85 results about "Image geometry" patented technology

The invention relates to the technical field of power equipment state monitoring, and particularly discloses a transformer fault diagnosis method and system based on image recognition, and the method comprises the steps: collecting a transformer multi-mode image sequence in real time, and carrying out the definition and part integrity evaluation and screening to form an initial image set; performing multi-scale space registration on the initial image set and a transformer normal state standard template to generate a reference image, and reversely deriving a displacement vector field based on pixel-level difference; carrying out smooth optimization and geometric reconstruction on the displacement vector field under the geometric constraint of the transformer structure, and generating a correction image with a real structure; fault feature enhancement is carried out in a gradient domain of the corrected image, a fault area is identified through matching of multichannel feature extraction and a transformer typical fault feature library, and a diagnosis report integrating fault types, confidence coefficients and geometric parameters is generated; according to the method, the problem of image geometric deformation caused by shooting condition differences is effectively solved, and the accuracy and reliability of fault identification are improved.

The invention relates to the technical field of image processing, and discloses a fluorescent lifetime microscopic image large-view-field splicing method and system and electronic equipment, and the method comprises the steps: constructing a tissue region mask for a to-be-spliced image, and constructing a vignetting model in a tissue region to complete vignetting correction; counting brightness indexes in an organization area to determine a global brightness reference, and realizing image group brightness unification through global zooming and single image brightness adaptive correction; further realizing alignment of adjacent images through feature point matching, and finishing image block splicing in combination with a minimum color difference suture fusion method to obtain an image band; an overlapping area is extracted from an image belt to generate an effective content mask, relative displacement is estimated by using a phase correlation method after low-frequency suppression, image belt splicing is completed by multiplexing a minimum color difference suture fusion method, a large-view-field spliced image is output, the automation degree and robustness of splicing are improved, the spliced image is geometrically consistent and visually seamless, and the splicing efficiency is improved. And the requirements of medical research on high-resolution and large-field-of-view fluorescence lifetime microscopic images are met.

The invention relates to a GIS (Gas Insulated Switchgear) equipment mechanical defect diagnosis method based on a Gramb angle field and a dual-channel PCNN-Attention neural network, and belongs to the technical field of gas insulated switchgear mechanical vibration defect diagnosis. The method solves the problems that traditional diagnosis depends on artificial feature extraction, so that subjectivity is high, information mining is insufficient, and defect severity evaluation is missing. According to the technical scheme, the method comprises the steps that a one-dimensional vibration signal is converted into a GASF two-dimensional image and a GADF two-dimensional image through a GASF field so as to completely reserve time sequence topological features; carrying out data enhancement by adopting an image geometric transformation technology so as to improve the generalization ability of the model; and a dual-channel PCNN-Attention model is constructed, and synchronous intelligent identification of defect types and severity is realized through parallel feature extraction and dynamic weight optimization of an attention mechanism. According to the method, the diagnosis accuracy, reliability and adaptive capacity are improved, and support is provided for equipment state operation and maintenance.

The invention relates to the technical field of glass bottle detection, and discloses a glass bottle bottom defect detection method which comprises the following steps: building a detection platform, and adopting a multi-view imaging module of an annular LED and oblique light supplement combined light source, a vertical camera and 3-4 oblique cameras; after the to-be-detected glass bottle is positioned, synchronously collecting a bottle bottom full-view image set; graying, adaptive median filtering, CLAHE histogram equalization and Otsu binarization preprocessing are carried out on the image; extracting geometric and textural features of the image, inputting the geometric and textural features into an improved YOLOv5 model for identification and classification, and judging defects by combining with multi-view result fusion; and sorting the glass bottles according to a detection result and generating a traceable report. The identification rate of tiny defects is larger than or equal to 99%, the detection time of a single bottle is smaller than or equal to 0.5 second, and the method is compatible with glass bottles with the diameter of 30-100 mm, is suitable for large-scale quality control in the fields of food and beverage, medicine packaging and the like, and has extremely high practical value.

The invention relates to the field of image recognition, discloses an image recognition method and an image recognition system, and systematically solves the core pain point of traditional document recognition through optical-topology fusion processing and a dynamic resource allocation mechanism. The image recognition method is composed of an acquisition module, a grid module, a phase module, a setting module and a distribution module, pixel brightness is calculated and coded based on document RGB data, and a two-dimensional coding matrix is generated; constructing a geometric correction grid, forming a nonlinear constraint field, and enhancing the anti-deformation capability of the image; generating spiral phase light waves in the constraint field by using a spatial light modulator, and generating a time-varying phase map; determining a local topology index by detecting the number of phase jump times; and extracting the closed boundary region as a character block, and generating an analysis result file. The system breaks through traditional limitation, improves image geometric correction precision, feature extraction sensitivity and boundary judgment accuracy, efficiently completes document analysis, and is suitable for scenes such as document digitization and information retrieval.

The invention relates to the technical field of industrial visual inspection, and discloses an asynchronous exposure double-view-field imaging system and method, and the system can achieve the merging of independent view field images collected by two parallel lenses into a single linear array camera through a reflector and a non-polarization beam splitting cube based on the asynchronous exposure double-view-field imaging system. Meanwhile, in cooperation with precise time sequence control, the system drives two independent stroboscopic light sources to be synchronous with line trigger signals of the camera, different view fields are alternately illuminated in odd-even line periods, the camera is made to perform asynchronous exposure, and two paths of image information are collected in a time-sharing mode; and finally, reconstructing two complete, independent and seamlessly spliced detection images from the interleaved data streams through an odd-even line separation algorithm. According to the invention, with the hardware cost of a single camera, the detection visual field is obviously expanded, the detection efficiency and the image geometric precision under the high-line scanning frequency are ensured, and the excellent balance of the performance, the cost and the efficiency is realized.

The invention discloses a target positioning method and system based on target height correction monocular distance measurement, and the method comprises the steps: carrying out the calibration of a ship-borne camera through a data set, employing a YOLO target detection model, carrying out the detection of video data collected by the ship-borne camera, and obtaining the bounding box information of a target; and secondly, according to bounding box information, calculating a yaw angle of the target relative to a main point of the shipborne camera, and obtaining an azimuth angle by combining an orientation angle of the main point of the shipborne camera. And then height correction is carried out to measure the distance between the target ship and the shipborne camera. And finally, according to the azimuth angle and the distance, combining the latitude and longitude coordinates of the shipborne camera to obtain the actual geographic latitude and longitude coordinates of the target ship. According to the method, the goodness of fit between distance measurement and actual image geometry is improved, the actual object distance is reflected more accurately, and accurate and efficient target positioning is carried out.

The invention discloses a multi-scale feature fusion-based atmospheric turbulence video correction method and system, and the method comprises the steps: extracting shallow texture, middle structure and high semantic information through a multi-scale feature pyramid, and carrying out the channel weighting processing of different frequency components through the combination of frequency adaptive enhancement and a deformable attention alignment mechanism; adaptive optimization is realized for turbulence sensitivity difference, feature level alignment is completed under the guidance of optical flow, tilt distortion is eliminated, and reference features are dynamically maintained and turbulence noise accumulation is inhibited by adopting a gating reference updating module. In addition, cross-frame space-time fusion is completed through a multi-head time channel self-attention mechanism, time consistency is improved, and image geometry and high-frequency details are gradually recovered by using a two-stage decoder. According to the method, the definition of the turbulence degraded image can be remarkably improved, and technical support is provided for high-precision imaging.

The invention belongs to the technical field of wireless communication, and particularly relates to a machine vision integrated communication enhancement integration method and system, a communication part and a machine vision part are equipped, and target equipment can be identified by the communication part and the machine vision part; when the machine vision-fused communication enhancement integrated system deals with massive equipment access, accurate identification and management of equipment are realized through a multi-mode multi-level identification system, a dynamic association algorithm and a collaborative management mechanism; the equipment identified by the communication part is matched with the equipment identified by the machine vision part to ensure that the target equipment identified by the communication part and the machine vision part is the same in the processing flow; perceptual data obtained by the communication part and including parameters such as distance, position / angle, speed and the like of the target equipment are fused with perceptual data obtained by the machine vision part and including parameters such as images, geometric features, position / angle, speed, distance and the like of the target equipment.

The application discloses a method for intelligent component identification and segmentation of an aircraft CAD model based on three-mode fusion, comprising: obtaining an aircraft CAD model with a boundary representation structure; constructing a multi-modal training data set that fuses surface images, geometric parameters and component semantic labels; training a multi-modal identification model; using the model for preliminary identification; for low confidence or small area surfaces, starting a geometric rule auxiliary module based on geometric topology and parameterization rules for semantic completion and correction to form a final segmentation result; automatically extracting key geometric feature lines; and converting and outputting into a structured engineering semantic format that can be used by grid generation software. The application realizes full automation of component identification, solves the problems of low efficiency and consistency caused by manual dependence, improves the identification integrity and engineering practicability of complex geometry, and the output result can directly drive subsequent grid generation, thereby significantly improving the automation level of the digital simulation pre-processing process.

The application provides a task-driven defogging perception method and system, and belongs to the technical field of cooperative perception of Internet of Vehicles. The method comprises the following steps: collecting a first foggy image stream and a second foggy image stream; obtaining a first processed image by intercepting and processing the first foggy image stream; obtaining a second processed image by intercepting and processing the second foggy image stream; respectively performing geometric alignment on the first processed image and the second processed image to obtain a first aligned image and a second aligned image; performing defogging and image enhancement processing on the first aligned image and extracting a first feature; performing defogging and image enhancement processing on the second aligned image and extracting a second feature; respectively repairing the first feature and the second feature by using a diffusion model; fusing the repaired first feature and the second feature, and determining a final perception result based on the fused feature. The application improves the robustness of automatic driving and balances the low power consumption of AI intelligent glasses and the high performance computing demand of a vehicle-mounted system.

The invention relates to the field of remote sensing image processing and terrain correction, and particularly discloses a radar geometric field-driven SAR distortion information adaptive neighborhood propagation identification method, which comprises the following steps of: S1, acquiring a Golden digital elevation model of a research area and a pixel-by-pixel ellipsoid incident angle graph generated by SAR satellite precision orbit parameters and imaging geometry, and carrying out geometric unified processing on the Golden digital elevation model and the pixel-by-pixel ellipsoid incident angle graph; s2, calculating a terrain parameter, an effective slope angle, an R geometric response field and an R field directivity gradient based on the unified raster data; s3, main distortion region classification is carried out according to the effective slope angle and the pixel-by-pixel ellipsoid incident angle; s4, defining an adjacent gradient and constructing a self-adaptive threshold value and a shadow self-adaptive threshold value; and S5, performing passive region propagation from the boundary of the main distortion region, and generating a final classification grid map after space overlapping correction. According to the method, the defect of blind propagation of the existing method is overcome, and high-precision automatic identification of SAR geometric distortion under complex terrains is realized.

The application discloses a kind of underwater concrete crack width automation measurement system and method based on double laser, it is related to underwater structure health monitoring and optical measurement technical field.The application includes: double laser projection module, for projecting a pair of circular laser points with fixed physical spacing to underwater concrete surface to be measured as physical scale reference;Image acquisition module, including underwater camera, for collecting image containing laser point and crack area;Data processing module, for performing underwater camera calibration, laser point detection, crack image segmentation and crack physical width calculation.The method corrects the image geometric distortion caused by "water-glass-air" refraction by establishing an accurate underwater multi-interface refraction imaging model;Nonlinear mapping relationship from image pixels to physical space is established using double laser points;Based on unsupervised crack segmentation model, crack area is automatically segmented, the physical width of each position of crack is calculated and width distribution cloud chart is generated.

This invention relates to a method for stitching and adjusting the grayscale of multispectral images using a filter array, belonging to the field of aerial image processing technology. The purpose of this invention is to utilize the SURF algorithm to register projected images, using the mean of the grayscale average ratio of overlapping regions in single-band images as the grayscale adjustment coefficient, and adjusting the grayscale of the registered images sequentially based on the grayscale of a reference image. This invention comprises three steps: determining image geometric relationships, adjusting image grayscale, and stitching images. This invention solves the problem of grayscale differences in single-band multispectral stitched images and minimizes the loss of spectral information of ground features. It not only effectively solves the problem of inconsistent grayscale of ground features in images but also maximizes the preservation of spectral information characteristics of ground features in different bands.

The application is suitable for the technical field of power equipment state monitoring, and provides a flow blocking valve state identification method, system, device and program product. The method comprises: acquiring a reference image and a to-be-inspected image of a flow blocking valve; identifying first structure key points from the reference image, and identifying second structure key points from the to-be-inspected image; performing geometric correction on the to-be-inspected image based on the first structure key points and the second structure key points, to obtain a corrected image consistent with the reference image in geometric space; calculating a reference angle in the reference image and a current angle in the corrected image respectively, and determining the state of the flow blocking valve according to the angle difference between the current angle and the reference angle. The accuracy and universality of flow blocking valve state identification are improved.

The application discloses an image processing method and an electronic device, and relates to the technical field of image processing, wherein the method comprises the following steps: extracting a first geometric edge line set of an initial depth image according to the initial depth image corresponding to a target monocular image; screening geometric edge lines with gradient scores greater than a target score threshold from a depth gradient image generated according to the initial depth image and an initial edge line set of the target monocular image to determine a second geometric edge line set; and fusing the first geometric edge line set and the second geometric edge line set to determine image geometric edge lines of the target monocular image and construct a final geometric edge line set. The method solves the problems of serious texture interference, insufficient utilization of depth information, difficulty in balancing the number and purity of edge lines, and insufficient real-time performance in the prior art, and achieves the technical effects of filtering texture noise while ensuring the number and robustness of edge lines and meeting the needs of multi-sensor fusion tasks.

The application discloses a tunnel water mist environment under inspection image de-scattering enhancement and reconstruction method, and belongs to the technical field of image processing. The method comprises the following steps: acquiring a polarized image sequence and pose information collected synchronously; decoupling a scene by using a fusion neural network to output a fog-free image, a geometric gradient map and a water mist concentration map; calculating a frequency domain fusion weight map by using a frequency domain attention generation network; performing multi-scale frequency domain decomposition on the fog-free image to obtain high-frequency components and low-frequency components; performing pixel-level weighted fusion and inverse transformation according to the weight map to generate a preliminary enhanced image and perform color correction. The application adopts a physical constraint deep learning and a frequency domain adaptive fusion mechanism, can effectively eliminate the interference of tunnel water mist scattering, and improves the definition, detail performance and color authenticity of the inspection image, thereby providing high-quality visual support for tunnel safety monitoring.

The application discloses a motor insert sheet defect detection method based on multi-sensor fusion, relates to the technical field of motor insert sheet defect detection, and comprises the following steps: in the presence of boundary penetration, calculating edge matching mismatch degree, edge jump rate, edge inversion proportion and edge closure deviation to form a geometric conflict feature set of an insert sheet edge, which is used for representing image geometric expression differences caused by boundary penetration; inputting the geometric conflict feature set into a fusion confidence evaluation function to obtain a fusion confidence evaluation factor, which is used for identifying whether the insert sheet edge has a fusion distortion risk. The application solves the structural expression conflict problem caused by boundary penetration in the multi-view imaging fusion process of the motor insert sheet, constructs a fusion confidence evaluation mechanism to identify the fusion distortion risk, dynamically limits the image participation fusion condition, and thus improves the recognition accuracy of the edge structure and the stability of the fusion result in the insert sheet defect detection.

This invention relates to a method and medium for detecting building changes based on UAV imagery. By using UAV images of the same target area acquired at different times, the method incorporates UAV ground control points and aerial POS information into an aerial triangulation system to jointly calculate the camera's exterior orientation elements and imaging geometry. Orthophotos of the preceding and following times are generated under a unified coordinate reference, and building probability maps of the preceding and following times are obtained through processing. The merged orthophoto obtained after combining the preceding and following times UAV images is used to construct multiple object units based on cross-temporal consistency. Furthermore, the average first and second semantic confidence scores corresponding to the preceding and following times orthophotos are obtained. Based on whether the two average confidence scores meet three preset judgment conditions, each object unit is determined and labeled as belonging to a specific type of building change area. Thus, stable and accurate detection of building changes based on UAV imagery is achieved.

The application relates to the technical field of industrial vision detection, and discloses an asynchronous exposure double-view field imaging system and an imaging method. Based on the asynchronous exposure double-view field imaging system, independent view field images collected by two parallel lenses can be combined into a single linear array camera through a mirror and a non-polarization beam splitting cube. Meanwhile, cooperating with precise timing control, the system drives two independent stroboscopic light sources and a camera line trigger signal to be synchronous, alternately illuminates different view fields in an odd-even line cycle, makes the camera asynchronously expose and collect two-way image information at different times, and finally reconstructs two complete, independent and seamlessly spliced detection images from the interwoven data stream through an odd-even line separation algorithm. The application realizes significant expansion of a detection view field with a single camera hardware cost, guarantees detection efficiency and image geometric accuracy under a high line scanning frequency, and realizes excellent balance of performance, cost and efficiency.

The application discloses a kind of hierarchical light field image encoding methods, comprising: enhancement layer coding network utilizes the enhancement layer coding strategy based on domain-specific reference to carry out interlayer prediction and residual compression to spatial feature, and angle feature is generated at decoding end, and the light field image is encoded;Interlayer spatial prediction module is constructed, the co-location and local self-similarity within reference spatial feature are modeled simultaneously, to implicitly reconstruct high-frequency details in high-resolution light field representation, predict spatial latent representation;Spatial-angle context-aware reconstruction module is constructed, long-distance and short-distance spatial-angle correlation is explored in long-range and short-range context-aware branch respectively to model light field image geometry and preserve local details, and the light field image of enhancement layer is reconstructed.The application encodes light field image of different spatial resolution in an end-to-end manner, and improves the encoding performance of spatial hierarchical light field image.

The invention discloses an image distortion real-time correction method and system for an industrial infrared camera, and relates to the technical field of image processing, and the method comprises the following steps: constructing a distortion expression of a lens, and carrying out the first correction of an image according to the distortion expression of the lens, and constructing a response distortion expression of the mobile monitoring object according to the expression of the lens distortion amount, and carrying out second correction on the image. According to the invention, through constructing the lens inherent distortion expression and the dynamic distortion constructed for the mobile monitoring object, the static lens distortion and the composite distortion introduced by the dynamic scene are respectively eliminated, the geometric accuracy of the image is obviously improved, the deformation of the mobile object caused by the lens distortion can be quantified in response to the distortion expression, and the accuracy of the image is improved. The misjudgment of distortion aggravated by object motion is avoided, the complex distortion is decoupled into two parts of lens inherent and object response by a step-by-step correction strategy, the calculation complexity of single correction is reduced, and the real-time requirement is considered.

The application relates to the technical field of power equipment state monitoring, and particularly discloses a transformer fault diagnosis method and system based on image recognition, which comprises the following steps: collecting a transformer multi-modal image sequence in real time, screening through clarity and component integrity evaluation to form an initial image set; performing multi-scale space registration on the initial image set and a transformer normal state standard template to generate a reference image, and inversely deducing a displacement vector field based on pixel-level differences; under the constraint of a transformer structure geometry, smoothing, optimizing and geometrically reconstructing the displacement vector field to generate a structure true correction image; strengthening a fault feature in the gradient domain of the correction image, identifying a fault area through multi-channel feature extraction and matching with a transformer typical fault feature library, and generating a diagnosis report of comprehensive fault types, confidence and geometric parameters; the application effectively solves the problem of image geometric deformation caused by differences in shooting conditions, and improves the accuracy and reliability of fault identification.

This application discloses an image processing method and electronic device, relating to the field of image processing technology. The method includes extracting a first set of geometric edge lines from an initial depth image corresponding to a target monocular image; filtering geometric edge lines with gradient scores greater than a target score threshold based on a depth gradient map generated from the initial depth image and the initial set of edge lines in the target monocular image to determine a second set of geometric edge lines; fusing the first and second sets of geometric edge lines to determine the image geometric edge lines of the target monocular image, and constructing a final set of geometric edge lines. This method solves the problems of severe texture interference, insufficient utilization of depth information, difficulty in balancing the number and purity of edge lines, and insufficient real-time performance in related technologies. It achieves the technical effect of filtering texture noise while ensuring the number and robustness of edge lines, meeting the requirements of multi-sensor fusion tasks.

This invention discloses an image geometric distortion correction method based on the Unet network, which solves the problem that current methods mainly target single distortion types and have significant limitations when correcting mixed distortions. The key technical points are: constructing a distortion correction model based on the Unet network; selecting place365 images and a pre-generated checkerboard image as image data sources; considering radial and tangential distortion to generate distorted images and their coordinate difference vectors as datasets; setting the distortion correction model according to the experimental environment parameters; training and testing the distortion correction model using the constructed dataset; acquiring the actual dataset and predicting the coordinate difference vectors of the distorted images; and using an image resampling algorithm to resample the corrected image based on the predicted coordinate difference vectors and the actual distorted image. This Unet network-based image geometric distortion correction method considers both radial and tangential distortions, has fewer limitations, and achieves better overall correction results.

This invention provides a method for extracting tall buildings and their heights from a single-view high-resolution SAR image. Specifically, the method involves preprocessing the SAR image; extracting salient features from the SAR image, including bright spots and bright lines; generating facade regularity components, facade bright line components, and secondary scattering components based on semantic information and salient features; formulating production system rules based on the spatial relationships between different components to generate tall building hypotheses; removing unacceptable tall building hypotheses to obtain the final tall building extraction result; calculating the range length of each tall building in the SAR image based on the final extraction result; and converting the range length of the tall buildings into their actual heights based on the imaging geometry in the SAR image. This method does not require a large amount of training data, only a single-view SAR image and its incident angle information, to quickly extract complex and diverse tall buildings and their heights.

The invention relates to the technical field of vehicle calibration, and discloses a calibration method and device of a vehicle-mounted image system, a vehicle and a storage medium. If yes, when the vehicle speed is reduced to a preset range and a ground marking line conforming to preset geometric characteristics is detected, triggering to collect to-be-calibrated images shot by at least two cameras installed on the vehicle in different directions; based on each to-be-calibrated image, identifying a ground marking line in a view field of a corresponding shooting view angle, and determining position deviation of the same physical ground marking line under different shooting view angles; and based on the position deviation, the to-be-compensated view angle image is translated by taking the ground marking position of the reference view angle image as a reference, and when the ground marking in a panorama obtained by splicing the translated images of all view angles meets a continuity requirement, an image geometric transformation parameter corresponding to translation is taken as a current target calibration parameter of the vehicle.

This invention belongs to the field of digital image processing and computer vision technology, and relates to a method and related equipment for geometric correction of document images. The method includes: preprocessing a distorted document image to extract document regions from the image; using a deep learning model to extract geometric features from the document regions, obtaining structural feature lines and local angular distortion information in the image; representing the local angular distortion information as a Beltrami coefficient field, and constructing a geometric constraint model based on the structural feature lines and the Beltrami coefficient field; obtaining a quasi-conformal mapping by solving the Beltrami equation according to the geometric constraint model, and mapping the distorted image domain to a standard plane image domain based on the quasi-conformal mapping, thereby obtaining the corrected document image. This invention extracts macroscopic structural feature lines and microscopic local distortion information through a deep learning model, and combines this with quasi-conformal mapping theory to construct a geometric constraint model, which can maintain the consistency of local character shapes while correcting the overall structure of the document image, effectively improving the quality of geometric restoration of the document image.

This invention relates to the field of remote sensing image processing and terrain correction, specifically disclosing a radar geometry-driven adaptive neighborhood propagation identification method for SAR distortion information. The method includes: S1, acquiring the Copernicus digital elevation model of the study area and a pixel-by-pixel ellipsoidal incidence angle map generated from the precise orbital parameters and imaging geometry of the SAR satellite, and performing geometric unification processing; S2, calculating terrain parameters, effective slope angle, R-geometric response field, and R-field directional gradient based on the unified raster data; S3, classifying the main distortion area according to the effective slope angle and pixel-by-pixel ellipsoidal incidence angle; S4, defining the neighboring gradient and constructing adaptive thresholds and shadow adaptive thresholds; S5, propagating the passive area from the boundary of the main distortion area, and generating the final classification raster map after spatial overlap correction. This invention overcomes the shortcomings of blind propagation in existing methods and achieves high-precision automatic identification of SAR geometric distortion under complex terrain.

The invention discloses an image recognition method and device and a program product, relates to the field of artificial intelligence, and is applied to the field of financial science and technology, and the method comprises the steps: carrying out the data preprocessing and feature extraction of an Nth image set, and obtaining a first feature vector, the Nth image set comprising a face image collected in an Nth time window; an Nth feature set is constructed according to the first feature vector and feature vectors of face images in an (N-1) th image set, and the (N-1) th image set comprises the face images collected in the (N-1) th time window; clustering the Nth feature set, and determining a target image template according to a plurality of clusters in an Nth clustering result; and recognizing the face synthesis image in the preset storage space according to the target image template. Through application of the method and the device, the problems that the generalization performance is reduced and the recognition result is inaccurate when a diversified face composite image attack is recognized due to insufficient robustness of image geometric transformation in related technologies are solved.