47results about How to "Solve the technical problem of low recognition accuracy" patented technology

The invention discloses a multidimensional characteristic passenger re-identification method, a device, a storage medium and a processor, wherein the method comprises the steps of establishing a firstconvolutional neural network model according to a preset multidimensional characteristic extracting method; training a first convolutional neural network model according to a preset public data set,and obtaining a second convolutional neural network model; training a model group which is composed of the second convolutional neural network model according to a preset ternary set data set, and obtaining a third convolutional neural network model, wherein the model group comprises three parallel second convolutional neural network models; inputting a target passenger image and a to-be-identified passenger image to a third convolutional neural network model, thereby obtaining a target passenger characteristic vector and a to-be-identified passenger characteristic vector; and calculating a vector distance between the target passenger characteristic vector and the to-be-identified passenger characteristic vector, and obtaining an identification result according to the vector distance. Themultidimensional characteristic passenger re-identification method, the device, the storage medium and the processor settle a technical problem of relatively low identification precision in the passenger re-identification mode in prior art.

The present invention provides an article identification method and apparatus, and an indoor map generation method and apparatus. The article identification method comprises the following steps of: conducting continuous grasping, updating and learning of information in an Internet webpage to establish an article feature database; and matching and identifying indoor articles based on the article feature database. The invention solves the technical problem of low article identification accuracy of the prior art and achieves the technical effect of effectively improving the article identification accuracy. In addition, the article identification method is used to generate an indoor map to locate the indoor articles, thereby improving the positioning accuracy.

The invention discloses an identification method and an electronic device. The method is applied to the electronic device. The electronic device comprises a display unit. The method comprises the steps that when an object to be identified is placed on the display unit, a default image which is stored in the electronic device is read, wherein the variety of the colors of the default image is less than a threshold value; the default image is displayed in a first region of the display unit, wherein at least a part of the object to be identified is placed in the first region; at least one image acquisition device which is built in or externally connected with the electronic device is used to acquire a first image, wherein the first image is an image which comprises the image of the object to be identified; the first image is processed; and the object to be identified is identified.

The invention discloses a face recognition method, a mobile terminal and a computer readable storage medium. The face recognition method comprises the following steps that: when a terminal detects a face recognition unlocking instruction, according to a preset camera, obtaining a face image input by a user on a preset prompt expression, and prompting the user by the terminal to input the preset expression through voice, wherein the preset prompt expression is extracted from a preset expression image; extracting face characteristics from the face image by the terminal, and extracting glasses characteristics in the face image; carrying out matching analysis on the face characteristics, the glasses characteristics and a preset standard characteristic by the terminal to obtain a first similarity; and when the terminal detects that the first similarity is greater than a first threshold value, executing a terminal unlocking function. The glasses characteristics are compared and matched, so that the user still can obtain a high face recognition success rate even if the user wears the glasses so as to improve the face recognition efficiency of the terminal to improve user experience.

The invention provides a training method and device for identifying a model, wherein, the method comprises the following steps: acquiring a first training set and a second training set; based on the first training set, training the ANN model through a plurality of iterative processes to obtain a plurality of trained ANN models; determining a test accuracy rate of each trained ANN model by the second training set, wherein the test accuracy rate is an average accuracy rate of all characters tested by the ANN model; according to the test accuracy of each ANN model after training, selecting the model with the highest accuracy as the global optimal recognition model. Since the scheme sets up a training set for training, and a training set for determining the accuracy rate, and the model with the highest accuracy rate is finally selected as the most final model, thus solving the technical problem that the recognition accuracy rate of the existing trained recognition model is low, and achieving the technical effect of effectively improving the recognition rate of the recognition model obtained by training.

The invention discloses an identity recognition method, device and system. The method comprises: acquiring voiceprint features and a first timestamp of a target object, wherein the first timestamp isused for representing the time when the voiceprint features are acquired; obtaining a pre-stored voiceprint feature successfully matched with the voiceprint feature and a second timestamp from a voiceprint library, the second timestamp being used for representing the time when the pre-stored voiceprint feature is obtained; determining a similarity threshold based on the first timestamp and the second timestamp; and obtaining a recognition result of the target object based on the similarity threshold and the similarity between the voiceprint features and the pre-stored voiceprint features. According to the invention, the technical problem of low identification accuracy of an identity identification method in the prior art is solved.

The invention provides a geographic information identification method and device based on deep learning. The method is based on a BiLSTM + CRF deep neural network model. an aggregation layer suitablefor a fine-grained labeling strategy is added; a unique geographic information analysis model (BiLSTM + CRF + AGG) is formed, and through careful improvement and optimization, the system has very highfault-tolerant capability, and relatively accurate geographic position information identification is realized on text information which is nonstandard and incomplete and has wrongly written characters; a plurality of complete and mutually independent auxiliary systems are developed; the deep learning model developed by python is seamlessly applied to a windows development platform, secondary development of mainstream languages such as C + +, C #, java and the like is friendly supported, and an implementation scheme from research to application is realized. The technical effect of improving the recognition precision is achieved.

The invention discloses a portrait infringement judging method and device, electronic equipment and a storage medium, and relates to the technical field of deep learning and image processing. The specific scheme comprises the following steps of inputting a to-be-detected image into a pre-trained infringement judgment model; obtaining face features of the to-be-detected image through the infringement judgment model; calculating a similarity score of the to-be-detected image and each celebrity image based on the face features of the to-be-detected image and the face features of each celebrity image in a pre-constructed celebrity image library; wherein the image library comprises at least one real celebrity image and celebrity images in a cartoon form corresponding to the real celebrity images; and in response to the fact that the similarity score is higher than or equal to the first predetermined score, determining that the to-be-detected image is an infringement image. According to theembodiment of the invention, the portrait in a cartoon form can be recognized, so that portrait infringement can be avoided to a greater extent.

The invention discloses a face recognition method and a device, and an establishment method of a face detection library. The method comprises the following steps: acquiring image information of a target object; extracting facial features of a face in the image information; matching the face feature with at least one preset face feature in a face detection library, and determining a preset face feature successfully matched with the target object, the face detection library comprising a preset face feature of the invited object and a preset face feature of an object detected in a preset area; and obtaining the identity information of the successfully matched preset facial features to obtain the identity information of the target object. The technical problem that in the prior art, face recognition is conducted on the target object, and the recognition precision is low is solved.

The invention discloses a method and a device for identifying insurance claim fraudulent behaviors. The method comprises the following steps: acquiring voice data of an insurance claim settlement user, the voice data being voice data in an insurance claim settlement process of the insurance claim settlement user; processing the voice data to obtain feature information of the voice data; and judging whether the insurance claim settlement user has a fraud behavior in the insurance claim settlement process or not according to the feature information. The technical problem that an existing method for identifying insurance claim fraudulent behaviors is low in identification accuracy is solved.

The invention provides a heart sound signal classification method and device, equipment and a storage medium. The classification method comprises: when the processor receives to-be-classified heart sound signals sent by a user side, storing the to-be-classified heart sound signals in the cache module, and determining a preset signal classification model corresponding to the to-be-classified heartsound signals in the storage module; the processor acquiring signal characteristics of abnormal standard signals and signal characteristics of normal standard signals in the signal classification model, and acquiring signal characteristics of the heart sound signals to be classified; and the processor comparing the heart sound signals to be classified with the signal characteristics of the abnormal standard signals and the normal standard signals to determine the types of the heart sound signals. According to the method, the heart sound signal classification model is constructed, and the heartsound signals to be classified are determined based on the signal characteristics corresponding to the abnormal standard signals and the normal standard signals, so that the abnormal heart sound signal identification efficiency is improved.

The invention discloses an urban functional area quantitative identification method based on open source geospatial vector data. According to the method, a framework system and a technical process foridentifying an urban functional area under the guidance of open source data are established. Firstly, data quality is evaluated through a qualitative analysis method, a mapping relation between various open source data sources and national classification standards is established through semantic analysis, and a fusion strategy is formulated. In the method, a research unit is divided by geographicnational condition road data, and classification reference data is composed of three parts: OSM fusion surface data, single building data extracted from an image and POI data. The city blocks are divided into five classes by judging the spatial relationship between the city division blocks and the classification reference data. The method provided by the invention can improve the category extraction precision and the identification precision.

The invention discloses a name recognition method and a device, computer equipment and a readable storage medium. The method comprises the following steps: inputting original name information into a preset network model for training to obtain a name labeling model; inputting name information to be identified into the name labeling model to obtain labeled name information; and extracting names in the marked name information according to a first preset rule. According to the method and the device, the technical problem of low identification accuracy caused by dependence on a word segmentation tool in a name identification method in related technologies is solved. According to the method and the device, the purpose of identifying the name without depending on a word segmentation tool is achieved, so that the technical effect of improving the name identification accuracy and identification efficiency is achieved.

The invention provides a method and device for identifying a large game application. The method comprises the following steps: detecting an occupied space of an installation package of a running gameapplication; detecting a display mode of the running game application; and identifying whether the game application is a large game or not according to the occupied space and the display mode. The technical problem that in the prior art, a large game is identified only according to the name of the game package, and identification accuracy is low is solved.

The invention discloses a deteriorated article identification method and device, equipment and a medium. The deteriorated article identification method comprises the following steps: receiving an image of a to-be-identified article, determining an initial identification result corresponding to the to-be-identified article based on the image of the to-be-identified article; and identifying the to-be-identified article through a preset first deteriorated article identifier corresponding to the initial identification result; obtaining a first discrimination result, and identifying the to-be-identified article through a preset second deteriorated article identifier corresponding to the initial identification result; obtaining a second discrimination result, fusing the first discrimination result and the second discrimination result to obtain a fused discrimination result of the to-be-identified article, and comparing the fused discrimination result with a preset freshness threshold value to determine the actual freshness of the to-be-identified article. According to the invention, the technical problem of low identification accuracy of deteriorated articles in the prior art is solved.