126 results about "Database graph" patented technology

A method (and system) for specifying the region layout of objects in an affine invariant manner as a set of affine intervals between pairs of regions, includes representing database and query regions using affine intervals along with their region identity, matching query region layout to layout of database image regions using an index structure, and retrieving relevant images of the database by hashing for dominant hit regions in the index structure.

Each image of a set of images is characterized with a set of sparse feature descriptors and a set of dense feature descriptors. In some embodiments, both the set of sparse feature descriptors and the set of dense feature descriptors are calculated based on a fixed rotation for computing texture descriptors, while color descriptors are rotation invariant. In some embodiments, the descriptors of both sparse and dense features are then quantized into visual words. Each database image is represented by a feature index including the visual words computed from both sparse and dense features. A query image is characterized with the visual words computed from both sparse and dense features of the query image. A rotated local Bag-of-Features (BoF) operation is performed upon a set of rotated query images against the set of database images. Each of the set of images is ranked based on the rotated local Bag-of-Features operation.

Systems, methods and articles solve queries or database problems through the use of graphs. An association graph may be formed based on a query graph and a database graph. The association graph may be solved for a clique, providing the results to a query or problem and / or an indication of a level of responsiveness of the results. Thus, unlimited relaxation of constraint may be achieved. Analog processors such as quantum processors may be used to solve for the clique.

The present invention is directed to providing a higher degree of association between nodes and links in a graph by creating data structures (spiders) that provide views into graphs that transcend the relatively static association of a conventional graph. A spider's variables bind to any number of nodes and links in the graph, enabling all of the bound nodes and links by addressing the spider. By adding constraints on the extent or degree of binding in a spider to a graph, a subset of the graph is identified. The spider can then used to address the subset of the graph as constrained by the spider. A spider can bind to a link in order to identify a parent / child structural subset of the graph. More specifically a spider is a collection of variables that create a template or pattern and bind to the nodes and links in the graph. A spider traverses a graph by binding its variables to various nodes and links in the graph.

The embodiment of the invention provides an indoor positioning method and device based on SLAM. The method comprises the steps of obtaining a to-be-positioned image uploaded by a user terminal, constructing a picture database with pose information rapidly through an SFM algorithm; extracting a feature vector of a to-be-positioned image and a feature vector set of a database image through a convolutional neural network; calculating the similarity between the feature vectors of the to-be-positioned image and the database image, completing loopback detection, and obtaining the image with the highest similarity with the to-be-positioned image in the image database and the first pose information corresponding to the image, thereby obtaining the accurate position and pose information of the userterminal. Compared with a traditional visual SLAM algorithm which adopts a word bag model and is weak in recognition capability, the deep features of the image are learned through the neural network,the higher recognition accuracy can be achieved, and the accuracy of loop detection is improved.

The invention provides a compact Hash code learning method based on semantic protection, and the method comprises the steps: dividing a data set, and obtaining a test sample set, a training sample set, and an image library; then constructing a deep Hash network model; adding a hidden layer (Hash layer) to the last full connection layer of a common convolutional neural network model, the number ofneurons of the hidden layer is the length of Hash codes, an activation function is a panh function, designing a constraint function, protecting the semantic similarity of images, and meanwhile it is guaranteed that the learned Hash codes are evenly distributed and the quantization error is small; extracting Hash codes of the query image and the database image through the trained model, and calculating Hamming distances between the Hash codes of the image and the Hash codes of all the images in the database; and finally, sorting the Hash codes in the database according to a sequence of distances from small to large, and sequentially outputting the original images corresponding to the Hash codes to obtain a similar image retrieval result. According to the method, retrieval of large-scale images is more accurate and effective.

A method and system of optimizing a complex manufacturing process performed to achieve one or more processing objectives for the process and / or a component produced by the process. The system includes a graphical user interface, a process module, and an optimization module. The process module includes a training module, an empirical relationships database, an analytical equations database, a heuristic knowledge database, and a process models database. The graphical user interface is used to input one or more processing variables and constraints for the processing objective. The training module generates empirical relationships from the processing variable and empirical data obtained from the manufacturing process. The process module generates a process model that takes into consideration heuristic knowledge of the manufacturing process, empirical relationships, and optionally analytical equations relating to the manufacturing process. The optimization module employs the process model to optimize the manufacturing process.

A system is presented with a search panel for specifying search criteria for searching a database of biological information. The search panel uses the extensible markup language (XML) to send search requests to the database. A database graph generation module linked to the biological database generates a database graph representing the database schema. Once the database schema is known, another module creates joins between the database tables in order to most effectively join data from one table to another. An SQL statement incorporating the optimized joins is then used to search the biological database.

The invention discloses a visual loopback detection method based on semantic segmentation and image restoration in a dynamic scene. The visual loopback detection method comprises the following steps:1) pre-training an ORB feature offline dictionary in a historical image library; 2) acquiring a current RGB image as a current frame, and segmenting out that the image belongs to a dynamic scene areaby using a DANet semantic segmentation network; 3) carrying out image restoration on the image covered by the mask by utilizing an image restoration network; 4) taking all the historical database images as key frames, and performing loopback detection judgment on the current frame image and all the key frame images one by one; 5) judging whether a loop is formed or not according to the similarityand epipolar geometry of the bag-of-words vectors of the two frames of images; and 6) performing judgement. The visual loopback detection method can be used for loopback detection in visual SLAM in adynamic operation environment, and is used for solving the problems that feature matching errors are caused by existence of dynamic targets such as operators, vehicles and inspection robots in a scene, and loopback cannot be correctly detected due to too few feature points caused by segmentation of a dynamic region.

The invention discloses a multi-channel topic model-based editable garment image search method. The method comprises the steps of firstly, finding a main body region of a garment commodity in a picture by using an object detection method, extracting multiple descriptors from the main body region and quantizing the descriptors into vectors in bag-of-word forms by using a bag-of-word model; secondly, according to search conditions, editing and modifying weights of visual words, fusing the vectors in the bag-of-word forms into retrieval features capable of describing high-level semanteme of the garment commodity by using a pre-trained multi-channel topic model, and establishing indexes; and during online detection, calculating vector similarity of a to-be-queried commodity sample image and animage of a database, and taking the commodity with the highest similarity as a search result. The information of vision, commodity text attributes and the like of the to-be-queried commodity can be re-edited; the demand of a user on the commodity is described more accurately; and the user-expected commodity is searched for through the multi-channel topic model.

A database graphical comparison tool allows comparing database items that include multiple database statements and graphically displaying the comparison results in a ranked list of database statements. The graphical comparison tool includes a graphical user interface that allows a user to easily configure the tool for both manual and automatic (or scheduled) comparisons. In addition, the user may specify one or more actions that may be autonomically performed when the comparison of database items meets predefined criteria. Database items that may be compared using the database graphical comparison tool include optimizer monitors and plan cache snapshots.

The embodiment of the invention discloses an image retrieval method and device, equipment and a medium, and relates to an intelligent search technology. The method comprises the following steps: extracting global features and local features of a to-be-retrieved image by utilizing a preset neural network model, and extracting global features and local features of a to-be-recalled image; determininga candidate image set through global feature matching and local feature matching between the to-be-retrieved image and the to-be-recalled image; and determining a retrieval result from the candidateimage set by performing local feature verification on the to-be-retrieved image and the candidate images in the candidate image set. According to the embodiment of the invention, the accuracy of imageretrieval under the condition of a large number of database images can be improved, and meanwhile, the retrieval efficiency is ensured.

The invention discloses an intelligent monitoring security data processing management system based on big data analysis. The system comprises a region division module, an image acquisition module, a time period division module, a feature extraction module, a storage database, an image preprocessing module, a danger feature database, an analysis cloud platform, a video retrieval module, an early warning module and a display terminal. According to the invention, a large amount of useless video image data is removed through the image preprocessing module, and dangerous behavior type characteristics are extracted by comparing the reserved video image data through the characteristic extraction module and the dangerous characteristic database, so that the problems that the existing video monitoring resource data is massive and redundant, occupies a large amount of storage and is inconvenient to use are solved. The method solves the problem of high monitoring hardware construction capital occupation, improves the comparative analysis operation speed of effective data, is suitable for monitoring of some long-term monitoring areas, reduces the hardware cost, saves a large amount of time, and promotes the normal implementation of security monitoring work of monitoring personnel.

The invention discloses an image retrieval method based on a hierarchical convolutional neural network, and mainly aims at solving the problem that in existing all-sky aurora image retrieval, the accurate rate is low. The method comprises the implementation steps that 1, local key points of all-sky aurora images are determined by adopting an adaptive polar barrier method; 2, local SIFT features ofthe all-sky aurora images are extracted, and a visual vocabulary is constructed; 3, the convolutional neural network is pre-trained and subjected to fine tuning, and a polar region pooling layer is constructed; 4, region CNN features and global CNN features of the all-sky aurora images are extracted; 5, all the features are subjected to binarization processing, and hierarchical features are constructed; 6, a reverse index table is constructed, and the global CNN features are saved separately; and 7, hierarchical features of a queried image are extracted, the similarity between the queried image and the database images is calculated, and a retrieval result is output. According to the method, matching of the local key points is achieved through the hierarchical features, the problem that inan existing image retrieval method, the false alarm rate is high is solved, the advantage of being high in retrieval accuracy rate is achieved, and the method is suitable for real-time image retrieval.

The invention provides a human face beauty prediction method based on multi-task learning. The method comprises the steps of constructing a multi-task human face database and constructing a multi-taskhuman face beauty prediction model. According to the invention, the accuracy of face beauty prediction is enhanced by adding expression recognition and age recognition. In the multi-task face database construction process, the constructed database image comprises three labels including a face expression attribute, an age attribute and a face beauty degree attribute, so that subsequent multi-tasktraining and prediction are facilitated. Network parameters are shared among tasks in a multi-task training process. Shared features are learned, so that the accuracy of learning a single task by a network is improved. The multi-task learning is carried out by using a deep learning network. A shared representation layer can enable tasks with generality to be better combined with correlation information. A task specific layer can independently model task specific information. Network parameters can be optimized by using samples from different tasks. Meanwhile, the multi-task performance is improved.

A method, system and computer program product for efficiently identifying images, videos, audio files or documents relevant to a user using binary search trees in attribute space for guiding relevance feedback. A binary tree is constructed for each relative attribute of interest. A “pivot exemplar” (at a node of the binary tree) is set for each relative attribute's binary tree as corresponding to the database image, video, audio file or document with a median relative attribute value among that subtree's child examples. A pivot exemplar out of the available current pivot exemplars that has the highest expected information gain is selected to be provided to the user. Comparative attribute feedback is then received from the user regarding whether a degree of the attribute in the user's target image, video, audio file or document is more, less or equal with the attribute displayed in the selected pivot exemplar.

The invention discloses an unsupervised pedestrian re-identification method based on three-data-set cross transfer learning, which comprises the following steps of: training three CNNs on a big data set for image classification to obtain three pre-training models; finely adjusting the three labeled source pedestrian data sets A, B and C respectively; Utilizing the three CNNNs to respectively extract features of label-free pedestrian images in a target data set, and using K-to obtain K-values; respectively clustering the extracted features by using a neighbor clustering algorithm; screening outpicture samples which are close to a clustering center domain after clustering of the three models, and labelling the picture samples; adding the three labeled sample data into another source pedestrian data set in a crossed and alternate manner, and then finely adjusting the model; inputting a pedestrian test picture into the three trained models to obtain three feature matrixes, and performingmaximum pooling operation to obtain a unique feature of the test picture; and the Euclidean distance between the unique feature and the picture feature in the database is calculated, and the identityof the database picture with the minimum distance is the identity of the test picture.

The invention relates to the area of computer vision video data analysis, in particular to the technologies aimed to search the required objects or events in the analyzed video originally received from a third-party device. An imported video analysis device consists of memory, database for metadata storage, a graphical user interface, and a data processing module. The data processing module is configured to upload a video in any available format into the memory and to import the uploaded video into software of the imported video analysis device. Software decompresses and analyzes the imported video to generate metadata characterizing the data in all objects in the video and to save the metadata in database. The search speed for the required event or object in the imported video received from a third-party device is increased.

The invention discloses an image retrieval method based on deep hash and quantification and a storage medium. Firstly, a training set and a test set are established, images needing to be recognized are preprocessed, then a convolutional neural network is constructed, an Alexnet model structure is adopted as a basic framework, data pairs are generated at any time through training samples, trainingis conducted according to the convolutional neural network, and a corresponding output value Zn is obtained. By processing an image category through a Glove model to obtain an embedded label V, calculating an error function of an output value of the convolutional neural network by combining the embedded label V, updating network parameters, processing a query image and a database image by using atrained model to obtain corresponding binary codes, calculating inner product similarity by using an asymmetric distance quantification method, a retrieval result is output. According to the method, the block coding module is introduced, and the elaborately designed hybrid network and the specified loss function are utilized to jointly learn the deep-view semantic tags, so that the accuracy of image retrieval is greatly improved.

The invention discloses an exhibition hall and exhibition hall intelligent management and control platform based on personnel dynamic distribution tracking analysis. The platform comprises a region division module, an identity information collection module, a personnel identity authentication module, a storage database, an image collection module, an image processing module, a personnel statisticsmodule, a personnel analysis module, an analysis server, a voice broadcast module, a temperature detection module, a temperature analysis module and an exhibition hall management and control platform. According to the invention, the identity information of each visitor entering the exhibition hall is obtained, the number of visitors in each sub-region in the exhibition hall region is counted, whether the number of visitors in each sub-region reaches the maximum is analyzed, the visitors in each sub-region which does not meet the requirements are screened, and personnel diversion is carried out, so that the visitor flow rate is improved. Meanwhile, the personnel density grade of each sub-region is analyzed, the air ventilation frequency of each sub-region is screened, the optimal temperature of each sub-region is calculated, and adjustment is carried out, so that a large amount of energy consumption is saved.

The invention discloses a three-dimensional visual automatic monitoring system platform taking a BIM technology as a carrier, which belongs to the field of engineering monitoring. The three-dimensional visual automatic monitoring system platform comprises a database, a chart module, a BIM module, an early warning module, an authority management module, a data module, an analysis and prediction module, a large screen module, a monitoring operation and maintenance module, an expert system module and a display interaction module. The system platform has the following outstanding characteristics:a BIM technology is used as a carrier of monitoring data, a monitoring report can be issued, a monitored object can be analyzed and predicted, and an expert system capable of intelligently giving a corresponding solution under emergency and early warning conditions is provided; bottom-layer monitoring data is transmitted to a system platform through a network after being collected, and a monitoring effect is presented at a PC end, a Pad end and a Phone end of the display interaction module. According to the intelligent engineering monitoring system, informatization and intellectualization of engineering monitoring are achieved, intuitive visualization of the monitoring effect is achieved, and engineering construction is protected.

The invention relates to the technical field of image retrieval and computer vision, in particular to a fine-grained image retrieval method based on visual attention mechanism weighting. The method comprises the following steps of: image preprocessing: setting the length of the longest side of an image to be 500 pixels; feature extraction: inputting the image into a convolutional neural network, and then selecting and outputting the features of the last convolutional layer; target feature selection: firstly, optimizing a local activation graph, and then selecting a local feature vector according to an activation graph result, so as to realize more accurate target feature selection; feature weighted aggregation: evaluating the importance degree of each feature, so as to enable the weightedfine-grained local features to still be embodied during pooling aggregation and improve the precision of fine-grained retrieval; and performing image retrieval, and calculating cosine similarity between the characteristic vectors of the queried image and a database image. An image feature extraction and coding detail graph is shown in figure 1. According to the method, fine-grained image retrievalcan be realized, and the retrieval accuracy is improved.

The invention provides a multi-user concurrent modeling method and system based on embedded software for graphical modeling, and belongs to the field of graphical modeling. The method comprises the following steps: constructing an edition control tool server in a system server; establishing an edition control tool for graphical modeling, and creating a project containing graphical modeling staff file folders; distributing graphical modeling staff rights by a creator; performing modeling by a graphical modeling staff, enabling the graphical model to be divided into sub models by an edition control tool and preserving the sub models by the edition control tool, and besides, preserving global resources in a database in the system server; updating modified contents of other graphical modeling staffs from the edition control tool by the graphical modeling staff; after modeling, submitting the modified contents to the edition control tool. According to the multi-user concurrent modeling method and system disclosed by the invention, a vast modeling project is divided into small modules, so that the graphical modeling speed is greatly increased, the right management is definite, the structure is clear, different graphical modeling staffs do not influence each other, and preserved files can be reused. Therefore, duplication of labor can be reduced, and the working efficiency is improved.

The invention belongs to the field of image recognition and aims to provide a quick and effective large-scale target identification method based on a mobile platform for mobile platform search, and the method can be used for quickly and effectively encoding SIFT (scale invariant feature transform) characteristic points into binary codes with a function of keeping local sensitivity by a Hash algorithm. The method comprises the following steps: obtaining label information of SIFI characteristics X of a database image, wherein the label information consists of '0' and '1'; defining normalized distance similarity and quantization errors; searching a binary label of a data point with minimum value of the sum of NS and MD; obtaining weak Hash functions; combining the weak Hash functions to obtain a strong Hash function. The method is a quick and effective mobile platform search method; the search scheme of mobile equipment can be adjusted according to network conditions, so that responses can be given in time under different network conditions.

The invention discloses a face recognition method based on a caffe deep learning framework. The face recognition method comprises the following steps: step (1), establishing a face recognition database; step (2), preprocessing the database image; step (3), establishing a convolutional neural network by using a caffe framework; step (4) training a deep neural network model by using a caffe-based parallel framework; and step (5) calling the trained caffe model for testing. According to the method, on the basis of a caffe framework-based deep learning framework, appropriate model parameters can be better and faster trained under an improved neural network model, and parallel computing pictures are used in the recognition process, so that the recognition efficiency under the condition of a large amount of data is greatly improved.

Various aspects of the subject technology relate to augmented reality image retrieval. An application running on a user device allows the user to visualize and modify design features for a real-world object, such as a billboard, a bus-mounted advertisement, or a wall of a room in a building. The application is communicatively coupled to an image server having an image database with millions of available images, and having intelligent processing to identify recommended images from the database, based on the design aspects from the user. The application provides a selectable option to view the design aspects, and / or the recommended database images, on a captured image of the real-world object. In this way, the user is provided with the ability to design the appearance of the real-world object by visualizing and / or modifying, in real time, server-recommended images as they will appear on the real-world object.