameters of the second candidate image processing model; and determining the target image processing model according to the performance evaluation index. The method can quickly determine the image processing model without relying on a large amount of training data, can greatly reduce the time for manually designing the model, and effectively improves the model iteration efficiency.

This invention discloses an apparatus for performing user authentication, comprising: a controller; and a display device. The controller is configured to display a user authentication image comprising a plurality of images on the display device. The plurality of images includes candidate images for user authentication. The conditions for user authentication include conditions for selecting a candidate image. The candidate image satisfies predetermined candidate image conditions. The candidate image conditions specify the positional relationship between the candidate image and adjacent images that are different from the candidate image and satisfy predetermined adjacent image conditions.

The application belongs to the technical field of data processing, and discloses a camera calibration method, device, equipment and storage medium; the method comprises the following steps: initializing a camera model to obtain an initial camera model, and inputting a calibration board image sequence of the camera model into the initial camera model; screening the calibration board image sequence through the initial camera model to obtain a reference frame image and a residual image; obtaining an initial calibration result of the current initial camera model according to the reference frame image, screening the residual image according to the initial calibration result to obtain a candidate image set corresponding to the initial camera model; determining a target image set from the candidate image sets corresponding to multiple initial camera models according to a preset evaluation function, and obtaining a calibration result of the camera model according to the target image set; the application adaptively selects a reference image from the calibration board image sequence, evaluates the angle difference of the reference image to obtain an optimal image set, optimizes the camera model based on the optimal image set, and obtains a more accurate calibration result.

This application discloses a motion focus switching method and display device. The method includes: acquiring a first image captured by an image acquisition device during user follow-up training; performing human body recognition and motion detection on the first image; acquiring a candidate image, which refers to an image of a person currently performing a preset action on a target part of the human body; determining a target image based on the candidate image and the first image currently locked by the motion focus; and switching the motion focus in the follow-up training interface displayed on the monitor to the target image. This application is designed for multi-person follow-up training scenarios, where any person can recall the motion focus back to themselves by performing a preset action, achieving dynamic and accurate matching and switching of the motion focus, improving the intelligence, flexibility, and accuracy of motion focus switching, thereby providing users with a better follow-up training experience.

The application discloses a target object detection method and device, a storage medium and an electronic device, and relates to the field of computer vision and wild animal monitoring. The method comprises the following steps: collecting positioning data and environment data synchronized with an image frame; mapping the positioning data and the environment data into a conditional vector through a multilayer perception machine; linearly modulating a target convolution layer of a convolutional neural network and a target network layer of a pyramid network in a feature extraction process through the conditional vector to obtain a conditioned feature; generating a plurality of candidate image frames on the image frame according to the conditioned feature, and screening a target image frame from the plurality of candidate image frames according to the size and proportion of each candidate image frame; and detecting feature information related to a target object in the image frame according to the target image frame. The application solves the problem of insufficient detection precision and robustness of the prior art under small sample conditions for target objects with complex backgrounds and high species similarity, such as long-armed apes in tropical rainforests.

A technique is provided to generate highly accurate localization of objects moving along a trajectory. In one technique, a specific image associated with a moving object is identified. A set of candidate images is selected from multiple images used to train a neural network. For each candidate image in the set of candidate images, (1) the output from the neural network is generated based on inputting the specific image and each of the candidate images into the neural network, (2) the predicted position of the specific image is determined based on the output and the position associated with each of the candidate images, and (3) the predicted position is added to the set of predicted positions. The set of predicted positions is aggregated to generate an aggregated position for the specific image.

This invention belongs to the field of artificial intelligence technology and relates to a closed-loop emotion-aligned image editing method and system based on a multimodal large model. The method includes: S1: data reception; S2: instruction parsing; S3: current emotion evaluation and emotion transformation index construction; S4: retrieval and reversible patch loading; S5: semantic mask generation; S6: attention mask generation; S7: mask fusion; S8: candidate image generation; S9: candidate image evaluation; S10: stopping criterion judgment, determining whether the stopping criterion is met. If the stopping criterion is not met, self-correction is performed, and after self-correction, the process returns to step S6 until the stopping criterion is met or the number of iterations is reached to obtain the edited image. It can stably generate edited images that conform to the target emotion while ensuring content consistency, providing a new technical path for building a highly reliable, controllable, and deployable multimodal emotion-oriented image editing system.

A kind of uninterrupted face detection method, electronic equipment and storage medium are extracted from original image single color component form initial image, with predefined face template on initial image sliding, extract to obtain the multiple pre-check images of different regions corresponding to initial image, pre-processing is carried out to pre-check image, obtain candidate image, pre-screening is carried out to pre-check image and / or candidate image, remove the image meeting screening condition, the image not meeting screening condition is put into image data set, the matching degree of image in image data set and face template is calculated, the best matching region of initial image is determined, face is judged according to the matching degree of best matching region, if meeting matching condition, then it is judged that face is detected in original image.The present application saves computing resources, reduces power consumption, improves detection efficiency, and is suitable for low-power and low-storage demand uninterrupted detection application scenarios.

A fast face capture system and process for identifying an individual as the individual walks through an area. A set of raw images is streamed from at least one camera for detecting individuals entering the area. The images are searched for a face. If a face is detected, a tracking ID is assigned; and a timer and face tracking is commenced to obtain a sequence of candidate images for each individual as the individual walks through the area. A maximum quality image is selected from the candidate images for each individual based on at least one quality metric, the elapsed time, and the quality select count of the candidate images. The maximum quality image is submitted for matching with a verified image of the person. The invention has particular application to security check points for quickly matching the face of the moving person with a previously acquired and verified identity.

This application discloses a method and apparatus for optimizing promotional images. The apparatus includes a promotional image optimization system, a computer-readable storage medium, and a computer program product. The optimization method includes: performing risk detection on the promotional image to be optimized to obtain a first risk detection result; generating modification suggestions for the promotional image based on the first risk detection result; adjusting the promotional image based on the modification suggestions to obtain multiple candidate images; performing risk detection on each candidate image to obtain a second risk detection result for each candidate image; and selecting the optimal candidate image from the multiple candidate images as the optimized promotional image based on the second risk detection results of each candidate image. This application's promotional image optimization method achieves simultaneous identification of security and legal risks, improving the comprehensiveness of risk management; simultaneously, by utilizing an automated processing mechanism, it effectively reduces manual intervention and further optimizes processing efficiency.

The invention discloses a multi-modal target re-identification method based on semantic logic and topology collaborative deduction, and relates to the technical field of computer vision, and the method comprises the steps: extracting a to-be-queried multi-modal target image and a structured text of each group of candidate multi-modal target images in a candidate image set through employing a large language model; respectively inputting the to-be-queried multi-modal target image, each group of candidate multi-modal target images and the corresponding structured texts into a pre-trained feature extraction network, and extracting to obtain multi-modal alignment features; and calculating distance metric values of the multi-modal alignment features of the to-be-queried multi-modal target image and each group of candidate multi-modal target images, and determining the identity category corresponding to the candidate multi-modal target image with the minimum distance metric value as the identity category of the to-be-queried multi-modal target image. According to the method, through the synergistic effect of semantic logic and topology, the robustness and discrimination capability of multi-modal target re-identification in a complex scene are remarkably improved.

The application provides an AI-based graphic design intelligent layout method and system, which comprises the following steps: performing semantic analysis on design text to extract a text semantic vector and a text semantic weight value; subsequently, combining the above results, performing semantic matching and visual carrying capacity analysis in a candidate image library to select the most suitable target image and its text carrying area; then, calculating a layout center based on the local visual activity distribution of the carrying area, determining the title area proportion according to the regional geometric characteristics and the visual activity average value, and generating a layout structure containing a title, an explanation and auxiliary information; finally, correcting the text box position according to the stability relationship between the layout structure and the layout center, completing the text layout and design file output. The application realizes the full-process automation from text semantic understanding to image matching and layout generation, and effectively solves the problems of low efficiency and incoordination between graphics and text in traditional manual or template layout.

The application relates to the technical field of automatic driving, and particularly provides a target identification method and device based on a visual language model and a medium, aiming to solve the problem of how to accurately identify a target. The method provided by the application comprises the following steps: acquiring an image of a driving environment of an intelligent device and a question text corresponding to the image, the question text being used for describing a question about a target to be identified in the driving environment; inputting the image and the question text into a visual language model for processing to obtain an answer text; the visual language model is acquired in the following manner: acquiring a candidate image containing the target to be identified from a driving environment video; checking the candidate image by using a to-be-trained model, if the checking is passed, taking the candidate image as an image sample, the to-be-trained model being a visual language model after pre-training; and performing incremental training on the to-be-trained model by using the image sample and its labeled information. Through the above method, target information of an environmental target can be accurately identified from the image of the driving environment of the intelligent device.

This disclosure provides a game image processing method. In response to a first setting instruction, at least one target trigger condition is determined based on at least one game information. When the target trigger condition is detected, a screenshot prompt is displayed in a graphical user interface. In response to a first screenshot instruction, an image selection interface corresponding to a first screenshot mode is displayed. The image selection interface includes a candidate image sequence determined based on the target game event. In response to a save instruction, the target game image determined from the image selection interface is saved according to the target format corresponding to the save instruction. This enables the system to intelligently identify the target trigger condition based on the game information set by the user, and actively prompt the user to take a screenshot when the condition is detected, thereby providing a candidate image sequence containing the screen before and after the trigger for the user to select and save.

Provided are a multimodal fusion-based high-definition map feature recognition method and apparatus, a device, and a medium. The method includes determining attribute characteristics, pixel registration characteristics, and hybrid registration characteristics of a target map feature based on point cloud data and at least two types of candidate image data of the target map feature; determining a pixel correspondence of the target map feature based on the pixel registration characteristics of the target map feature and determining a hybrid correspondence of the target map feature based on the hybrid registration characteristics of the target map feature; and fusing the attribute characteristics of the target map feature based on the pixel correspondence and the hybrid correspondence to obtain a fused characteristic of the target map feature and determining a feature category of the target map feature based on the fused characteristic of the target map feature.

This invention discloses a method for precise focusing of microbial microscopic images, relating to the fields of clinical microbiology microscopic image processing and computer vision technology. The method includes: acquiring multiple microbial microscopic images at different objective focal lengths; calculating a sharpness score using a gradient statistical method combining spatial weighted feature enhancement and adaptive threshold noise reduction for overall coarse selection; performing spatial-frequency dual-domain feature enhancement on each candidate microbial microscopic image; performing microbial region detection and generating a mask that only covers the microbial region; using the mask to block pixels outside the microbial region, calculating the sharpness score, and selecting the candidate image with the highest sharpness score as the precisely focused image. This invention ensures the acquisition of clear microscopic images of the microorganisms themselves, improving the accuracy and reliability of microbial detection.