8 results about "Visual dictionary" patented technology

The application provides a knowledge graph representation method for key neuron hierarchical relationship in a deep neural network, comprising: calculating the confidence of each neuron in a high layer convolution layer of the deep neural network; deriving the importance of each neuron in the high layer convolution layer according to a classification result; calculating the influence degree of each neuron in the high layer convolution layer of the deep neural network on a current image; determining a key neuron according to the influence degree of each neuron in the high layer convolution layer on the current image; visualizing the key neuron; marking each obtained neuron visual feature map as a semantic feature of the neuron, extracting the weight and layer information of the neuron from the network, constructing a relationship according to the weight and layer information of the neuron in a triple mode, matching the relationship with a visual dictionary, generating a semantic dictionary, and generating a visual knowledge graph. The application realizes the explainability and user readability of the deep neural network.

The invention relates to a line fault reason intelligent identification method based on a recording waveform image, and the method comprises the steps: carrying out the feature extraction of a fault waveform image through employing a scale invariant feature transform (SIFT) feature extraction method, and mining key feature information in the fault waveform image; a clustering method (K-means) is utilized to classify similar feature descriptors according to clusters to form a visual dictionary, and on this basis, an image pyramid technology is utilized to realize multi-scale feature fusion of image vectors; and a support vector machine (SVM) classifier is introduced, and vectorized image features are utilized to carry out accurate identification on fault causes. The line fault reason identification model constructed by the invention can effectively predict various power grid line fault reasons, and improves the accuracy and timeliness of fault reason identification; good robustness is shown in an actual case test, and generalization performance and stability of the model are ensured through cross validation and monitoring of model performance indexes; and powerful support is provided for fault diagnosis and maintenance of a power system.

The invention discloses a method and system for recognizing the semantic position of an automatic driving automobile by using an image retrieval technology, and the method comprises the steps: receiving a subjective position label inputted by a user, and obtaining and presenting a related candidate image from a classification visual dictionary based on the subjective position label; a user selects from the candidate images to obtain a query image and then pre-processes the query image; extracting low-level visual features and color information from the preprocessed query image to form a query feature vector; performing similarity calculation on the query feature vector and feature vectors of a plurality of storage images in a database; determining a semantic position image based on a result of the similarity calculation; and deciding the autonomous vehicle according to the determined semantic position image. Required semantic positions are tracked and retrieved by using significant features and color information, so that the semantic position recognition capability of the automatic driving vehicle is enhanced, and the autonomy and reliability of the vehicle in a complex driving environment are improved.

Embodiments of the present application provide an ancient character deciphering method, device, equipment and storage medium, and relate to the technical field of ancient text deciphering. An ancient character visual dictionary composed of multiple key-value pairs is obtained, a feature encoder is used to perform feature extraction on each synthetic ancient character image in the ancient character visual dictionary to obtain ancient character image features, and the ancient character image features constitute an ancient character feature library; a target ancient character image to be deciphered is obtained, and the feature encoder is used to obtain target image features corresponding to the target ancient character image; the similarity values of the target image features and each ancient character image feature in the ancient character feature library are calculated, and the deciphering result of the corresponding target ancient character image is determined according to the similarity values. The visual dictionary is used to generate search samples covering multiple writing style variants, the defect of a small scale of a real ancient character database is overcome, the coverage range of a search benchmark is expanded, the overall accuracy and automation efficiency of large-scale ancient character deciphering are improved, and strong interpretability is achieved.

The invention discloses a method for processing an uplink signal time-frequency graph of an unmanned aerial vehicle, which comprises the following steps of: under a preset environment condition, acquiring spectrum data of multiple types of unmanned aerial vehicles needing to be identified in at least two periods for multiple times, and determining a distribution rule of uplink frequency hopping signal points in a picture based on the spectrum data, according to the characteristics of the uplink frequency hopping signal points and the characteristics of the formed image, the distribution rule of the uplink frequency hopping signal points generated by the unmanned aerial vehicles of different remote control models and needing to be identified in the image and the characteristics of the formed image are obtained; the method comprises the following steps: generating a corresponding uplink signal time-frequency graph template based on the distribution rule of uplink frequency hopping signal points generated by unmanned aerial vehicles of different remote control models and needing to be identified in a picture and the characteristics of the formed image, and finally obtaining an unmanned aerial vehicle visual dictionary; the unmanned aerial vehicle visual dictionary is used as a template for evaluating and matching collected unmanned aerial vehicle remote control uplink signals in an algorithm framework for identifying the unmanned aerial vehicle.

A method for image generation includes: processing an input text sequence by using a trained language model to obtain an output sequence output by the language model, the output sequence including a plurality of indices in a language dictionary associated with the language model, the language model being trained on the language dictionary, the language dictionary including at least an index set corresponding to text encodings in a natural language and an index set corresponding to image encodings; constructing image encodings corresponding to the plurality of indices in the output sequence into a target feature map; and determining, by using a trained image decoder, a target image matching the text sequence from the target feature map, the image decoder being trained on a visual dictionary including the index set corresponding to the image encodings.

This invention discloses a robot relocalization method, apparatus, and robot. In the method, laser point cloud data and image data acquired by the robot at its current position are received; keyframe information corresponding to the image data is determined according to a visual dictionary; physical location information corresponding to the keyframe information is obtained from a visual map; feature point information of the image data is matched with feature point information corresponding to the keyframe information; and calculation is performed by combining the physical location information of the feature points in the visual map and the pixel location information of the feature points in the image data. After obtaining the robot's first pose data, a predetermined area is set based on the first pose data, and iterative matching calculations are performed within the predetermined area using the laser map and laser point cloud data to obtain the robot's second pose data. By using the above scheme, the iteration range of laser matching can be reduced by utilizing the initial pose data, thus accelerating the matching speed and quickly achieving robot relocalization.

The application particularly relates to a vector character recognition method based on a bag-of-words model feature point retrieval, which comprises the following steps: S100, reading character contour information of any vector graph by reading vector graph data; S200, analyzing the character contour information into control point coordinates; S300, drawing the control point coordinates into a control point grayscale graph; S400, extracting an ORB feature vector according to the control point grayscale graph; S500, taking the ORB feature vector as input, and searching for a character ID with the highest similarity from a visual dictionary through a bag-of-words tree index; and S600, obtaining a font and a unicode code corresponding to the vector character through a character ID mapping relationship. Through the above scheme, the vector graph file can be directly subjected to character recognition without format conversion, and meanwhile has the following multiple advantages: first, the character recognition range is large, the accuracy is high, and the character set can be extended to be larger; second, the character recognition speed is fast, and the single character recognition speed is about 1.5 ms; and third, the font can be judged while the character recognition is performed.