291 results about "Picture recognition" patented technology

The invention provides an adaptive welding seam detection and three-dimensional reconstruction method based on deep learning and binocular vision. The adaptive welding seam detection and three-dimensional reconstruction method comprises the steps of S1, collecting samples and making a training data set; s2, the picture of the sample to be welded is processed, a feature region is recognized, the image quality of the region to be welded is analyzed and evaluated through wavelet transform and local variance, and the noise level and the contrast ratio are calculated; s3, dynamically generating edge detection parameters and model fitting parameters according to the image quality; s4, using an edge detection algorithm to extract edge point cloud of the welding seam area; s5, performing RANSAC linear fitting, weighted least square fitting and polynomial curve fitting on the edge point cloud in parallel; s6, selecting an optimal fitting result based on an image quality adaptive dynamic scoring model; and S7, carrying out three-dimensional coordinate conversion in combination with the three-dimensional matching model IGEV-Stereo, and outputting a final welding seam three-dimensional coordinate. According to the invention, automatic detection of the position and size of the welding seam can be efficiently and accurately realized.

The invention relates to the technical field of image recognition, in particular to a lightning arrester fault thermal imaging picture recognition method and system combined with gaze prediction, and the method comprises the steps: carrying out the down-sampling of an original thermal imaging image to a fixed size, sequentially passing through a multi-layer convolution and a Spatial Softmax layer, and outputting a predicted gaze point track sequence, extracting a key area and a non-key area according to the intensity of the fixation point track sequence; reconstructing the key area to obtain a compressed and recombined target area; inputting the non-key region into a variational auto-encoder to obtain a low-dimensional potential feature vector; and inputting the image-structure joint feature representation into an EffiCroprViT model, and finally obtaining a lightning arrester fault classification result. Local and global features are efficiently fused, the calculation complexity is reduced, background noise interference is effectively suppressed, real-time and accurate identification and early warning of the fault state of power equipment are realized, and the fault classification method has the advantages of high efficiency, high reliability and high reliability. Therefore, the safety and stability of power grid operation are ensured.

The invention discloses a method for improving the model picture recognition capability based on a multi-modal large model, and relates to the technical field of image recognition, and the method comprises the steps: building an infrastructure of the multi-modal large model, and training the multi-modal large model as a basic model through a large number of pictures; newly added images are collected, and corresponding standard questions and answers describing picture contents are prepared in combination with the images; and according to a picture recognition task requirement, designing a corresponding instruction, namely a cue word, and guiding the large model to carry out picture recognition. The multi-modal large model is trained by using massive pictures, the accuracy of a picture recognition technology can be optimized by fully utilizing a picture feature library of the multi-modal large model when the multi-modal large model is applied to the field of picture recognition, and the magnitude order of training pictures can be greatly reduced by using fine tuning training of the multi-modal large model, so that the accuracy of the picture recognition technology is improved. After composite fine tuning training is carried out, the multi-modal large model can have the recognition capability of generalizing trained pictures, and the number of pictures needing to be used for training can be greatly reduced.

The invention provides an interface element positioning method based on picture recognition. According to the method, interface elements are recognized through an existing automatic user interface operation technology and numbered, and then screen pictures containing numbers are intercepted and transmitted to a large language model. And the big language model recognizes an operation target by utilizing the graph recognition capability and the natural language understanding capability of the big language model, obtains coordinate information of operation elements through UI automation attributes, and finally executes corresponding operation. According to the method, the recognition and positioning accuracy of the target operation element in the RPA system based on natural language interaction can be remarkably improved, and the error rate is reduced.

The invention relates to the technical field of picture recognition, and discloses a picture detection method and device, equipment and a medium, and the method comprises the steps: carrying out the portrait direction confirmation of a certificate picture, obtaining a certificate picture in a straightening state, obtaining a highlight region of the certificate picture in the straightening state, judging whether the highlight region is a light reflection region, and if not, carrying out the recognition of the certificate picture; if yes, sending the certificate picture in the straightening state to a server, otherwise, acquiring a non-reflective area picture of the certificate picture in the straightening state, identifying character area incomplete features and portrait area incomplete features in the non-reflective area picture, and identifying a complete character area according to the character area incomplete features by utilizing a character area boundary identification model; identifying a complete portrait area according to the incomplete features of the portrait area by utilizing a portrait area boundary identification model, segmenting a key information area of the certificate picture in the straightening state according to the complete character area and the complete portrait area to obtain a key information picture, and sending the key information picture to a server, and the accuracy and efficiency of picture detection are improved.

The invention provides a crop disease and insect pest judgment method and system based on image feature recognition, and relates to the technical field of picture recognition processing, and the method comprises the steps: carrying out the disease and insect pest recognition of a to-be-judged crop image through a pre-trained disease and insect pest analysis model, and obtaining a first disease recognition result; presetting similar pest and disease combination and corresponding combined spot features, judging whether the similar pest and disease combination exists in the first pest and disease recognition result, if not, outputting the first pest and disease recognition result, and if the first pest and disease combination exists, extracting a spot area of the to-be-judged crop image; traversing spot areas of the to-be-judged crop image, and if joint spot features appear in the spot areas, taking two diseases and pests in the similar disease and pest combination as output results; if the combined spot features do not exist, the pest and disease damage with the minimum confidence degree in the similar pest and disease damage combinations is removed; and updating the first pest identification result.

The invention discloses a reservoir dam safety monitoring method based on multi-source information, and the method comprises the steps: carrying out the high slope inspection through employing an unmanned aerial vehicle A and an unmanned aerial vehicle B, carrying out the high-definition camera and an infrared thermal imager, calculating the pitching angle between the unmanned aerial vehicles through GPS synchronization, image recognition and other technologies, and correcting the fusion image recognition potential safety hazards; path planning is carried out by adopting a cattle farming method, and the problems of unmanned aerial vehicle inspection path independence and task coverage are solved; the pixel value of the unmanned aerial vehicle B in the image A is compared with the normal horizontal state image, the pitching angle lambda is calculated and used for correcting the image A and the image B, and the potential hidden danger of the reservoir dam is fused and recognized; the method is low in cost and easy to maintain, a comprehensive inspection result is combined with a picture recognition deep learning algorithm by combining a manual inspection verification result, long-term tracking of a potential safety hazard part is established, and powerful support is provided for reservoir dam monitoring.

The invention discloses a code generation method and device based on a large model, a storage medium and computer equipment. The method comprises the steps that a UI prototype draft of a target service is acquired; performing element identification on the UI prototype draft through a picture identification large model, and determining element information in the UI prototype draft; generating a code generation cue word based on the element information and a preset cue word, inputting the code generation cue word into a code generation large model to obtain an intermediate code corresponding to the UI prototype draft, and arranging a target service code of the target service based on the intermediate code. The automatic process from UI design to code generation is achieved, the tedious process of manually writing UI components and debugging interaction logic in traditional development is avoided, repeated work and human errors are reduced, developers can put more energy into core service function development, then development efficiency is improved, code quality is guaranteed, and the development cost is reduced. And rapid and high-quality delivery of the project is promoted.

The invention provides a gastroscope image recognition and auxiliary diagnosis system and method based on deep learning and large language model fusion, and the system comprises a data collection and processing module which is used for obtaining and preprocessing various types of gastroscope images; the gastroscope picture recognition and classification module is used for inputting the preprocessed gastroscope pictures into a GastroHAF model for training and outputting recognition and classification results of the gastroscope pictures, and the GastroHAF model comprises a weak supervision anatomical mark positioning sub-module, a space co-occurrence context reasoning sub-module and an anatomical atlas guided hierarchical classification sub-module; and the auxiliary diagnosis module is used for outputting stomach diagnosis results and examination suggestions based on the fine-tuned Chinese medical big language model. The GastroHAF model provided by the invention can quickly, effectively and accurately judge the position of the gastroscope in combination with prior anatomical knowledge, and can effectively assist doctors in stomach examination in combination with a large language model.

The invention discloses a quantitative decomposition logic storage architecture for simplifying picture recognition, which relates to the technical field of computers, and comprises the following steps: step 1, collecting picture and video data; step 2, feature value extraction; 3, dividing by using a letter sequence in combination with the data; step 4, establishing an alphabet sorting form quantitative decomposition logic frame storage structure; step 5, matching the to-be-identified picture and video data; step 6, identifying and judging; according to the method, an independent logic storage space and independent identity letter sorting information are generated by constructing a letter sorting form quantitative decomposition logic storage architecture, the operation process is simplified, the address information transmission speed is increased, data processing can be efficiently operated under common hardware configuration, the recognition precision is greatly improved, and the recognition efficiency is improved. And the letter sorting form quantitative decomposition logic storage architecture has an encryption effect, so that the privacy leakage risk in the data distributed storage identification process can be effectively reduced, the data security is improved, and the management is more convenient.

The invention discloses an aviation intelligent test platform, and relates to the technical field of aviation test. According to the method, a locally deployed domestic large model and a knowledge base are introduced, a large model fine tuning technology is integrated to realize automatic generation from software requirements to test cases, then on the basis of graph or video identification of a multi-modal large model, characteristic graphs of avionics display control are identified through pictures, a test result is obtained, and the test efficiency is improved. Therefore, automatic execution of the whole testing process is achieved, many problems caused by manual testing are avoided, high efficiency is achieved by reducing labor input, the accuracy of a testing result is improved through an image characteristic recognition technology, and testing requirements of different industry fields can be met by building an exclusive knowledge base and a directional fine tuning model.

The invention provides a zero-code report management method and system. The method comprises the following steps: 1) identifying a shot paper table picture to obtain a table title and table items; processing the spreadsheet to obtain a sheet title and a sheet item; 2) processing the obtained table titles and table items to obtain database names and field names, and establishing a database table and in-table fields; outputting the processed database name and field name; 3) dynamically generating report management data according to the input database name and field name so as to realize the functions including adding, deleting, modifying and querying of the report data; 4, the needed UI template is selected according to the input database name and the input field name, and a report management interface and an editing interface are dynamically generated according to the input item and the selected UI template style.The requirement for zero-code report management is met by photographing and recognizing a traditional paper report, importing an electronic report and generating the needed report through one key.

The present invention provides a crop pest and disease identification method and system based on image feature recognition, which relates to the technical field of image recognition and processing. The present invention uses a pre-trained pest and disease analysis model to identify pests and diseases in a crop image to be identified, and obtains a first pest and disease identification result; presets similar pest and disease combinations and corresponding joint spot features, and determines whether there is a similar pest and disease combination in the first pest and disease identification result; if there is no similar pest and disease combination, outputs the first pest and disease identification result; if there is the first pest and disease combination, extracts a spot area of ​​the crop image to be identified; traverses the spot area of ​​the crop image to be identified, and if a joint spot feature appears in the spot area, both pests and diseases in the similar pest and disease combination are used as output results; if there is no joint spot feature, eliminates the pest and disease with the lowest confidence in the similar pest and disease combination; and updates the first pest and disease identification result.

The invention relates to the technical field of picture recognition methods, in particular to an abnormal picture automatic recognition method, which comprises the steps of reading monitoring video files from object-oriented storage in batches, and performing distributed frame extraction on the video files; stratified sampling is carried out on the picture after frame extraction, and a training set which is all positive samples is formed to train a PatchCore model; determining an abnormal score threshold, and reasoning all the pictures subjected to frame extraction to obtain an abnormal score and an abnormal thermodynamic diagram of each picture; judging whether each picture is an abnormal picture, and inputting the abnormal picture and the abnormal thermodynamic diagram into a multi-modal large model; the abnormal thermodynamic diagram is divided into nine areas according to nine grids, abnormal areas in the nine areas are input into a multi-modal large model, abnormal description of an abnormal picture and a specific coordinate frame of the abnormal areas are recognized, and result information is stored in object-oriented storage; according to the method, the recognition accuracy of the abnormal data can be improved under the condition of lack of abnormal samples.

The invention discloses a GIS-BIM-based in-service road disease section marking cooperation platform and method, and the method comprises the steps: constructing an EBS structure tree model, importing the EBS structure tree model into a GIS system to form a GIS-BIM platform, and transmitting a to-be-inspected line and a stake number coordinate table to an unmanned plane; the unmanned aerial vehicle inspects and shoots a long picture and matches a stake number label, and the server cuts the picture and identifies disease features; and the GIS-BIM platform maps the disease pictures to the corresponding BIM model segments, and deletes the maps if no disease pictures are inspected again. The stake mark coordinate table is generated according to stake mark progressive increase, disease recognition adopts a convolutional neural network model, and color gradient distinguishing is performed according to severity. The platform integrates EBS and WBS dual-core structure trees, realizes full-period collaboration through stake mark interval dynamic mapping, improves the disease marking efficiency and precision, and realizes full-process digital operation and full-period automatic collaboration.

The application relates to an intelligent clipping application method through picture recognition, which comprises the following steps: acquiring medical image video data to be clipped; performing grouped image recognition on the medical image video data to be clipped to obtain a key frame index; and performing clipping on the medical image video data to be clipped based on the key frame index. The application does not require a clipping personnel with medical knowledge to perform clipping, and the clipping personnel can complete the video clipping operation without watching the whole content of the video or repeatedly watching the video for multiple times, so that the efficiency of the video clipping is improved. The application also relates to an intelligent clipping application device through picture recognition, a storage medium and equipment.

The invention provides a document processing method, a document question and answer method and a computer program product, and relates to the technical field of document question and answer. The document processing method comprises the steps of performing picture recognition and text processing on a to-be-processed document in a target format to obtain a plurality of text segments; wherein the to-be-processed document comprises a table and / or a picture; based on the plurality of text segments, generating multi-level abstracts and multi-level recommendation questions corresponding to the multi-level abstracts; and performing associative storage on the plurality of text segments, the multiple levels of abstracts and the multiple levels of recommendation questions to obtain a document database. The document question-answering method comprises the steps of obtaining an input question of a user; based on the input question, performing retrieval in a document database, and determining a recalled text segment and a recalled recommendation question; and generating an answer result in combination with the recalled text segment and the recalled recommendation question. According to the method and the device, characters, pictures, tables and the like existing in the document can be processed, and question and answer recall processing is carried out based on the extracted multi-level abstract and the generated recommendation question.

The invention relates to the related technical field of inspection, in particular to a photovoltaic intelligent unmanned aerial vehicle inspection system. The system comprises an SaaS platform for providing annotation platform management, merchant platform management, platform setting function and content management. The AI training system comprises a training system and an identification system. The training system is used for model training and continuously optimizing model performance through annotation data; the recognition system is used for recognizing defects, such as spot spots, multiple spots, battens, missing, no-load and shielding, in photovoltaic pictures shot by the unmanned aerial vehicle. And the labeling platform provides project management, task management, employee management, role management and picture labeling functions and is used for data labeling and preprocessing work. And the merchant platform performs photovoltaic project management, surveying and mapping management, map management, unmanned aerial vehicle management, task management, an infrared picture AI identification function and a photovoltaic defect report generation function.

The invention discloses an image recognition system for marking the appearance of a mobile phone. The system obtains an area of ​​interest marked on the appearance of a mobile phone by a user, wherein the area of ​​interest is a subset of an image to be detected, and is a partial area extracted from the image, and the partial area is composed of a plurality of irregular shapes. The image to be detected is segmented into two or more non-overlapping areas according to a similarity criterion of preset grayscale, color, texture and shape features. The areas segmented from the image are used as target objects for subsequent feature extraction according to connected components, and the two-way components are pixels connected together according to preset rules in a binary image. The mapping relationship of data from input to output is obtained, and the best mapping function for reflecting the target task is found through training. High-dimensional abstract features are used to represent low-dimensional target features. A network training module includes a convolution layer and a pooling layer. Different data amplifications are used to increase the data volume, so that data samples are richer and the precision and accuracy of defective part extraction are improved.

The invention provides a picture recognition method which can be applied to the technical field of big data and the technical field of artificial intelligence. The image recognition method comprises the steps that based on a first image recognition model, multiple visual features of a target image are extracted, and the target image is an image of a business bill; based on a second image recognition model, extracting a plurality of layout features used for representing character layout information of the business bill in the target picture; performing feature fusion processing on the plurality of visual features and the plurality of layout features by using a pre-constructed feature fusion network to obtain a target fusion feature; and performing specific information extraction processing on the target fusion feature by using the region generation network, and outputting a picture recognition result. The invention further provides a picture recognition device and equipment, a storage medium and a program product.

The application discloses a picture recognition model training and picture recognition method, system, device and medium. The method comprises the following steps: obtaining a target picture, performing feature extraction on the target picture based on a feature extraction network to obtain target features of the target picture; judging a picture type of the target picture based on a binary classification network and the target features; and determining a processing strategy for the target features according to the picture type obtained by the judgment. That is, after the feature extraction on the target picture, the target features obtained by the extraction are classified, different processing strategies are adopted for different types of picture types, the flexibility of the face recognition model training is improved, and the performance of the face recognition is improved.

The invention relates to the technical field of mobile terminal plant target detection, in particular to a mobile terminal plant target detection and recognition method in an off-line state, which comprises the following steps: turning on mobile terminal equipment, and selecting a proper plant and a proper shooting angle; acquiring a multi-angle to-be-detected plant picture by using a mobile terminal; based on the internal model of the mobile terminal equipment, using a YOLO model to train a regional plant to obtain a specific model, and extracting image plant feature data; comparing the target features with the model data features to obtain plant type probability distribution and plant position information; comparing the identification results of the plurality of pictures; marking the type and the position of the plant in the picture, and returning the identification result; through the trained picture recognition model, the plant picture is recognized and analyzed, the plant variety and related information are judged, the related information is further analyzed, simple, accurate and intelligent plant recognition is achieved, and great convenience is brought to field network-free scientific investigation.

The present disclosure relates to the technical field of face recognition, and provides a face recognition method and device in the presence of occlusion. The method comprises: acquiring a base data set, wherein the base data set contains multiple base pictures; generating an occlusion for each base picture to obtain an occluded picture corresponding to each base picture; when a face picture to be recognized is detected, extracting a first feature of the face picture, a second feature of each base picture, and a third feature of the occluded picture corresponding to each base picture by using a face recognition model; calculating the similarity of the first feature and the second feature of each base picture and the third feature of the occluded picture corresponding to the base picture respectively; determining a target base picture with the highest similarity to the face picture according to the similarity of the first feature and the second feature of each base picture and the third feature of the occluded picture corresponding to the base picture, and recognizing the face picture as a target object corresponding to the target base picture.

The invention relates to a distributed energy storage site surveying and evaluating method and system based on multi-model collaboration. The method comprises the steps that picture description and a structure questionnaire form are collected through a mobile terminal; matching according to the picture description and the structure questionnaire form to generate a first cue word; calling a large model, identifying setting information and electrical characteristics in the picture description, and obtaining a picture identification result and a text analysis result; constructing a feature request by combining the form data to generate evaluation model parameters; the evaluation model parameters, the existing energy storage equipment configuration and the load prediction data are fused, and the station exploration evaluation result of the machine room is output through the evaluation model; matching and generating a second cue word according to the station exploration evaluation result; calling the large model based on the matched second cue word, and generating an evaluation report according to the station exploration evaluation result; by applying the mode and the method of the invention, the processing efficiency is greatly improved and the data accuracy is improved for the design of a distributed energy storage site survey evaluation scene with a large site number scale.

The present invention discloses a picture recognition method, a recognition model training method, a device and an electronic device. Among them, the method includes: obtaining a target 3D picture to be recognized; inputting the target 3D picture to be recognized into a first recognition model, where the first recognition model is used to recognize the target 3D picture to be recognized to obtain the picture type of the target 3D picture to be recognized, and the convolutional blocks of the first recognition model are the same as those of the second recognition model, and the second recognition model is a model obtained by training an original recognition model using target training samples, and the target training samples include cubes obtained by rotating and sorting N target cubes obtained from 3D sample pictures, and N is a natural number greater than 1; obtaining the first type of the target 3D picture to be recognized output by the first recognition model. The present invention solves the technical problem of low model training efficiency in the related art.

The present invention relates to an NFC tag-based application wake-up method and an NFC acceptance terminal. The NFC acceptance terminal comprises: an NFC tag module, used to provide an NFC tag supporting read and write forms; a camera, used to acquire a pose image of a mobile terminal approaching the NFC acceptance terminal; and a pose recognition module, used to recognize, on the basis of the pose image, an approaching pose of the mobile terminal approaching the NFC acceptance terminal, look up a correspondence table on the basis of the approaching pose to obtain application configuration information of an application corresponding to the approaching pose, and write the application configuration information to the NFC tag, the correspondence table storing one-to-one correspondences between a plurality of approaching poses of the mobile terminal approaching the NFC acceptance terminal and application configuration information of a plurality of applications. According to the present invention, a plurality of different applications can be supported on one NFC acceptance terminal without modifying the mobile terminal.

The present invention discloses a self-supervised pre-training method based on a region screening module and multi-level comparison. Since manually annotating dataset images is both expensive and time-consuming, using a self-supervised method without labels to pre-train deep learning models for image classification and object detection has indispensable research significance in the field of computer vision. In order to directly perform effective self-supervised pre-training in this type of dataset, the present invention proposes to pre-train the model based on an instance region screening module, whose function is to screen out a specific number of patch graphs that are most likely to contain instance information from the two enhanced graphs generated from the images for contrastive learning and match them. The present invention proposes a module for multi-level contrastive learning of global, local, and global-local comprehensive information. The model pre-trained with this method using a multi-instance dataset has achieved a good improvement in accuracy in downstream tasks such as image recognition and object detection.

The present application provides a method, apparatus, device, and readable storage medium for training an identification model. The method includes obtaining a preset proportion of special sample pictures from a preset number of basic sample pictures and a preset number of special sample pictures corresponding to documents of a preset type, where the basic sample pictures are intercepted pictures of corpus fields in a basic corpus, and the special sample pictures contain the same type of fields in the documents of the preset type; training a basic model based on the preset number of basic sample pictures and the preset proportion of special sample pictures among the preset number of special sample pictures to obtain an identification model for document picture identification. On the premise of meeting the model accuracy rate, this method can achieve the effect of saving resources and training time for training the model.