2266 results about "Multiple image" patented technology

The invention relates to an image processing technology, and discloses an efficient panoramic image splicing method and system based on multi-view fusion, and the method comprises the steps: collecting a plurality of images from different views; performing multi-scale feature extraction on each image, generating a feature descriptor for each extracted feature point, and matching a corresponding feature point pair; performing multi-view geometric constraint screening on the feature point pairs; according to the screened feature point pairs, estimating a homography matrix between adjacent images, and carrying out global optimization on the homography matrix; aligning all the images into the same coordinate system; determining an overlapping region between adjacent images; and according to the pixel information in the overlapping areas, fusing the overlapping areas by adopting a self-adaptive weighted fusion algorithm so as to splice the plurality of images into a panoramic image. The invention further discloses a control device and a computer readable storage medium. The invention aims to improve the efficiency and accuracy of generating the multi-view fused panoramic image.

The invention relates to the technical field of road engineering detection, in particular to a highway pavement crack image intelligent detection system and method, and the system comprises an image acquisition module, a deep learning module, an image enhancement module, a crack measurement and calculation module, a crack development trend prediction module and an information visualization module. The core innovation of the invention lies in that a differential geometry theory is introduced to construct a crack development trend prediction module, and the module comprises a crack characterization model based on a differential manifold, a multi-scale crack evolution tensor field analysis model and a nonlinear space-time crack development prediction model. A pavement is regarded as a two-dimensional differential manifold, cracks are represented as singular curves on the manifold, a multi-scale tensor field analysis technology and a non-linear kinetic equation are combined, accurate prediction of the future development trend of the cracks is achieved, the system supports multiple image acquisition modes, transverse, longitudinal and net cracks can be accurately detected, and the detection precision is high. The method adapts to complex illumination and background conditions, and predicts the expansion rate and severity change of the crack.

The invention relates to the technical field of machine vision structure health monitoring, and discloses a steel structure weld fatigue cracking real-time monitoring system based on machine vision. The system comprises a space-time registration and fusion module, a multi-scale feature analysis module, a health monitoring module, a crack deduction calculation module and a regulation and control strategy generation module. Performing space-time registration and pixel-level fusion through the visual data of the plurality of image sensors to generate a synchronous multi-source image stream; a multi-level welding seam characteristic spectrum is constructed through multi-scale characteristic analysis, and a welding seam structure knowledge base is dynamically updated; the knowledge base and the real-time characteristic spectrum are used for monitoring the welding seam health state, and abnormity is recognized; deducing a crack initiation position and an evolution path in combination with historical damage data; and real-time load information is fused to pre-estimate the remaining service life, and a structural integrity regulation and control strategy is generated online. According to the invention, high-precision fusion of the multi-source visual data and active prediction of the crack trend are realized, and the monitoring accuracy and the early warning capability are improved.

The invention provides an industrial product defect automatic classification method and system, and the method comprises the steps: collecting original industrial product defect image data, and constructing a labeled image sample set and an unlabeled image sample set; constructing a training image sample set based on the labeled image sample set and the unlabeled image sample set in combination with a plurality of image synthesis strategies; based on the training image sample set, introducing a transfer learning strategy and fusing an attention mechanism, and constructing and optimizing an industrial product defect classification model; performing semi-supervised joint training and online learning based on the training image sample set and the real-time small-batch image sample set; and constructing an industrial product defect identification log based on the real-time image flow sample set, the classification model parameters and the corresponding classification prediction function. On the basis of multi-strategy image enhancement and semi-supervised training, transfer learning and a channel attention mechanism are fused, expansion of industrial product defect image samples and fine defect identification are achieved, and the method is suitable for an intelligent defect detection system in various industrial manufacturing fields.

The invention relates to a crack tracking and predicting method based on point cloud registration, belongs to the technical field of image processing, and solves the problem that an existing three-dimensional point cloud model is low in precision and lacks automatic crack tracking capability. The method comprises the following steps: after preprocessing a plurality of images regularly acquired by an unmanned aerial vehicle, identifying the plurality of images by using a deep learning model to obtain crack positions, and then extracting crack features; performing dynamic blocking and differential feature extraction and matching on the plurality of images, and constructing a point cloud model; mapping the crack position and the crack feature obtained in each period to a point cloud model constructed in a corresponding period, and storing the point cloud models in a point cloud model library according to a time sequence; the point cloud models of every two adjacent periods are registered, and a change area of crack characteristics is analyzed; and predicting the change trend of the crack in the future time according to the environmental data of the time sequence, the crack characteristics and the crack characteristic change rate. And improvement of the accuracy of the point cloud model and automatic tracking of the crack are realized.

The invention discloses a cradle head tracking method, system and device, a storage medium and a program product, and relates to the technical field of cradle heads, and the method comprises the steps: obtaining the target position information of a tracking target in continuous multi-frame image frames collected by an image sensor, and the size of a detection frame, determining a pan-tilt rotation angle of a target pan-tilt carrying an image sensor when multiple image frames are acquired; determining a target position offset between the target position and the center of the image based on the target position information, and mapping the size of the detection frame into a target distance between the tracking target and the image sensor; inputting the target position offset and the holder rotation angle corresponding to the multiple image frames into a motion estimation model to obtain a predicted position; and determining a pan-tilt angle value of the target pan-tilt based on the predicted position and the target distance, and driving the target pan-tilt to rotate based on the pan-tilt angle value. According to the invention, the cradle head keeps stable tracking while quickly responding to target movement, so that the stability and reliability of cradle head tracking control are improved.

The invention relates to the field of image processing, in particular to a fire-fighting equipment fault automatic detection method and system based on image processing, and the method comprises the steps: collecting a plurality of images in a plurality of regions at the same time interval, and carrying out the graying of the images; pixel points are classified according to gradient distribution of gray level images in the area, and corrosion probability parameters of the area are obtained; obtaining a water leakage probability parameter according to the time sequence change of pixel points in the suspected corrosion area; and comparing the change conditions of the pixel points of the suspected corrosion area and the non-suspected corrosion area to obtain a water leakage correction parameter, obtaining a fault early warning parameter by combining the parameters, setting a threshold value, comparing the threshold value with the fault early warning parameter, judging whether a water leakage condition exists or not, and completing automatic fault detection of the fire-fighting equipment. The method can effectively distinguish environment condensate water interference and real leakage, and has anti-noise capability.

The invention relates to a glass flaw recognition method, system and equipment based on image enhancement and a medium. The method comprises the following steps: acquiring a multi-view image set of to-be-detected glass; respectively carrying out region-of-interest extraction on the vertical dark field image, the horizontal dark field image, the vertical bright field image and the horizontal bright field image to obtain multiple groups of image blocks; combining a plurality of image sub-blocks with the same position in each group of image blocks to obtain a standard image sub-block group; inputting the standard image sub-block group into a pre-trained glass flaw detection model to obtain a detection result and a classification probability corresponding to each image sub-block at the same position; and based on a preset channel fusion weight, performing weighted fusion on the detection results at the same position according to the corresponding classification probability to obtain a glass flaw recognition result at the corresponding position. According to the method, through region-of-interest extraction, channel separation feature extraction and fusion probability output under an independent view angle, the accuracy of glass flaw automatic identification is improved.

The invention discloses a detection method and device for intelligent visual detection of parts and a storage medium, belongs to the technical field of industrial quality detection, and aims to solve the problems that surface and internal defects of complex parts are difficult to identify synchronously and the identification precision is low. The method comprises the following steps: acquiring multi-modal image data obtained through structured light three-dimensional imaging and laser ultrasonic scanning; performing geometric registration and scale normalization on the image to generate a fused image; carrying out image preprocessing and edge analysis, and extracting a region of interest; dividing the candidate region of interest into a plurality of image blocks, and inputting the image blocks into an anomaly detection network; calculating a reconstruction error between the original image block and the reconstructed image block, and generating an abnormal scoring graph; and finally, extracting a defect area through image post-processing, and outputting information such as a defect type, a spatial position, a geometric dimension and a severity level. According to the method, the robustness and accuracy of multi-modal defect identification are improved, and the method is suitable for an online visual inspection task of an industrial production line.

The invention discloses a multi-degraded image recovery method based on frequency domain decomposition, and aims to solve the problems that a single model is difficult to deal with various image degradation and recovery processes of different frequency domains are mutually coupled in the prior art. According to the method, a degraded image is decomposed into a high-frequency space and a low-frequency space through fast Fourier transform, and a double-branch network architecture is adopted for targeted processing: for the high-frequency part, a high-frequency feature adaptive processing module HFPM is designed, and detail texture features are effectively extracted and interference is suppressed through feature enhancement and cross-layer fusion technologies; and for the low-frequency part, constructing a low-frequency feature conversion enhancement module LTEM, and capturing global context information by using cyclic convolution to improve the integrity of the structure contour. According to the method, decoupling processing of frequency domain features is realized, and the image restoration performance of the model in various degradation scenes such as rain removal, noise removal and defogging is remarkably improved through the synergistic effect of high-frequency detail enhancement and low-frequency structure optimization. Experimental results show that the method has excellent recovery effect and robustness when a plurality of image degradation tasks are processed at the same time, and can be effectively applied to visual tasks such as traffic accidents with high image quality requirements.

The invention aims to provide a multi-condition diffusion model-based Shengtuo calligraphy rubbing fusion restoration method, which comprises the following steps of: firstly, acquiring a plurality of Shengtuo calligraphy rubbing images from different sources, preprocessing the images, and constructing a standardized training data set; designing a multi-image feature extraction and alignment module to obtain a fused structural feature tensor; constructing a core network structure of the multi-conditional diffusion model, and supporting guidance of a plurality of conditional vectors; jointly inputting the obtained fusion structure feature tensor and a plurality of condition guide items into a diffusion model to generate a fused and repaired potential image; and finally, restoring the potential image obtained by reconstruction into a final output image through a reverse mapping module. According to the method, the problems of blurring, pen lacking, breakage and non-uniform styles during fusion and restoration of the rubbing image in the prior art are solved.

The invention discloses a texture perception state space modeling method for an image restoration task, and the method comprises the steps: 1, constructing a region selection mechanism based on texture complexity, and enabling the region selection mechanism to be used for distinguishing a flat region and a high-texture region in an image; 2, introducing a texture modulation mechanism, and performing explicit adjustment on a state transition matrix in the state space model; 3, enhancing the context modeling capability of the model through a multi-direction sensing module; and 4, by combining position embedding and a sequence modeling structure, the capability of the model in the aspects of image structure understanding and spatial information maintenance is improved. The method can effectively alleviate the problem of information loss when a traditional image restoration method processes texture details, improves the structure restoration capability of a complex region, gives consideration to the restoration quality and the calculation efficiency, is suitable for multiple image restoration scenes such as image super-resolution, image rain removal, low-light image enhancement and the like, and improves the image restoration efficiency. And the method has good engineering adaptability and actual deployment value.

The invention provides a three-dimensional reconstruction method, a data processing method, a rendering method and a device applied to the field of artificial intelligence, which are used for dividing a plurality of images in a scene into a plurality of areas in the scene and performing scene reconstruction based on at least one image of each area so as to realize block scene reconstruction. The method comprises the following steps: acquiring a plurality of images for a scene, wherein the scene comprises a plurality of areas; dividing the plurality of images into the plurality of areas according to the poses of the shooting equipment corresponding to the plurality of images to obtain at least one image of each area; obtaining at least one three-dimensional Gaussian body of each region according to the at least one image of each region, wherein the at least one three-dimensional Gaussian body is used for representing the region in the scene; and combining the at least one three-dimensional Gaussian body of each region to obtain a combined three-dimensional Gaussian body, and representing the scene by the combined three-dimensional Gaussian body.

The invention relates to a defect identification method based on a point cloud model, belongs to the technical field of image processing, and solves the problem that an existing three-dimensional point cloud model is low in precision and lacks dynamic defect identification capability. The method comprises the following steps: carrying out image dynamic partitioning and differentiation feature extraction and matching on a plurality of images acquired by an unmanned aerial vehicle, constructing a point cloud model, and storing the point cloud model in a point cloud model library according to a time sequence; obtaining a difference region by comparing the newest point cloud model with a historical point cloud model in a point cloud model library; and acquiring a local image of the difference region by using an unmanned aerial vehicle, preprocessing, and identifying a plurality of defects in the local image by using a deep learning model. The improvement of the accuracy of the point cloud model and the automatic identification of defects are realized.

Systems described herein use either multiple image sensors or one image sensor and a complementary sensor to provide underwater depth perception. The systems include a computing system that provides the underwater depth perception based on data sensed by the multiple image sensors or the one image sensor and the complementary sensor. The systems can include a submersible device (such as a submersible mobile machine) that includes a holder configured to hold the one image sensor. The holder can be configured to hold the computing system in addition to the one image sensor. And, in some embodiments, the holder is configured to hold the complementary sensor in addition to the computing system and the one image sensor. Alternatively, in some embodiments, the holder is configured to hold the multiple image sensors. And, the holder can be configured to hold the computing system in addition to the multiple image sensors.

The embodiment of the invention provides an ink line detection method and system, electronic equipment and a storage medium, and relates to the technical field of building construction, and the method comprises the steps: obtaining point cloud data corresponding to a building; generating a panorama according to the point cloud data; segmenting the panorama into a plurality of image blocks, and respectively inputting the image blocks into a pre-trained ink line detection model to obtain a corresponding detection result image; and obtaining the position of a target ink line in the building according to the point cloud data and the detection result graph corresponding to each image block. In this way, the panorama is generated through the point cloud data, the panorama is input into the ink line detection model for detection, the detection result graph is output, the position of the target ink line is restored according to the point cloud data corresponding to the ink line in the detection result graph, automatic and accurate detection of the ink line is achieved, manual detection is replaced, and the detection efficiency is improved. Error accumulation caused by ink line mixing under the conditions of cracks, recesses and the like is reduced, the measurement difficulty under the condition that the ink lines are abraded or shielded is reduced, and the detection efficiency and precision of the ink lines are improved.

The invention discloses a meniscus injury prediction method and system based on a neural network, and the method comprises the steps: obtaining a knee MRI image, carrying out the automatic segmentation of the knee MRI image based on a convolutional neural network, extracting a plurality of image features based on the segmented image, and obtaining a meniscus injury prediction result. And constructing a meniscus damage prediction model based on the SNN network model, and carrying out model training and verification. The meniscus injury diagnosis accuracy and diagnosis efficiency are improved, future injury risk prediction is achieved, valuable prediction information is provided for clinicians, prevention and treatment schemes are helped to be formulated, meanwhile, radiomics feature and load structure feature heat maps and a mixed attention mechanism are introduced, the clinical interpretability is enhanced, and the diagnosis accuracy and efficiency of meniscus injury are improved. And a full-automatic diagnosis process is realized.

A method of checking wing position in a CMS includes performing a calibration of a wing position supporting a camera relative to a vehicle to provide a desired field of view by capturing multiple images at different lighting conditions, extracting and storing a reference feature from each of the multiple images, triggering a wing position check, capturing a current image from the camera having a current position of the reference feature; sensing a current lighting condition at which the current image is captured, determining that one of the different lighting conditions is more similar to the current lighting condition, comparing the current position of the reference feature to the stored reference feature from the one of the multiple images generated under the one of the different lighting conditions, and outputting a result of the wing position check if a difference from the comparing step exceeds a threshold value.

The invention relates to the technical field of image data processing, in particular to an image adaptive optimization processing method and system for laser printing output, and the method comprises the steps: carrying out the multi-scale feature analysis and fusion of an input original scanning image, and obtaining a fused feature distribution mapping matrix; performing adaptive contrast enhancement based on the fused feature distribution mapping matrix to obtain an enhanced contrast image; performing edge feature extraction on the enhanced contrast image by using an improved multi-direction edge detection algorithm to obtain an edge feature image with an enhanced edge; performing adaptive local texture analysis and enhancement on the edge feature image to obtain a texture enhanced image; and carrying out printing adaptability optimization based on the texture enhanced image to obtain a final optimized output image. According to the technical scheme, the quality of the image printed and output by the laser printing equipment is comprehensively and remarkably improved from multiple image processing dimensions.

The invention discloses a real haze image defogging method based on a haze degradation model. The method comprises the following steps: constructing a haze degradation model fusing multiple scattering effects and multiple image degradation factors; using the haze degradation model to construct a training data set including the clear image and the corresponding pseudo haze image; constructing a defogging network for a real haze scene; training the defogging network by adopting the training data set until a preset loss function is converged; and inputting a to-be-defogged image into the trained defogging network to obtain a defogging result. According to the haze degradation model constructed by the invention, the difference between a synthetic domain and a real domain is effectively relieved; a designed space-frequency hybrid module improves the adaptability of the model to complex degradation characteristics; the prior-guided feed-forward network fully excavates and fuses dark channel prior information, and the sensing and modeling capability of the model to the haze area is effectively enhanced.

The invention relates to the technical field of image processing, in particular to a concrete member surface defect detection method and system based on image segmentation, and the method comprises the steps: obtaining a surface image of a concrete member, and dividing the surface image into a plurality of image blocks; and performing frequency domain transformation on any image block to obtain a power spectrum. According to the method, the feature space period of each image block is analyzed and calculated through frequency domain transformation, and adaptive weighted fusion is carried out on texture features at different distances by using Gaussian weight on the basis of the feature space period. According to the method, the feature extraction process can dynamically adapt to the physical scale of image local textures, namely, small distance analysis is automatically emphasized on fine textures and large distance analysis is automatically emphasized on rough defects, so that scale-perceived composite texture features are constructed; and the accuracy of identifying the concrete surface defects under the complex texture background is obviously improved.

The invention discloses a pavement crack identification method and system based on YOLOv8-Seg, and belongs to the technical field of road engineering pavement maintenance. The method comprises the following steps: identifying a pavement crack slice image by using a trained YOLOv8-Seg segmentation model, and generating a mask area of a crack; extracting attribute information of the crack, wherein the attribute information comprises a center point coordinate, a bounding box, a crack area and a mask binary image; performing space-time continuity analysis on a plurality of images continuously acquired in the same road range, and judging whether cracks with similar forms and similar positions exist or not; and if cracks with similar forms and similar positions exist in the images of the at least two different point positions, determining that the cracks are real cracks, and outputting an identification result. According to the method, crack pixel-level segmentation and attribute extraction are realized, cross-image space-time continuity analysis is combined, crack authenticity is judged, the false detection rate is effectively reduced, and the identification accuracy and engineering applicability are improved.

The invention provides an image processing method, a training method of an image evaluation model and electronic equipment, and relates to the field of computer technology and artificial intelligence. The method comprises the steps that image data and inquiry text data are acquired, the image data are generated by an image generation model, and the inquiry text data are used for describing the requirement for quality evaluation of the image data by adopting a natural language; respectively inputting the image data and the inquiry text data into a plurality of image evaluation models, and respectively carrying out quality evaluation on the image data from corresponding evaluation dimensions by utilizing the plurality of image evaluation models to obtain a plurality of evaluation results; and integrating the plurality of evaluation results to obtain a quality evaluation result of the image data. According to the method and the device, the technical problems of relatively low accuracy and meticulousness of image quality evaluation in related technologies are solved.

The invention discloses a color restoration method and device for a multi-color-temperature mixed light source scene, equipment and a medium, and the method comprises the steps: dividing a to-be-restored image into a plurality of image blocks, and calculating the white balance gain offset value of each image block; all the white balance gain offset values are corrected based on the scene change coefficient, and each target white balance gain offset value is obtained based on the proportionality coefficient and the corrected white balance gain offset values; calculating target CCM saturation based on the proportionality coefficient and the color temperature value of the white balance block in the image block, and determining a target CCM based on the target CCM saturation and the calibration CCM; correcting the corresponding image block based on each target white balance gain offset value; and performing color restoration on the corrected image block by using the target CCM. According to the method, the local white balance gain and the dynamic CCM are used for performing color rendition on the image blocks, color temperature areas do not need to be distinguished, meanwhile, the real-time performance and the effect of color rendition are guaranteed, and the user experience is improved.

The invention provides a code rate control method, and the method comprises the steps: updating the deviation information between an accumulated data volume in a preset time period corresponding to a first image frame and a target data volume according to the data volume of the first image frame currently encoded by an encoder; wherein the accumulated data volume is determined according to the total data volume of the plurality of image frames coded in the preset time period, and the target data volume is determined according to the target code rates of the plurality of image frames; under the condition that it is determined that the first image frame belongs to a waiting frame, if the deviation information is larger than a forced adjustment threshold value, a code rate adjustment instruction is triggered, and the code rate adjustment instruction is used for instructing the encoder to adjust the encoding code rate of the to-be-encoded image frame; wherein the forced adjustment threshold value is greater than a conventional adjustment threshold value, and the conventional adjustment threshold value is used for determining whether to trigger a code rate adjustment instruction or not under the condition that the current image frame does not belong to a waiting frame. According to the mode, dynamic balance of code rate control can be realized through a double-threshold triggering mechanism.

The invention provides an ophthalmic disease prediction model training method and electronic equipment, and the method comprises the steps: obtaining ophthalmic image data and label data corresponding to the ophthalmic image data; preprocessing the ophthalmology image data to obtain a plurality of image blocks with different scale levels; learning the space offset of each sampling point on each image block through the convolution operation of the stacked adaptive gradient modulation convolution layer, adjusting the original convolution kernel weight of the stacked adaptive gradient modulation convolution layer based on the space offset, and transforming the plurality of image blocks with different scale levels to obtain a morphological enhancement feature map; inputting the morphological enhancement feature map into a mixed attention feature extraction module to obtain a small focus enhancement feature map, and performing dimensionality reduction to obtain a feature vector; and inputting the feature vector and the tag data into a full-connection classifier, outputting a disease category prediction probability, and when a training stop condition is met, completing training of the multi-branch convolutional neural network, forming an ophthalmic disease prediction model, and improving the recognition accuracy of minimal lesions.

The invention relates to the technical field of image processing, and discloses a pathological image classification method and system based on graph neural feature fusion, a terminal and a storage medium, and the method comprises the steps: obtaining multi-scale digital image data, carrying out the extraction of the multi-scale digital image data, obtaining a plurality of image features, fusing the plurality of image features through a pyramid-shaped multi-scale feature fusion model to obtain a multi-resolution feature; obtaining a slide level label, constructing an initial graph neural network model based on a dynamic graph construction mechanism, a self-attention mechanism, a channel reduction mechanism and multi-resolution features, and training the graph neural network model according to the slide level label to obtain a target graph neural network model; and obtaining to-be-detected pathological image features, inputting the to-be-detected pathological image features into the target image neural network model for classification, and outputting a classification result. According to the method, features of the to-be-detected pathological image are classified according to the target image neural network model, and efficient and accurate classification is realized.

The embodiment of the invention discloses a picture drawing method and device, equipment and a storage medium. The picture drawing method comprises the steps that multiple original pictures uploaded by a user are acquired; the original picture is converted into a to-be-displayed picture to be added in a canvas, the size of the to-be-displayed picture is equal to the to-be-displayed size of the original picture in a viewport, and the viewport comprises an area, displayed in a display screen, in the canvas; a texture image set is created, the texture image set is used for storing the texture of each to-be-displayed picture displayed in the viewport, and the size of the texture image set is determined according to the size of each to-be-displayed picture in the viewport and the position coordinates of the to-be-displayed pictures; drawing the to-be-displayed picture into the texture image set; and drawing the texture atlas into a canvas area in the viewport. According to the technical means, the technical problem that in the related technology, when a user uploads a plurality of pictures in a collaborative drawing board, the collaborative drawing board renders each picture, generated system memory resource occupation is too high is solved.

The invention provides an image-based bolt performance detection method and device, and belongs to the field of image classification and recognized.The method comprises the steps that a first image set is acquired, and the first image set is preprocessed to obtain a second image set; wherein the first image set comprises a plurality of images of different angles of the target bolt; the second image set is input into a trained MobileNetV3 classification network, a third image set is obtained, and the third image set is the second image set which is obtained through classification and is provided with category labels; and inputting the third image set into a defect identification network to obtain a detection result of the target bolt. According to the image-based bolt performance detection method and device provided by the invention, background interference can be eliminated through mask preprocessing, targeted defect detection is realized in combination with bolt category information, the recognition precision is remarkably improved, the false detection rate is reduced, meanwhile, the bolt performance is accurately evaluated by adopting quantitative calculation, and the efficient and accurate requirements of industrial quality inspection are met.

The invention discloses an AI-based logistics document intelligent identification and filling system, and the system comprises an image collection module which is used for obtaining the image information of a logistics document, supporting a plurality of image input modes, such as scanner, mobile phone photographing, camera capturing and the like, carrying out the preprocessing of a collected image, including the operation of image enhancement, denoising, graying, binaryzation and the like, and obtaining the image information of the logistics document; the image quality is improved, and preparation is made for follow-up recognition; the document type identification module is used for carrying out document type judgment on the preprocessed document image based on a convolutional neural network (CNN) algorithm in deep learning, and can automatically identify common logistics document types; the invention relates to the technical field of logistics information, and the AI-based logistics document intelligent identification and filling system can efficiently and accurately identify various logistics documents, realize intelligent filling and improve the automation level and accuracy of logistics document processing.