70results about How to "Guaranteed recognition efficiency" patented technology

The invention discloses an expression identification method fusing a depth image and multi-channel features. The method comprises the steps of performing human face region identification on an input human face expression image and performing preprocessing operation; selecting the multi-channel features of the image, extracting a depth image entropy, a grayscale image entropy and a color image salient feature as human face expression texture information in the texture feature aspect, extracting texture features of the texture information by adopting a grayscale histogram method, and extracting facial expression feature points as geometric features from a color information image by utilizing an active appearance model in the geometric feature aspect; and fusing the texture features and the geometric features, selecting different kernel functions for different features to perform kernel function fusion, and transmitting a fusion result to a multi-class support vector machine classifier for performing expression classification. Compared with the prior art, the method has the advantages that the influence of factors such as different illumination, different head poses, complex backgrounds and the like in expression identification can be effectively overcome, the expression identification rate is increased, and the method has good real-time property and robustness.

The invention discloses a radar and video fused large-scale monitoring system and method. The device comprises a central control device and monitoring terminals which are connected with the central control device and respectively arranged at needed monitoring areas; the monitoring terminals comprise a controller, a video monitoring module and a radar monitoring module composed of more than one radar monitor, wherein the video monitoring module and the radar monitoring module are respectively connected with the controller; all radar monitors are arranged at the needed directions where the needed monitoring areas can be completely covered within each corresponding coverage area; when a target radar monitor monitors a target, the controller controls the video monitoring module to adjust an acquisition direction and transmit an acquired target image to the central control device to perform recognition and tracking. The method uses the system to realize large-scale intelligent monitoring. The radar and video fused large-scale monitoring system and method provided by the invention have the advantages that omnidirectional intelligent monitoring of a large-scale area can be realized, the target can be quickly and accurately tracked and recognized and the monitoring precision is high.

The invention provides a design method of an active defense system based on white list technology and a system, belonging to the technical field of safety. According to the design method of the activedefense system, the reliability of a white list source is guaranteed through collecting the Hash value of a legal copy operation system mirror image; network visits are reduced through generating andcustomizing a local white list library for a customer, the recognition efficiency of a program is improved to some extent, and the starting speed of a run program is guaranteed; the pragmaticality and recognition efficiency of the white list are guaranteed through acquiring a patch updated by an operating system in real time; a recognition function is trusted through an unknown program, the resource requirement on traditional white list technology is solved, and the non-limitation of the white list is guaranteed. The method and the system can be used for effectively solving the problem whicha blacklist confronts, and are reliable, efficient, pragmatic and non-limited in the realization of the active defense system design of the white list.

The invention relates to a method and a system for identifying the bar code of a server-assisted mobile terminal, belonging to the technical field of the mobile communication value-added services. The method comprises the following steps of: collecting bar code images, resolving the bar codes, invoking the server-assisted resolving capability, transmitting through the network connection, managing the network connection, assisted-resolving the bar codes, processing resolution results, and finally showing identification results to a user at the bar code identifying client. The invention provides a two-stage identification method including the mobile terminal direct identification and the server-assisted identification. The method provides a two-stage identification method consisting of mobile terminal direct identification and server-assisted identification and comprising the following steps of: firstly, using a camera of the mobile terminal to directly scan the bar codes for identification, and judging whether to invoke the server function for assisted identification on the basis of the identification results, thereby ensuring the identification efficiency of the terminal bar codes and simultaneously achieving the goals of improving the identification success ratio and the stability. Therefore, the implement of the method is practical and feasible.

The invention relates to an image number character recognition method. An algorithm structure is optimized, and influences of illumination and inclination are processed in an image preprocessing stage, thereby ensuring that a better binary character image is acquired before character recognition; then, a two-dimensional PCA algorithm is adopted to extract primary characteristics in an image so as to reduce the dimension and operation complexity of the image; a decision tree algorithm and an Adaboost algorithm are combined to optimize a classification algorithm, thereby ensuring the recognition efficiency and reducing the over-fitting risk; and finally, a nine-one division method is used for verifying the generalization ability of a classification model so as to ensure the stability of a classifier. By virtue of algorithm optimization, the accuracy rate and speed of number character recognition are obviously increased.

The invention provides a micro-expression recognition method based on the main direction of optical flow, which comprises the following steps: interpolating a micro-expression sequence formed by framing a micro-expression video; positioning the eyes and nose of each frame, and calculating the position of the middle point of the eyes as the origin of the coordinate axis; calculating the distance dem and dmn from the origin of the coordinate axis of each frame to the center of each eye and the nose; constructing normalized mesh according to dem and dmn; calculating the optical flow direction ofeach pixel region in all adjacent frame grids and clustering to obtain the optical flow main direction; a rectangular characteristic region being formed in the main direction of the optical flow, andthe pixel region in the main direction of the optical flow being assigned with weight coefficients to obtain the optical flow field optimized by adjacent frame grids, and the curves formed by connecting the related data drawing points being smoothed to obtain the global optical flow field; the pixel region in the main direction of the optical flow being assigned with weight coefficients. A training classifier inputs each optical flow characteristic region and compares the sum of similarities between the traversed-level microexpression sequence and the trained microexpression sequence with S toobtain a classification result.

The invention is suitable for the technical field of computers, and provides a fraud number identification method and device, computer equipment and a storage medium. The method comprises the following steps: obtaining communication feature information of a to-be-identified number; processing the communication feature information according to a preset fraud number recognition model to generate a fraud number recognition result, wherein the preset fraud number recognition model is generated by training in advance based on a semi-supervised learning self-training classification algorithm. According to the fraud number identification method provided by the invention, the fraud number identification model is trained and generated by using the self-training classification algorithm which does not need to depend on a large amount of labeled sample data in the training process, so an optimal identification model can be obtained through training, and the method has good adaptability in the field of fraud phone identification with insufficient sample data; the fraudulent number recognition result obtained by processing the communication feature information by using the fraudulent number recognition model is high in accuracy.

The invention discloses a voiceprint recognition method and a device, computer equipment and a storage medium, and belongs to the technical field of artificial intelligence, and the method comprises the steps: obtaining a to-be-recognized voice, carrying out the framing of the to-be-recognized voice, and obtaining an audio frame set; performing Fourier transform on the audio frame to obtain sequence information of the voice data; inputting the sequence information into a pre-trained self-attention voiceprint recognition model, and obtaining a voiceprint feature matrix of the audio frame; and splicing the voiceprint feature matrixes to obtain a voiceprint recognition result of the to-be-recognized voice. In addition, the invention also relates to a blockchain technology, and the to-be-recognized voice can be stored in a blockchain. According to the method, training and prediction of the voiceprint recognition model are carried out by adopting the neural network, the voice features are mapped to the hidden spaces by adopting the self-attention mechanism to be processed respectively, and finally the processing results of all the audio frames are spliced to obtain the recognition result of the to-be-recognized voice, so that parallel operation can be realized, and the identification precision is considered while the identification efficiency is ensured.

The invention provides a method and device for identifying fraud and electronic equipment, and the method comprises the steps: identifying to-be-identified behavior data based on a classification model; when the to-be-identified behavior data is abnormal, determining an auditing result of the to-be-identified behavior data; taking the to-be-identified behavior data as a positive sample and addingthe positive sample into a preset sample set; taking the to-be-identified behavior data as historical behavior data, and extracting outlier historical behavior data corresponding to outliers in the historical behavior data based on a preset outlier detection model; and when the outlier historical behavior data belongs to fraudulent behaviors, adding the outlier historical behavior data as a positive sample into the sample set, and then training the classification model again based on the updated sample set. Through the technical scheme provided by the embodiment of the invention, the number ofsamples in the sample set is increased, the classification model can identify novel fraud behaviors in time, and the identification effect of the classification model is improved.

The invention provides a speech recognition method and system based on an end-to-end deep learning model. The method comprises a step of mapping simple or compound vowels with similar pronunciations to the same code and mapping initial consonants with similar pronunciations to the same code to form a rule mapping table, a step of performing data encoding on a corpus by using the rule mapping ruleand representing Chinese characters of the corpus by using codes in the rule mapping table, a step of training the encoded corpus by using a hybrid end-to-end model to obtain 'pinyin' and 'phoneme' acoustic models, wherein the hybrid end-to-end model comprises a 'pinyin' end-to-end model and a 'phoneme' end-to-end model, a step of encoding multiple words to be applied by using the rule mapping table to form a vocabulary library, a step of identifying speech by using the acoustic models, and a step of comparing codes outputted by the acoustic models and codes of vocabularies in the vocabulary library by using editing distances and finding a minimum editing distance, and a corresponding vocabulary is an identification result. According to the method, the recognition efficiency of the systemis improved.

The invention relates to the technical field of automatic identification, and discloses a generation method, a generation device, an identification method and an identification device of a dynamic three-dimensional code. The generation method comprises the steps: carrying out the coding according to the preset embedded information, and obtaining a three-dimensional code threshold layer; adding color information and / or gray scale information to the three-dimensional code threshold layer to obtain a visual code layer; carrying out re-coding on the visual code layer to obtain a three-dimensionalcode picture; carrying out fusion processing on the three-dimensional code picture and more than one frame of background picture to obtain a fused three-dimensional code. The information capacity is larger, and the forms are more diversified and the user experience is better on the basis of ensuring the recognition efficiency and accuracy. Particularly, a plurality of dynamic three-dimensional codes can be superposed to form a three-dimensional code group, so that the management and the use are more convenient.

The invention relates to a curved surface object corner recognizing and spraying method. The method includes that on the basis of realization of fragmenting of a complex free curved surface, aiming at an included angle between adjacent surface fragment combinations, geometric relation with a spraying flare angle of a spray gun is adopted, and a law model of influence, of the included angles between the surface fragment combinations, on extreme thickness values of a coating during vertical spraying is built to recognize corner features. By the method, nonexistence of coatings and coating waste at junctions of concave surface fragments during vertical spraying at junctions of the surface fragment combinations can be effectively avoided, so that spraying efficiency is improved while spraying quality is guaranteed.

The invention relates to the technical field of charging equipment for identification, in particular to data line anti-falling safety charging equipment for radio frequency identification payment, which comprises an equipment main body, electromagnets I are fixedly connected to the left side wall and the right side wall in the equipment main body, and a positioning frame is fixedly connected to the right side of the electromagnets I in the equipment main body; a moving plate is slidably connected to the interior of the front surface of the positioning frame, a limiting plate is movably connected to the position, located on the left side of the moving plate, of the interior of the moving plate, a data line in the middle is limited through two limiting clamping plates, and therefore the equipment is prevented from falling off in the using process, and normal power connection of the equipment can be well guaranteed; meanwhile, the identification efficiency of the equipment can be well guaranteed, and the signal lamp is turned off by separating the clamping plate from the bayonet lock below, so that a worker is reminded that the equipment goes wrong, and the situation that parts in theequipment are damaged, and the service life of the equipment is affected is avoided.

The present invention relates to an ultrasound-based voice assistant wake-up method and apparatus, a computer device and a storage medium. The method comprises the steps of receiving an ultrasonic wake-up instruction from a wake-up device; analyzing the received ultrasonic wake-up instruction to obtain a spectrum analysis result; comparing the spectrum analysis result with the spectrum of a presetultrasonic wake-up instruction to determine whether the spectrum analysis result is matched with the spectrum of the preset ultrasonic wake-up instruction or not; and if so, waking up and activatinga voice assistant, and returning confirmed activation information. According to the scheme of the present invention, an intelligent voice device in a standby state is configured to collect ultrasonicaudio, so that the intelligent voice device is prevented from monitoring the user conversation in real time during a standby state, and the user privacy can be protected; and in addition, difficulty of implementation of the ultrasonic audio recognition technology is controllable, so that compared with an existing method for using a statistic model or a deep neural network model to perform wake-upkeyword recognition, the CPU performance and memory capacity requirements are greatly reduced, and the setup cost of a recognition device is reduced while ensuring recognition accuracy and efficiency.

The invention discloses a terminal application data flow identification method and system, and the method comprises the following steps: S1, carrying out the preprocessing of a to-be-identified terminal application data flow collected in real time, and obtaining a plurality of types of data package sets after classification according to quintuple information; s2, extracting a first data feature ofthe obtained data packet set of each category, respectively matching the first data feature with the features of a plurality of preset application services, if the first data feature can be successfully matched with the features of any preset application behavior, determining that the application services to which all data packets in the data packet set of the category belong are the preset application services, and ending the operation; otherwise, turning to the step S3; S3, extracting a second data feature of the obtained data packet set of each category, inputting the second data feature into a pre-trained data stream identification model, and identifying the data stream. According to the invention, through hierarchical data flow identification, the accuracy of data flow identificationcan be greatly improved while the identification efficiency is ensured.

The embodiment of the invention discloses a training method, training equipment, a using method and using equipment of a multi-task recognition model, and a storage medium. An initial feature sharingmodel is trained through employing a sample image which comprises a plurality of face attributes with independent features, then a sample image which comprises a plurality of face attributes with coupled features is added for training, finally, a decoupling model part is added to the model, and a sample image containing a human face posture is added for training, so that the training of the multi-target task recognition model can be realized by gradually increasing human face attributes, and the sample image with the human face attributes with independent features can be fully utilized to realize feature sharing; and the decoupling effect on the coupling features is realized by adding the decoupling model, so that the recognition accuracy of the multi-task target recognition model can be effectively ensured, and the recognition efficiency can be ensured while the multi-face attribute recognition precision of the single model is ensured.

The invention provides an information recognition device, system and method, and relates to the technical field of information. The device comprises a scanning assembly and a disc type storage assembly, and the scanning assembly and the disc type storage assembly are connected with a server; the disc type storage assembly is located on one side of the scanning assembly. At least one container boxis stored in the disc type storage assembly, and each container box is provided with an electronic tag. The disc type storage assembly is used for rotating when receiving a rotation instruction from the server, so that the electronic tag on the container box in the disc type storage assembly is aligned with the scanning assembly; and the scanning assembly is used for scanning the electronic tag when receiving a scanning instruction from the server to obtain scanning information, and sending the scanning information to the server to enable the server to identify the scanning information. According to the invention, the accuracy of information recognition is effectively improved while the recognition efficiency is ensured.

The invention provides a method and a system for constructing an efficient quality inspection model of a self-learning and updating training sample, wherein two SqueezeNet Pro supervised learning models are adopted, fast k. The means algorithm acts as a preselected tag; The loss function of each model is optimized by comparing the two data enhancement algorithms. Conducting self-learning and training models; And update the weights of the parent model. The invention effectively improves the identification efficiency or the identification accuracy, improves the speed of the algorithm calculation, does not increase the time complexity of the operation, and can effectively prevent the occurrence of the local optimal situation.

The invention provides an industrial production data entity identification method based on Merkley-tree. The method comprises the following steps of: 1) for the floatability of industrial production data, performing corresponding standardized processing on the data by utilizing the properties of a matrix and a vector to ensure that numerical attribute values of the same entity are the same; 2) calculating the information entropy of each attribute column, obtaining attribute sensitivity information, removing attributes with low sensitivity, and sequencing the rest attributes in a descending order according to the sensitivity; 3) proposing a chain structure called St-Chain; performing progressive Hash coding based on the St-Chain , and dividing the entities with the same Hash value into thesame block; and 4) for the structure obtained in the step 3), continuing to calculate the hash value of the subsequent attribute in each tuple, and repeatedly dividing the tuples into blocks accordingto the difference of the hash values to finally obtain an entity identification result, thereby providing an entity identification method with moderate algorithm operation efficiency and high identification precision.

The invention relates to the technical field of stage scene adjusting systems, and discloses an intelligent scene adjusting system capable of changing along with forms of stage actors, which comprises a stage scene module, limb induction, sole induction, hand induction, leg induction, stage contacts, vibration feedback, a communication unit, a monitoring center CPU, sound equipment and lighting equipment. According to the intelligent scene adjusting system capable of changing along with the form of the stage actor, the digital twinborn technology is utilized, the variation of gravity change, time, position, music and the like transmitted by the actor body through the communication unit 1 is transmitted to the monitoring center cpu, and the variation is matched and compared with data stored in a database of the monitoring center cpu, so that the scene can be automatically adjusted. When the variation of gravity change, time, position, music and the like transmitted by the communication unit 1 is matched with the data in the database, the monitoring center cpu controls the sound equipment 1, the lighting equipment 1, the lightening unit 1 and the light-emitting element 1 to perform corresponding actions.

The invention provides an image content identification method and device, relates to the image identification technical field, and solves the image content identification problems in the machine vision field; the method comprises the following steps: receiving a to be indentified image; identifying the image according to an image identification model, wherein the image identification model comprises a first feature set having content information used for identifying image content; outputting an image identification result. The image content identification method and device can be applied to image content identification equipment having high accuracy requirements for image content identification.

The embodiment of the invention provides a method and a device for real-time positioning of personnel, which are applied to an identification system comprising a plurality of position points. The method comprises the following steps of: acquiring real-time images of each position point; Identifying the image based on a preset database to determine an object corresponding to the image, wherein thedatabase stores a correspondence relationship between the human face features and the identity information of the object; updating a position of the object to a position of the position point. In theway of active image recognition, the recognizer does not need to carry other equipment, interference is not liable to occur, large-scale infrastructure foundation is not needed, there is no need to carry on the large-scale transformation to the original environment, the cost is saved, the efficiency of recognition is ensured, and the adaptive range is expanded.

Embodiments of the invention provide a method and a device for intercepting a multimedia message in a mobile device. The method comprises steps of extracting multimedia data from a multimedia message when it is monitored that the mobile device receives the multimedia message; determining the multimedia message is illegal or not according to multimedia data; and if yes, intercepting the multimedia message. Automatic adaption to change in illegal multimedia message is achieved, and the interception efficiency and the interception accuracy are improved.

The invention relates to the technical field of backplane bus communication, in particular to a communication system and method based on an SPI bus. According to the system and the method, a master station and at least one slave station connected with the master station in a daisy chain mode are adopted; the master station is used for sending a starting frame to the slave station; the slave station is used for generating and outputting a response frame corresponding to the starting frame after receiving the starting frame; the master station is further used for confirming the number of the slave stations based on all the response frames when the master station receives the starting frame and all the response frames, the communication system and method based on the SPI bus can achieve a slave station feedback mechanism, and the recognition efficiency and accuracy of the number of the connected slave stations are guaranteed.

The invention relates to the technical field of image processing, and discloses a slide cell scanning method based on a visual target marker, which comprises the following steps of: performing leukocyte identification on a 20-time objective lens based on cell color, and converting slide step numbers of a high-power lens and a low-power lens after the identification is completed; carrying out region division on the picture obtained by the 20-time objective lens, and carrying out reset scanning and cell calibration on the white blood cells by utilizing a high-power lens; if the set number of cells are not obtained, a conventional process is switched to carry out cell acquisition until the set number of cells are obtained, and meanwhile, in order to ensure the determination of the position before shooting, the debugging of the starting point is carried out in advance, so that the problem that the white blood cell position cannot be accurately found in the conventional slide cell scanning process is solved; and therefore, the identification efficiency can be ensured.

A method and device for identifying content on a target, a video acquisition method and device, a storage medium, and computer equipment, the method comprising: acquiring a part of images in an initial video stream as initial images, the initial video stream comprising multiple frames of images arranged according to a shooting time sequence; performing target identification on the initial image frame by frame according to a shooting time sequence, wherein the step of performing target identification on each frame of initial image comprises the following steps: carrying out target detection on each frame of initial image to obtain an image containing a to-be-identified target; identifying a target area where content is located from an image containing a to-be-identified target, wherein the content is located on the to-be-identified target; and performing content identification on the target area to obtain identification information. Through the scheme, the identification efficiency can be improved while the accuracy of the identification result is ensured.

The invention discloses a fingerprint image preprocessing method and device. The method includes the steps that a received fingerprint image is smoothed so that a first image can be obtained; the first image is processed through first enhancement so that a second image which is locally enhanced can be obtained; the second image is normalized so that a third image can be obtained; the fingerprint line direction angle and the fingerprint line frequency of the third image are calculated; the third image is processed through second enhancement so that a fourth image which is locally enhanced can be obtained; the fourth image is normalized so that a preprocessed output image can be obtained. According to the method and device, because the received fingerprint image is enhanced twice, the image quality of the fingerprint image enhanced twice can be high, the fidelity factor of the fingerprint line direction and the fingerprint line frequency of the preprocessed fingerprint image is high, and the identification accuracy and the identification efficiency of subsequent fingerprint recognition applications can be conveniently guaranteed.

The invention relates to a glass curtain wall crack detection method. The method comprises the following steps: step (1), acquiring a glass curtain wall image; (2) converting the glass curtain wall image into a gray level image; (3) calculating a first feature of the grayscale image based on an improved ULBP descriptor; step (4), carrying out dimension reduction on the first feature of the grayscale image to obtain a macro feature; and step (5), inputting the macro features into a classifier, and detecting whether cracks exist in the glass curtain wall image through the classifier. According to the invention, the crack of the glass curtain wall can be effectively detected, and the detection precision is high.

The invention provides a finger vein recognition method and system fusing basic features and deformation features. The recognition method mainly comprises the steps of image preprocessing, basic feature extraction, objective function construction and optimization and matching, deformation information extraction and basic feature and deformation feature fusion. According to the method, in the recognition process, deformation information is utilized, basic characteristics, matching scores and other information needed when the deformation characteristics are obtained are considered, and a fingervein recognition framework integrating the basic characteristics and the deformation characteristics is constructed. According to the method, the problem that when a single feature is used, the feature representation ability is insufficient is solved, the accuracy and robustness of finger vein recognition are improved, and the recognition efficiency is ensured.