422results about How to "Easy to understand implementation" patented technology

The invention provides a three-dimensional reconstruction-based unmanned aerial vehicle image stitching method. The method comprises the following steps of: obtaining unordered images shot by an unmanned aerial vehicle camera in a shooting area; carrying out feature point extraction on the obtained unordered images, carrying out feature matching on the extracted feature points to obtain matched feature point pairs; and sorting the unordered images according to the matched feature points and a sequence; analyzing the matched feature points so as to obtain camera attitude parameters, distortionparameters and space coordinates of the feature points; carrying out image clustering on the arranged images according to the camera attitude parameters and the space coordinates of the feature pointsso as to form a plurality of image clusters, respectively extracting dense point clouds of the image clusters, and completing dense matching; and carrying out ortho-rectification on the dense matchedimages according to the dense point clouds, the camera attitude parameters and the distortion parameters, and stitching the ortho-rectified images to obtain a stitched image. According to the method,the stitching speed is improved and projection distortion is not caused. The invention furthermore provides an unmanned aerial vehicle image stitching system.

The embodiments of the invention provides a defensive driving strategy generation method, device and equipment and a storage medium. The method includes the following steps: detecting types of obstacles in a perception range of an autonomous vehicle; determining whether collision risk exists between the autonomous vehicle and the obstacles by using collision risk detection modes corresponding to the types; and when the obstacles having the collision risk exist in the perception range, determining a defensive driving strategy of the autonomous vehicle. The defensive driving strategy generationmethod, device and equipment and the storage medium of the embodiments of the invention can reduce the probability of occurrence of road hazard.

The invention provides a video review method and device and a server. The method comprises the following steps of intercepting a multimedia fragment of a to-be-reviewed video according to a scene in the to-be-reviewed video; detecting whether a content of the multimedia fragment belongs to a prohibited category or not; and if yes, sending prompt information of the to-be-reviewed video to the prohibited category, wherein the multimedia fragment comprises an image frame, an audio fragment, a speech fragment and/or a text fragment. The method can be realized by intercepting the multimedia fragment of the to-be-reviewed video according to the scene in the to-be-reviewed video and detecting whether the content of the multimedia fragment belongs to the prohibited category or not, wherein the multimedia fragment comprises the image frame, the audio fragment, the speech fragment and a character fragment, the to-be-reviewed video is cut into the multimedia fragment according to the scene, and when one fragment has relatively strong characteristic performance, the to-be-reviewed video can be determined to belong to the prohibited category, and the technical scheme can greatly improve the review efficiency under a condition that the video is longer.

The invention provides a method and a device for determining video tags, a storage medium and a terminal device. The method includes the steps: determining an acquisition mode of an associative text of a video to be processed according to the field of the video to be processed, and extracting the associative text of the video to be processed; extracting candidate tags of the video to be processedfrom the associative text of the video to be processed; performing ordering on the candidate tags; selecting tags consistent with the video to be processed from the candidate tags according to an ordering result. The description accuracy of the video tags is improved by the method.

The invention provides a method, a device, a storage medium and a terminal device for detecting a line-of-sight point, wherein, the method comprises the following steps: obtaining a face image when anobserver looks at a display screen; extracting feature points of a face region and an eye region from the face image; tracking an initial picture sequence of the face image, iteratively calculating afeature point of the face region to obtain a head posture of the observer; determining a line-of-sight angle and a line-of-sight confidence parameter of the observer according to a characteristic point of the eye region; according to the head posture, the line-of-sight angle and the line-of-sight confidence parameter, and the distance of the observer to the display screen, determining the line-of-sight of the observer to be at a point-of-arrival position of the display screen. By adopting the invention, the position of the line of sight falling point of the observer can be conveniently, quickly and accurately determined.

The invention discloses a method, device, storage medium and terminal device for product defect detections and locations, wherein the method includes: according to the image data characteristics of products to be tested and the product defect complexity, selecting the neural network structure of detection models to be trained; obtaining training data, wherein the training data include training product pictures and standard categories and standard positions of product defects in the training product pictures; according to the training data, training the selected detection models of each neuralnetwork structure; selecting a best fitting detection model from the detection models after training to provide a detection server with the product defect detection, wherein the detection models are used for calculating according to the received product pictures to obtain predicted categories and predicted locations of the product defects in the received product pictures. Adopting the method, device, storage medium and terminal device for product defect detections and locations is capable of improving the accuracy of detecting the product defects.

The embodiment of the invention provides a multi-tone area speech recognition method and device, and a computer-readable storage medium. The multi-tone area speech recognition method comprises the steps: collecting speech in a speech acquisition environment using a microphone module, wherein a target tone area is set for each microphone in the microphone module in advance, the target tone area being one of a plurality of independent tone areas in the speech acquisition environment set in advance; Judging whether the speech is from a target sound region; Suppressing speech from a sound region other than the target sound region; the suppressed speech is recognized. As the embodiment of the invention use a microphone module to isolate different sound regions, the voice use by all the staff inthe voice acquisition environment is realized by isolating a plurality of sound regions, and the isolation degree of each sound region can reach 15-20dB, the ideal isolation effect is achieved.

The invention provides a vehicle diagnosis method, device and system, equipment and a computer readable storage medium. The method comprises the following steps: obtaining a vehicle diagnosis requestsent by a terminal, the vehicle diagnosis request comprising the vehicle information of a target vehicle; obtaining a test task corresponding to the target vehicle according to the vehicle information; and sending the test task to the target vehicle, so that the target vehicle analyzes the test task and sends the test task to a corresponding to-be-diagnosed object on the target vehicle to executea test. According to the embodiment of the invention, the test task needing to be executed by the to-be-diagnosed object of the target vehicle can be quickly and accurately obtained through the vehicle information. Moreover, since the analysis process of the test task is carried out at the vehicle end, the stability of the diagnosis test and the efficiency of sending the test task to the vehicle are effectively improved, and the task processing cost is saved.

The invention discloses a volume adjusting method applied to an audio playing device. The method comprises the collecting step, wherein surrounding environment sound is collected; the filtering step,wherein playing sound of the audio playing device is filtered out to obtain a surrounding environment noise value; the adjusting step, wherein the playing volume value is adjusted. By dynamically monitoring noise of the outside environment, the volume is automatically adjusted, then users do not need to manually adjust the volume, and the user experience can be improved.

The embodiment of the invention provides a block chain intelligent contract verification method, a device and a storage medium. The method comprises the following steps: obtaining block header information and transaction list of a designated block from a first node in a block chain network, the transaction list comprising a transaction identification and an execution result of an intelligent contract, the block header information comprising a root node of a first Merkel tree, a root node of a second Merkel tree and a last block identification of the designated block; obtaining a block header information and a transaction list of the designated block from the first node in the block chain network. Verifying a root node and a transaction identifier of the first Merkel tree; The root node ofthe second Merkel tree and the execution result of the intelligent contract are verified; Verifying whether the previous block identification of the specified block is in a pre-stored block header chain structure; If the above verification passes, it is determined that the execution result of the smart contract is valid. The embodiment of the invention can judge the validity of the execution result of the intelligent contract, save resources and improve efficiency without downloading the full amount of block information and repeatedly executing the intelligent contract in the lightweight nodelocally.

The embodiment of the invention provides a storage device and system and a control method. The storage device comprises multiple goods shelves, a first robot channel and a second robot channel. Each goods shelf comprises at least one temporary storage laminate, at least one storage laminate and multiple stand columns arranged in the horizontal direction at intervals. The storage laminates and thetemporary storage laminates are arranged in the vertical direction at intervals through the stand columns. The temporary storage laminates are used for providing multiple temporary storage stations, and the storage laminates are used for providing multiple storage stations. The first robot channel is used for allowing a first robot to run, and the first robot is used for storing goods on the temporary storage laminates. The second robot channel is used for allowing a second robot to run, and the second robot is used for carrying goods between the temporary storage laminates and the storage laminates. By means of the technical scheme, the efficiency of putting cargoes in and out of warehouses can be improved.

The embodiment of the invention provides a model training method and device and a recommendation information selection method and device. The model training method comprises the steps: obtaining the scene characteristics and a historical recommendation list of a historical user, wherein the historical recommendation list comprises a plurality of historical recommendation items which are arranged according to a time sequence, and the real feedback values of all the historical recommendation items; associating the scene feature, the real feedback value of the first historical recommendation item and the second historical recommendation item to construct a training sample, wherein the second historical recommendation item is located in the next time sequence of the first historical recommendation item; and training the initial model by using a plurality of training samples to obtain a prediction model which is used for obtaining a prediction feedback value of the recommendation item. According to the embodiment of the invention, the real feedback value of the previous time sequence historical recommendation item and the next time sequence historical recommendation item are used as the training sample training model; therefore, rapid adaptive learning can be realized based on feedback of the previous time sequence, and a model for accurately obtaining the prediction feedback value of the recommendation item can still be obtained under the condition that the training sample size is small.

The embodiment of the invention provides a speed planning method and apparatus for an autonomous vehicle. The method comprises: before lane changing of an autonomous vehicle, a plurality of to-be-selected speed trajectories are generated for the autonomous vehicle, wherein the to-be-selected speed trajectories include to-be-selected speeds corresponding to a plurality of time points within a planned time range; a predicted speed and a predicted position corresponding to an obstacle at a plurality of time points are obtained; according to the to-be-selected speed trajectories as well as the predicted speed and the predicted position, a cost function value of each of the to-be-selected speed trajectories is calculated; and the to-be-selected speed trajectory with the minimum cost function value is selected as a planned speed trajectory of the autonomous vehicle. Therefore, the speed of the autonomous vehicle before lane changing can be adjusted and the distance between the autonomous vehicle and the obstacle of a target lane is increased to create a safety distance between the lane changing opportunity and the lane changing, so that the lane changing success rate is increased.

Embodiments of the invention provide a dialogue interaction method, a dialogue interaction device, and a computer readable storage medium. The method comprises the steps of converting to-be-recognizedvoice into a first text; inputting the first text into a semantic analysis model to obtain intention information and word slot information of the first text; and inputting the intention information and the word slot information of the first text into a dialogue state machine to obtain interaction information corresponding to the first text. According to the embodiment of the invention, the intention information and the word slot information of the first text are directly obtained from the first text by utilizing the semantic analysis model. The step that the semantic analysis model needs to be connected behind the language model in the prior art is omitted, so that the time is shortened. The user can be responded more quickly. Meanwhile, according to the scheme, the calculation complexityis reduced, and the cost of the whole system is reduced.

The embodiment of the invention provides a workstation for a cleaning robot, and the cleaning robot. The workstation for the cleaning robot comprises: a sewage recycling device which is used for recycling sewage from a sewage box of the cleaning robot; and a clear water supply device used for injecting clear water into the clear water box of the cleaning robot. According to the technical scheme, clear water can be automatically injected into the clear water box of the cleaning robot, and sewage in the sewage box of the cleaning robot can be automatically recycled and cleaned, so manual labor is reduced, and user experience is improved.

The embodiment of the invention provides a training method and device for a semantic relation recognition model and a terminal, and the method comprises the steps: inputting a sample data set into aninitial pre-training model, outputting the representation information of sample sentences, and enabling the sample data set to comprise a plurality of sample semantic units; obtaining a plurality of feature words, and splicing the plurality of feature words to obtain representation information of the spliced feature words; inputting the representation information of the sample sentences and the representation information of the splicing feature words into an initial classifier, and outputting semantic relationship categories among the sample semantic units; adjusting the initial pre-training model and the initial classifier to obtain a new pre-training model and a new classifier; and establishing a semantic relation recognition model according to the new pre-training model and the new classifier. And the feature words are used as strong features in the chapter relationship, so that the classification effect on the specific semantic relationship can be improved. When the semantic relation recognition model is used for predicting the semantic relation category, the prediction time is shortened, and the prediction efficiency is improved.

The invention provides a method and a device for updating a product defect detection model, a storage medium and terminal equipment. The method comprises the following steps: receiving a first detection model generated by training; replacing the second detection model in a detection system with the first detection model, and making the ratio of the number of servers running the first detection model to the number of servers running the second detection model in the detection system after replacement conform to a replacement proportion; calculating the detection accuracy of the detection systembefore replacement and the detection accuracy of the detection system after replacement; and if the detection accuracy of the detection system after replacement is higher than the detection accuracyof the detection system before replacement and the difference is greater than an adjustment threshold, adjusting the replacement proportion based on a preset proportional step size increment, replacing the second detection model in the detection system with the first detection model, and making the ratio of the number of servers running the first detection model to the number of servers running the second detection model in the detection system after replacement conform to the replacement proportion after incremental adjustment. According to the invention, the applicability and detection accuracy of the detection model can be improved.

The embodiment of the invention provides a single sign-on method and device and a computer readable storage medium. The method comprises the steps that a single sign-on system receives an authentication request from a service system, the authentication request comprises a session primary key of the service system and a token of a client side, and the token is generated after the single sign-on system verifies that login information of the client side is legal; the single sign-on system authenticates a session primary key of the service system and the token; and under the condition that the authentication is passed, the service system is controlled to execute the service request of the client. And if the client accesses the service system and the service system carrier jumps to the single-point service system, the single-point service system indicates that the login information of the client is authenticated under the condition that the authentication session main key and the token arepassed. Therefore, the safety is further improved.

The invention provides a method, an apparatus, a storage medium and a terminal device for determining a line-of-sight area, wherein the method comprises the following steps of: acquiring a driver's face image; extracting feature points of a face region and an eye region of the driver from the face image; tracking an initial picture sequence of the face images to iteratively calculate feature points of the face region to obtain a head posture of the driver; estimating a classification result of the line-of-sight region observed by the driver based on a characteristic point of the eye region; and correcting the classification result of the line-of-sight area according to the head posture of the driver, and using the corrected line-of-sight area as the line-of-sight area of the driver. By adopting the invention, the line of sight area of the driver can be accurately obtained.

The present invention provides a method and a device for parking prediction of a parking lot, a storage medium and a terminal device. The method comprises the steps of: receiving a parking query request of a user terminal, wherein the parking query request comprises a parking destination and parking time of a user; performing statistics of the query quantity of query requests received at present and being the same as the parking query request; obtaining parking lot information meeting the parking destination from a navigation system; performing calculation of the query quantity and the parkinglot information through a parking prediction model to obtain a parking prediction result; and returning the parking prediction result to the user terminal. According to the invention, the operation and maintenance cost are low, the applicability is wide and the prediction accuracy is high.

The present invention relates to a dynamic brain function detection method and system synchronous with transcranial magnetic stimulation. The system comprises a transcranial magnetic stimulation device for magnetically stimulating a targeted brain region of a subject to achieve a purpose of activating or inhibiting functions of a corresponding brain region; a functional near-infrared spectrometerfor collecting brain blood oxygen signals of the targeted brain region of the subject and other function connection related brain blood oxygen signals; an analysis module for analyzing blood oxygen signals of multiple-frequency bands of the brain blood oxygen signals; and a synchronous marking module for synchronously marking a stimulation moment in near infrared time domain signals when the transcranial magnetic stimulation (TMS) device starts to perform magnetic stimulation so as to instantly collect changes of the blood oxygen signals when the magnetic stimulation starts. The method provides a feasible scheme for researching brain disorder and cortical changes caused by the TMS and can also objectively measure an effect of the transcranial magnetic stimulation on different regions of abrain on a premise that the effect of the transcranial magnetic stimulation on the different regions of the brain cannot be quantified.

The invention provides an intelligent monitoring method and device for infant monitoring, and a server. The method comprises the following steps: collecting state information of an infant, wherein thestate information comprises cry frequency, cry decibel, ambient sound decibel, body temperature, environment temperature, pulse and geographical location of the infant; analyzing the collected stateinformation of the infant according to an analytic model, and acquiring an analysis result of the analytic model; and sending alarm information to a curator when the analysis result output by the analytic model displays that the state of the infant is abnormal. The method provided by the embodiment of the invention can realize the following situations: the technical scheme provided by the embodiment of the invention analyzes the collected state information of the infant according to the analytic model, and sends the alarm information to the curator when the analysis result output by the analytic model displays that the state of the infant is abnormal. The monitoring on the infant state can be realized through the technical scheme provided by the invention, thereby timey helping the infantto recover the good state when the infant is in abnormal state.

The invention discloses an electric switching monitoring method, a device, equipment and a computer readable medium. The method comprises the steps: generating a switching operation order of an electric system according to an operation task to be executed, wherein the switching operation order comprises operation steps and an execution sequence of the operation steps; sequentially outputting voiceoperation instructions corresponding to all the operation steps according to the operation steps and the execution sequence of the operation steps in the switching operation order; utilizing operational motion images aiming at the voice operation instructions; judging whether the operational motion images can meet requirements of corresponding operation steps. According to the electric switchingmonitoring method disclosed by the embodiment of the invention, the electric switching operation efficiency of a data center is improved; in addition, reliability in the operation steps is improved, and the human factor risk is reduced; meanwhile, safety in the operation process is improved.

The invention provides a photographing method based on augmented reality. The method includes a photographing step of photographing a real-world image; an information acquisition step of acquiring identification information and real-world action information according to the real-world image; a matching step of matching an interactive object from an interactive information base according to the identification information and matching interactive actions of the interactive object from the interactive information base according to the real-world action; and an image generating step of merging thereal-world image, the interactive object, and the interactive action of the interactive object to generate an augmented reality-based image. The technical scheme of the invention can meet the needs of a user to have a photo or interaction with stars or other characters, animals, or articles. In addition, the invention also provides a photographing apparatus based on augmented reality, a photographing device and a computer storage medium.

The embodiment of the invention presents a driving planning method and device based on a road side unit, a road side unit and a storage medium. The method comprises the following steps: determining apre-selected path for a target vehicle according to the vehicle information of the target vehicle, wherein the target vehicle is a vehicle within the communication range of a road side unit; acquiringobstacle information corresponding to the pre-selected path according to the sensing data of sensors, and generating a dynamic map for the target vehicle according to the pre-selected path and the obstacle information; carrying out path planning and speed planning for the target vehicle according to the dynamic map to generate planned trajectory information; and determining the target trajectoryinformation of the target vehicle according to the planned trajectory information. According to the method in the embodiment of the invention, the path and speed of the target vehicle are planned based on the road side unit, and the information of obstacles such as pedestrians is considered in the planning process and added to the dynamic map, so that automatic driving of the target vehicle can berealized or driving decision service can be provided for the driver of the target vehicle.

The invention provides an automatic testing method and device and a server. The method comprises the steps that operating data of an online environment is collected; Mock simulation operation is performed according to first request information of a first preset type in the operating data to obtain first return information, corresponding to the first request information, in the online environment;and the first request information and the first return information are utilized to perform regression testing on an offline testing module corresponding to the online environment. Through the technical scheme, first, the operating data of the online environment is collected and used for regression testing, Mock simulation operation is performed on the first request information of the first presettype of return data which cannot be normally collected, so that the first return information, corresponding to the first request information, in the online environment is obtained, and then the operating data is utilized to perform regression testing on the offline testing module corresponding to the online environment. In this way, testing cost can be reduced, and meanwhile code invasion to online data is avoided.

The embodiment of the invention provides a user behavior information processing method and device based on a block chain. The method comprises the following steps: receiving behavior information of atarget user from each user behavior information providing node; correspondingly storing the behavior information of the target user and the identification information of the target user into an accounting node of a block chain; obtaining a target user node corresponding to the identification information of the target user; and sending the behavior information of the target user to the target usernode from the accounting node. According to the embodiment of the invention, the behavior information of the user from each user behavior information providing node is stored in the block chain; according to the method and the device, the corresponding user behavior information is issued to the user node, and the user behavior information is persistently stored and managed by utilizing the uniqueness, tamper resistance and traceability of the block chain technology, so that the integrity and authenticity of the information can be ensured, and a user can conveniently know various behavior conditions of himself/herself.

The invention provides an entity question answering method and device based on a neural network and a terminal, wherein the method comprises converting a word contained in a question and a candidate document into a word vector, and generating a corresponding question word vector sequence and a candidate document word vector sequence; inputting the problem word vector sequence and the candidate document word vector sequence into the long-short-term memory network model, and outputting the word coding sequence of the problem and the candidate document; matching the word encoding sequence of theproblem and the word encoding sequence of the candidate document to generate a candidate document representation based on matching information, wherein the candidate document representation comprisesa plurality of word representations; selecting start and end words from all word representations and generating entity answers based on the start and end words. The method reduces explicit computationand cumulative error, effectively utilizes semantic representation between questions and documents, and improves positioning accuracy of entity answers.

The invention provides an anti-peeping method and device for a display interface, a storage medium and a terminal device. The method includes the steps of in response to startup of an application in the display interface, turning on a front-mounted camera to obtain an image; obtaining positions of the human face pupils in a preset visual area according to the obtained image; according to the obtained positions of the human face pupils, determining whether or not a user corresponding to the human face pupils is a user who peeps a screen; when the user corresponding to the human face pupils is the user who peeps the screen, determining whether or not changes of the human face corresponding to the human face pupils meet peeping standards; if yes, determining that there is a peeping behavior;if not, determining that there is no peeping behaviors; when it is determined that there is a peeping behavior, controlling the application to be displayed in an anti-peeping mode. By adopting the anti-peeping method and device for the display interface, the storage medium and the terminal device, the accuracy of discriminating the peeping behaviors in a densely populated area is effectively improved, and the probability of misjudgment is decreased.

The invention provides a distributed database test method. The method comprises the steps of generating test data; executing abnormal simulation to generate an abnormal environment; performing corresponding operations on multiple database instances of a distributed database at the same time according to a data test model and the test data in the abnormal environment; and performing verification onservice availability and data correctness. The distributed database supporting multi-point concurrent data reading/writing can be subjected to the service availability and data correctness verification under the condition that a cluster is abnormal, so that concurrent verification of the database instances is realized and the blank that a single database instance only can be verified at present is made up for; and cluster operation and an abnormal simulation function are integrated, so that full automation of abnormal testing can be realized, and the manpower cost of test personnel is greatlyreduced.