36results about How to "To achieve the recognition effect" patented technology

The invention relates to a finger clicking character recognition method and a translation method based on artificial intelligence. The recognition method comprises the following steps of (1) respectively constructing and training each neural network; (2) using the acquisition device to acquire the current image of the pre-detection area in real time, and continuously inputting the image into the pre-trained finger positioning neural network to obtain the finger position information under the finger click state; 3) taking that position of the user's fin as the center, intercepting the image ofthe box area, inputting the angle recognition neural network, and outputting the rotation angle of the text in the image area; 4) rotating the rotation angle to intercept the frame area image with theposition of the user's fin as the center, and outputting the position information and the size information of the detected character area; (5) intercepting a corresponding image, inputting the OCR recognition neural network, and outputting the recognized text content. The invention not only improves the identification efficiency but also enables the identification artificial intelligence to be realized.

The invention provides an electronic nose gas identification method based on source domain migration extreme learning to realize drift compensation. According to the source domain migration extreme learning to realize drift compensation, a source domain migration extreme learning machine framework is proposed from the perspective of machine learning and used for solving the problem of sensor drift instead of direct correction for single sensor response; a source domain data set and a target domain data set are built according to labeled gas sensor array sense data matrixes collected by an electronic nose before drift and after drift respectively and are taken as inputs of an extreme learning machine for training an identification classifier of the electronic nose, so that the tolerance performance of the identification classifier on gas identification after the electronic nose drifts is improved, and the purposes of drift compensation and gas identification precision improvement are achieved; besides, technical advantages of the extreme learning machine are kept, and accordingly, the method has better generalization performance and migration performance. Therefore, based on the source domain migration extreme learning machine framework provided by the invention, one learning framework with good learning capacity and generalization capacity is built.

The embodiment of the invention provides a face recognition method, a face recognition system, a medium and electronic equipment, and relates to the technical field of big data. The method comprises the steps of acquiring a plurality of face samples to serve as a first training set, a first model based on the first training set is trained through a clustering method, and a first recognition resultof a face sample to be recognized is obtained through the first model; Determining a plurality of triplets according to the first recognition result, calculating the similarity of the plurality of triplet samples, and training a second model by taking the sample set with the similarity of the samples meeting a first threshold condition as a second training set to obtain a second recognition result; Taking the sample set with the similarity meeting a second threshold condition in the second recognition result as a third training set, and training a third model; And identifying the to-be-identified face based on the trained first model, second model and third model. According to the technical scheme of the embodiment of the invention, the recognition precision of the face recognition systemcan be improved.

The invention discloses a method for realizing smoke detection by using a deep learning classification model. The method comprises the following steps of: getting a frame of smoke image from the videostream; processing the smoke image with a Gaussian mixture model to obtain a motion region of the smoke image to be removed; processing the image using a dark channel smoke removing algorithm to obtain a smokeless image model; obtaining a difference image between the smoke image to be removed and the smokeless image model; binarizing the difference image to obtain a suspected smoke area; obtaining the intersection area between the motion area and the suspected smoke area, entering the intersection area into the trained deep learning classification model to get the final smoke recognition result; marking the smoke area in the smoke image to be removed according to the smoke recognition result. The method and system for realizing smoke detection by using deep learning classification model use a lightweight deep learning classification model to achieve higher accuracy and detection rate, reduce the false detection rate, and realize the effect of real-time detection.

The invention discloses an end-to-end lightweight deep license plate recognition method, and belongs to the field of image processing and deep learning. The method comprises the steps of collecting animage containing a license plate, and constructing a license plate data set; constructing a license plate detection network; acquiring a license plate area image; constructing a license plate recognition network; regarding the license plate detection network and the license plate recognition network as a whole, and performing end-to-end training on the whole network; and recognizing the license plate in the to-be-detected license plate image by using the trained network. According to the method, end-to-end training is carried out on the model, the calculated amount and time of the license plate recognition process are reduced through the lightweight convolutional neural network, and the method has the advantages of being low in requirement for the shooting angle of the to-be-recognized image, accurate in license plate positioning, high in license plate character recognition accuracy, high in recognition speed and the like.

The invention discloses preparation of a supermolecule gel factor based on three-column [5] aromatic hydrocarbon. The supermolecule gel factor is a complex organic gel factor which is synthesized fromamide modified column [5] aromatic hydrocarbon and trimesoyl chloride through a nucleophilic substitution reaction and is based on column aromatic hydrocarbon. According to the gel factor, blue organic gel with aggregation state induction florescence can be formed in cyclohexanol through C-H...pi,pi...pi accumulation and Van der Waals' force. In the organic gel, water solutions of Hg<2+>, Ca<2+>,Mg<2+>, Ni<2+>, Cr<3+>, Cd<2+>, Pb<2+>, Ag<+>, Zn<2+>, Ba<2+>, La<3+>, La<3+>, Eu<3+> and Tb<3+> are respectively added, only addition of Hg<2+> can quench blue florescence of the organic gel, and addition of other cationic ions does not remarkably influence fluorescence intensity of the organic gel, so that specific fluorescence detection on Hg<2+> can be achieved, and a detection limit is 1.02*10<-8>M.

The present invention provides a machine vision-based handwriting recognition method and a machine vision-based handwriting recognition system. The method comprises: collecting a target image, obtaining and storing the characteristic points of the target image, generating trace information of the characteristic points, and processing and recognizing the trace information; recognizing characters or symbols written by a user by using information of the handwriting trace captured by a camera. In this way, the traditional devices such as touch screens, electromagnetic induction screens, handwriting boards, and mouse are replaced, and desired characters can be written by waving fingers instead of walking to a computer or a touch media to operate.

The invention discloses a character assembly-oriented graphic verification code identification method and device, and aims at providing the graphic verification code identification performance. The method comprises the following steps of: obtaining a to-be-identified graphic verification code which consists of characters, and preprocessing the graphic verification code; cutting the graphic verification code to obtain at least one graphic; determining the quantity of characters included in each cut graphic; for each cut graphic which includes one character, matching the graphic with a preset sample set, and identifying the character included in the graphic; for each cut graphic which includes more than one character, matching the graphic with the sample set by adoption of a bidirectional sliding and downward deviation combined matching method, and identifying the characters included in the graphic; and combining identification results of the characters included in all the graphics to obtain an identification result of the graphic verification code.

The invention provides a triboelectrification-based intelligent key and an intelligent keyboard. The intelligent key comprises a sensing component; the sensing component comprises an insulating layer, an upper electrode layer and a lower electrode layer stuck to the upper and lower surfaces of the insulating layer separately, and a touch layer stuck to the upper electrode layer; the touch layer is in contact with a knocking object to generate electrical signals between the upper electrode layer and the lower electrode layer; and the surfaces of the materials for the touch layer and the knocked object have different electron gaining and losing capabilities. Correspondingly, the invention further provides a touch pen used with the intelligent key and the intelligent keyboard, wherein the materials for a contact of the touch pen and the touch layer have different electron gaining and losing capabilities. The mechanical energy of knocking the keyboard can be directly converted into electrical signals via the sensing component to drive a wireless transmitting device to realize alarm of the keyboard. A resistor is arranged in each key of the keyboard and connected with an output end, so that the keystroke position of the keyboard can be positioned and the keystroke mode can be identified.

Provided is a method for identifying a motion state. The method includes the steps of collecting the three-dimensional acceleration data of a target through an acceleration sensor, calculating resultant acceleration according to the three-dimensional acceleration data, and extracting the feature information of the resultant acceleration; inputting the feature information into a decision tree model, using the node of the decision tree model to identify the feature information, and determining the motion state of the target. According to the method, the three-dimensional acceleration data of thetarget is collected by the acceleration sensor configured on the to-be-identified target, and therefore the feature information is extracted from the resultant acceleration after the resultant acceleration of the target motion is calculated, and input into the decision tree model to be determined and identify the motion state of the target; the problem is solved that in the traditional technology, the identifying efficiency is low since a large amount of image data needs to be collected, and the technical effect of efficiently identifying the motion state of the target is achieved. The invention further provides a system for identifying the motion state and an animal behavior identifying system.

The invention discloses a pyroelectric human body transducer-based direction recognition device and a recognition method thereof. The pyroelectric human body transducer-based direction recognition device comprises an optical system, a pyroelectric infrared transducer, a signal amplifier, a signal comparator and a voice prompt device, wherein a Fresnel lens is adopted by the optical system; only one pyroelectric infrared transducer is available; the signal comparator is provided with two output ends TG1 and TG2; the TG1 and the TG2 are connected with the voice prompt device respectively. The pyroelectric human body transducer-based direction recognition device and the recognition method thereof have the beneficial effects that a double-probe recognition effect or an effect that a single probe is matched with a strong signal processor is achieved through a physical isolation and probe output signal analysis combined mode, the cost is saved and the device can be used on general household or unsafe and unreliable occasions.

The invention provides a dictation interaction method, system and device based on AI vision. The method comprises the steps of: S100, obtaining a collected target image in real time; S200: constructing and training a plurality of convolutional deep neural networks and cyclic deep neural networks, or based on a Transformer deep neural network combined structure of a self-attention mechanism, carrying out comprehensive weighting calculation on a plurality of combined structure output results for handwritten font recognition by utilizing a dynamically planned common substring matching algorithm,and recognizing action information and text information in the target image; S300, according to the identified action information, executing a dictation control task or a dictation control task; S400,controlling to play dictation content of the dictation task; and S500, controlling and displaying prompt content and a dictation result in the dictation task. Through multiple convolutional deep neural networks, interaction between gestures and dictation equipment is realized, the recognition accuracy is improved, the recognition speed is increased, and the use experience of a user is enhanced.

The invention discloses a network navy identification method and device, a storage medium and a processor. The network navy identification method comprises the following steps: capturing text data information related to a target keyword in a preset time period from the Internet; inputting the text data information into a network navy identification model, wherein the network navy identification model is trained by using multiple groups of historical text data through machine learning, and each group of historical text data in the multiple groups of historical text data comprises historical text data information and network navy characteristics identified in the historical text data information; and identifying whether network navy participates in the text data information from an output result of the network navy identification model. Through the network navy identification method and device, the technical problem that the network navy is difficult to identify in related technologies is solved.

The invention discloses a voice orientation recognition method, device, and system and a home controller. The voice orientation recognition method comprises: a voice sent out by a target sound sourceis processed to obtain audio data; the audio data are calculated based on a preset algorithm to obtain voice location information, wherein the voice location information is information of the locationof the e target sound source; characteristic data of the audio data are obtained, wherein the characteristic data are data corresponding to password information of the voice; whether the characteristic data matches preset characteristic data is determined; if so, the voice location information is outputted; and if not, not outputting is carried out. Therefore, the voice orientation is recognizedeffectively.

The invention provides a writing font copying auxiliary method, system and device based on AI vision. The writing font copying auxiliary method comprises the steps that S1, an image including finger action information and a writing medium in a view field range is acquired in real time; s2, constructing and training a plurality of convolutional neural networks; identifying finger action informationin the image and a target image pointed by the finger; recognizing text content and glyph information thereof in the target image, and outputting a text image, the text content and the glyph information thereof; s3, searching a reference font image from a preset font library according to the text content and the font information of the text content, and performing similarity comparative analysison the reference font image and the text image; and S4, outputting the text content and the comparison result. The copying auxiliary method is not limited to writing media, and the writing practice effect is effectively improved.

The embodiment of the invention discloses an image recognition method and device, computer equipment and a storage medium, and the method comprises the steps: obtaining a first image and a second image containing a target object, carrying out the category and position prediction of the target object in the first image through an initial recognition model, and obtaining a first prediction categoryand a first prediction position; carrying out convergence on the first prediction category and the target category, carrying out convergence on the first prediction position and the target position, carrying out adversarial learning on the first image and the second image through the initial recognition model, and obtaining a candidate recognition model; obtaining a target category and a pseudo target position corresponding to the target object in the second image through the candidate recognition model; inputting the second image into a candidate recognition model for category and position prediction to obtain a second prediction category and a second prediction position; and converging the second prediction category and the pseudo target category, and converging the second prediction position and the pseudo target position to obtain the trained recognition model, so that the accuracy and reliability of model training are improved.

The invention discloses a dish identification method, a cashier method, a dish order urging method, a related device, computing equipment and a medium. The dish recognition method comprises the stepsthat a dish image to be recognized is input into a first dish category recognition model to be processed so as to obtain image features output by a bottleneck layer in the first dish category recognition model, and the bottleneck layer comprises all processing layers in front of the last processing layer in the first dish category recognition model; and inputting the image features into a second dish category identification model for identification so as to obtain the category of the dish in the dish image.

The invention provides a cross-domain face recognition algorithm, a storage medium and a processor. The cross-domain face recognition algorithm method comprises the following steps of: establishing aFacenet neural network; adding an average pooling layer and a Faltten layer at the highest dimension layer of the feature vector of the Facenet neural network, and converting the feature vector into asingle-channel one-dimensional vector; calculating the maximum mean value difference loss by using a single-channel one-dimensional vector; and adding the maximum mean value difference loss into a loss function of the Facenet neural network, and making the maximum mean value difference loss and the loss function jointly participate in back propagation and gradient derivation. A network structureof an original Facenet algorithm is improved, MMD values of different domains are calculated at the highest layer of the feature dimension, and the MMD values are added into the synthetic loss function; due to the fact that the inter-domain statistical distribution difference is eliminated through the improved algorithm, a cross-domain face recognition effect is achieved.

The invention discloses poker cards and a recognition device and a recognition method thereof. The poker cards comprise a plurality of card faces with different suits and values. Each card face is provided with a plurality of encoding positions capable of providing different sensitization values, the encoding positions are arranged sequentially, each encoding position is coated with the sensitization values, and values corresponding to the card faces are obtained through calculation according to a preset sequence. In an automatic dealing machine, photovoltaic conversion components are used for converting light signals reflected by sensitization materials on each poker card face into electric signals. A sensitization value line is converted into an electric signal line, and finally recognition of the poker cards is achieved. The recognition method is simple, modification of the existing poker cards is small, and recognition cost is low. Simultaneously, recognition efficiency of the sensitization materials is high, fault rate is low, and the recognition device and the recognition method can effectively improve poker card recognition efficiency and reliability in the automatic dealing machine.

The present invention provides an electronic nose gas identification method based on source domain migration extreme learning drift compensation. It proposes a domain migration extreme learning machine framework from a machine learning perspective to solve the sensor drift problem, instead of directly correcting a single sensor response, using The labeled gas sensor array sensing data matrices collected by the electronic nose before drifting and after drifting were used to construct source domain data sets and target domain data sets respectively as inputs to the extreme learning machine, and the recognition classifier of the electronic nose was carried out. Learning to improve the tolerance performance of the recognition classifier for gas recognition after the electronic nose drifts, to achieve the purpose of drift compensation and improve gas recognition accuracy, and maintain the technical advantages of the extreme learning machine, making this method have better Generalization and transfer performance. It can be seen that the source domain transfer extreme learning machine framework proposed in the method of the present invention establishes a learning framework with good learning ability and generalization ability.

Embodiments of the present invention provide a face recognition method, a face recognition system, media, and electronic equipment, and relate to the field of big data technology. The method includes: acquiring a plurality of human face samples as a first training set, training a first model based on the first training set by a clustering method, and using the first model to obtain a first recognition result of the human face samples to be recognized Determining a plurality of triplets according to the first recognition result, calculating the similarity of the multiple triplet samples, and using the sample set whose similarity of the samples satisfies the first threshold condition as the second training set to train the first Two models to obtain the second recognition result; use the sample set whose similarity meets the second threshold condition in the second recognition result as the third training set, and train the third model; based on the trained first model, the second The model and the third model recognize the face to be recognized. The technical solutions of the embodiments of the present invention can improve the recognition accuracy of the face recognition system.

The invention provides a human body identity gait recognition system based on combination of visual sense and tactile sense and a recognition method thereof, and relates to an information recognition device. The system is composed of a plantar pressure information acquisition device, a gait information acquisition device, a signal conditioning circuit, an A / D converter, a microprocessor, a transmission module and an upper computer provided with an upper computer recognition work procedure, wherein a pressure measurement slab is arranged at a position, on which human body plantar pressure information measurement needs to be carried out, of a walking channel of a test site, a camera lens is arranged at a position, aligned to the space on the front upper portion of the walking channel of the test site so that gait recognition of a recognized person can be collected, human body identity gaits of the recognized person can be recognized as long as the recognized person normally walks on the pressure measurement slab on the walking channel of the test site, and the defects that according to an existing gait identity recognition technology, after image sequences are obtained through the camera lens, picture processing is carried out, features are extracted, recognition is conducted, and breakthrough of bottlenecks restricting the system in practical application cannot be achieved are overcome.

The embodiment of the present application discloses an image recognition method, device, computer equipment and storage medium. The first image and the second image containing the target object are obtained, and the category and position of the target object in the first image are predicted through the initial recognition model. Obtaining the first predicted category and the first predicted location; converging the first predicted category with the target category, converging the first predicted location with the target location, and performing adversarial learning on the first image and the second image through the initial recognition model, Obtaining a candidate recognition model; obtaining a target category and a false target position corresponding to the target object in the second image through the candidate recognition model; inputting the second image into the candidate recognition model for category and position prediction to obtain a second predicted category and a second predicted position; The second predicted category is converged with the pseudo-target category, and the second predicted position is converged with the pseudo-target position to obtain a post-training recognition model, which improves the accuracy and reliability of model training.

The invention discloses a method for realizing smoke detection by using a deep learning classification model. The method comprises the following steps of: getting a frame of smoke image from the videostream; processing the smoke image with a Gaussian mixture model to obtain a motion region of the smoke image to be removed; processing the image using a dark channel smoke removing algorithm to obtain a smokeless image model; obtaining a difference image between the smoke image to be removed and the smokeless image model; binarizing the difference image to obtain a suspected smoke area; obtaining the intersection area between the motion area and the suspected smoke area, entering the intersection area into the trained deep learning classification model to get the final smoke recognition result; marking the smoke area in the smoke image to be removed according to the smoke recognition result. The method and system for realizing smoke detection by using deep learning classification model use a lightweight deep learning classification model to achieve higher accuracy and detection rate, reduce the false detection rate, and realize the effect of real-time detection.

The invention discloses a liquid storage box which is internally provided with a sealed cavity for storing additives, at least one concave-convex part is arranged on the outer wall of the liquid storage box, and the combination of the number and the position of the concave-convex parts arranged on the liquid storage box correspondingly represents the type of the additives stored in the liquid storage box. Meanwhile, the invention further provides an additive feeding module, the additive feeding module is provided with at least one containing part for installing the liquid storage box, a plurality of contact switches are arranged in the containing part, and the contact switches cover the corresponding contact positions of the concave-convex parts arranged on the different liquid storage boxes. And each concave-convex part arranged on the liquid storage box is respectively contacted with a contact switch arranged in the accommodating part. Through the arrangement, the types of the additives in the liquid storage box are determined based on the information of the contact switch in contact with the concave-convex part, so that the effect of accurately identifying the types of the additives in the liquid storage box is achieved.

The invention discloses a deep learning oral pill identification method based on multiple views and data expansion. A database is established by adopting a multi-view and data augmentation method, and a data set is perfected from multiple angles. A lightweight network is used, and a practical model embedded into mobile equipment and small and medium-sized equipment is designed. And combining multiple views with a two-dimensional model, and completing the construction of a practical model after transfer learning. Meanwhile, an incomplete oral pill identification channel is established, and incomplete pills are subjected to template matching to be restored into complete pill pictures and then are identified. The method effectively classifies the medicines with highly similar shapes and colors, assists medical staff in sorting the medicines, and reduces and even avoids life safety problems of patients caused by wrong medicine classification. The overfitting problem caused by small data volume is solved through multi-view database building, data augmentation and transfer learning, a lightweight model MobileNetv2 is adopted as a basic framework, an attention module mechanism is introduced, the parameter quantity of the model is greatly reduced compared with that of a three-dimensional model, and the method is convenient, practical and easy to popularize.

The invention discloses an electronic ticket business data integration method, which comprises remote movement, a ticket business center, a scanning module, ticket information, a data module, an information confirmation module, an adjustment module and a mobile terminal.The integration method achieves the effect that seats can be purchased or replaced without arriving at a station through cooperation of the remote movement and the ticket business center. According to the integration method, the tedious problem that a user needs to arrive at a station to replace a seat in the past is solved, through cooperation between the scanning module and the ticket information, the effect of identifying the ticket information through multiple modes of bar codes, two-dimensional codes, face identification and identity card information is achieved, and the integration method is suitable for popularization and application. According to the integration method, the problem that a trip is delayed due to the fact that a ticket needs to be replaced at a window when the ticket is lost in the past is solved, and through cooperation between the data module and the adjusting module, passengers can know and change departure shifts and seat information through intelligent mobile operation; the problem that in the past, passengers cannot know the current shift and all seat information of the shift randomly due to travel changes is solved.

The invention provides a method for manufacturing and identifying a plastic cup and a plastic bowl; the edge surface of the plastic cup and the plastic bowl manufactured by the method is provided witha plurality of concave-convex marks. The process thickness of a material at the joints of the concave-convex shapes and the cup edge is smaller than that of the cup edge, so the concave-convex edgesare softer and easier to bend than the cup edge. In addition, the connection boundary of the concave-convex shapes and the cup edge adopt a point breaking process, so that the concave-convex shape boundary is easier to break off and the cup edge is separated. A user only needs to slightly press the concave-convex shapes with fingers and break the concave-convex shapes under the action of point disconnection stress, so that the convex recess is sunken to form a concave shape, the boundary of the concave-convex shape parts is broken, the connection with the cup edge and the plastic elasticity are lost, and rebounding and restoring cannot be achieved. Therefore, concave-convex comparison is formed with other convex shapes which are not pressed, a self-set recognition effect can be achieved incooperation with character or pattern marks required for recognition, and the plastic cup can be distinguished from other plastic cups. And a plurality of concave-convex marks are matched and combined, so that the possibility of repetition is basically avoided. Therefore, a good recognition effect of the self-set marks is achieved.

Provided is a method for identifying a motion state. The method includes the steps of collecting the three-dimensional acceleration data of a target through an acceleration sensor, calculating resultant acceleration according to the three-dimensional acceleration data, and extracting the feature information of the resultant acceleration; inputting the feature information into a decision tree model, using the node of the decision tree model to identify the feature information, and determining the motion state of the target. According to the method, the three-dimensional acceleration data of thetarget is collected by the acceleration sensor configured on the to-be-identified target, and therefore the feature information is extracted from the resultant acceleration after the resultant acceleration of the target motion is calculated, and input into the decision tree model to be determined and identify the motion state of the target; the problem is solved that in the traditional technology, the identifying efficiency is low since a large amount of image data needs to be collected, and the technical effect of efficiently identifying the motion state of the target is achieved. The invention further provides a system for identifying the motion state and an animal behavior identifying system.