95results about How to "Reduce order of magnitude" patented technology

The invention provides an urban rail transit panoramic monitoring video fault detection method based on depth learning. The method comprises a data set construction process, a model training generation process and an image classification recognition process. The data set construction process processes a definition abnormity video, a colour cast abnormity video and a normal video in an urban rail transit panoramic monitoring video. A training set and a test set are classified. The model training generation process comprises model training and model test. The model training is to train a fault video image recognition model based on a convolution neural network. The convolutional neural network comprises a plurality of convolution layers and a plurality of full connection layers. The model test is to calculate the test accuracy. If expectation is not fulfilled, the fault video image recognition model is optimized. The image classification recognition process comprises the steps that a single-frame image to be recognized is input into the model, and the fault video image recognition model outputs an image classification result to complete the fault image detection of the urban rail transit panoramic monitoring video.

The invention discloses a terahertz dangerous article detection method based on depth learning. The terahertz dangerous article detection method comprises steps that a dangerous article sample image database is established, and images are processed to be gray level images, which have the same size suitable for training and testing; a CNN neural network model is trained, and a final network model is generated and tested; dangerous article detection is carried out, and terahertz equipment is used to acquire terahertz images of to-be-detected objects are acquired, and the CNN neural network modelis used to detect the terahertz images of the acquired to-be-detected objects, and then detection results are acquired; and at the same time, the terahertz images of the acquired to-be-detected objects are added to the dangerous article sample image database. The CNN neural network is directly used for the training and the learning of the sample images, and complexity of selection of network parameters is reduced, and the obvious characteristics of the sample image data can be directly learned, and therefore problems of image classification and mode identification can be solved; working efficiency of security staff can be improved, workload of workers can be reduced, and the abovementioned method and the abovementioned device are suitable for security check of large flow of people.

The invention discloses a content-based robust image copy detection method, which has the following steps: to extract the feature vector of the testing image; to select the integral DCT transform coefficients of a plane Y of the original image YCbCr color space, to calculate the order measurement of the coefficients to obtain a coefficient sequence which is used as the feature vector of the testing image; then to establish testing image representative vector libraries; to execute clustering analysis to the feature vector sets of the testing image representative vector libraries to select the feature vectors closest to the clustering center as the clustering representatives to constitute a clustering representative vector library; to respectively search the matched testing image representative vector libraries to inquire the image feature vectors and the image feature vectors after rotation compensation, and to determine the belonged class; then to execute sequential matching search to each image feature vector in the clustering, and to determine whether the copy of the inquired image exists. The invention shows a higher robustness to help improve the inquiring efficiency, and has practical value as well as wide application in digital image database arrangement, digital image copyright protection and piracy tracking.

The invention discloses a user interest learning-based vertical search method. The method specifically comprises the steps of determining interests and preferences of a user according to attribute contents aiming at personal interest demands of the user and filled by the user during registration, and building an initial user interest model; if the user performs screening query according to attribute types, adding attribute type query records to the user personal interest model to finish model updating; calculating interest degree values of the user to all physical resources, and then performing sorting and storage according to the interest degree values from high to low; for a personalized search, performing screening query according to keywords or attributes selected by the user to generate an initial query result list; and for physical resources in the initial query result list, in combination with the interest degree values of the user to the physical resources, sorting the physical resources in the list according to the interest degree values from high to low to generate a final query result list. Through user interest learning, a result consistent with a current interest is actively provided for the user.

The invention provides a deacidification and dehydration method of 3-chloropropene. 3-chloropropene is subjected to absorption treatments of a drying agent A and a drying agent B in sequence, wherein the drying agent A is one or more of anhydrous cupric sulfate, clay, alumina and soda lime, and the component B is one or more of soda lime, molecular sieve, montmorillonite and silica gel. In an absorption column, 3-chloropropene is subjected to chemical and physical absorption with the drying agents so as to reduce the contents of water and acid. The deacidification and dehydration method of 3-chloropropene does not bring in new impurities to 3-chloropropene, and affects the product quality.

The invention provides a big data storage and search method and apparatus. The storage method comprises the steps of obtaining original logs of big data and analyzing specific log content of the original logs; classifying the original logs according to the specific log content, generating a document by a specified number of original log sets, and establishing a mapping relationship between the document and a segmented word, wherein the segmented word is matched with the specific log content of the document; combining documents to generate a combined file, wherein a plurality of segmented words corresponding to the number of combined documents can be provided when the file is searched; and storing the file, instead of the original log, in a distributed storage system framework. With the adoption of the method and apparatus, the data redundancy can be greatly reduced, so that the waste of server resources is reduced and the utilization rate of storage resources is increased.

The invention provides a method and system for processing error prompt messages. The method comprises the following steps: receiving error codes and scene identifiers; positioning configuration prompt messages corresponding to the received error codes and scene identifiers; and if the positioning is successful, setting error prompt messages as the positioned configuration prompt messages. Through implementing the method and system provided by the invention, error prompt messages can be set on the basis of the error codes and the scene identifiers, so that the same error code can correspond to different error prompt messages and then the order of magnitude of the error codes is decreased.

The invention is to provide a method and device for determining black margin position information of a video. The method specifically comprises the following steps: cutting out at least one original video image from a video to be processed; carrying out boundary sampling processing on the original video image to obtain a boundary sampling image corresponding to the video; determining first black margin position information corresponding to the boundary sampling image; and determining second black margin position information corresponding to the video according to the first black margin position information. Compared with the prior art, the method carries out boundary sampling processing on the original video image captured from the video to obtain the corresponding boundary sampling image, so that the second black margin position information corresponding to the video is determined according to the determined first black margin position information of the boundary sampling image; and the black margin position of the video is determined by determining the black margin position of the boundary sampling image, so that time complexity of the processing is reduced, detection efficiency and detection accuracy of the black margin position are improved, and detection robustness is enhanced.

The invention discloses a variational Bayesian parameter learning method based harmonic current detection algorithm for an electric vehicle charging station. The algorithm includes establishing a charger equivalent circuit model, and judging whether a charging pile can be randomly accessed into a power distribution network system by utilizing a random number algorithm and a comparison module; performing ideal superposition calculation on the harmonic current meeting gaussian normal distribution so that an ideal harmonic superposition coefficient calculation method can be obtained; and performing sampling on harmonic phase to form two sets of random set of state space and measurement space, obtaining model parameters through the lower bound of a logarithmic edge likelihood function by usinga variational Bayesian parameter learning method, continuously maximizing the lower bound of the likelihood function, and iteratively updating variational phase parameters until approximate distribution approaches the true posterior distribution of the parameters so that harmonic phase superposition detection can be realized. Actual harmonic superposition coefficients can be obtained by substituting harmonic phase distribution into the harmonic superposition coefficient calculation method, so that accurate detection of multi-harmonic current superposition at same time in a charging station can be realized.

The invention discloses a method for measuring a liquid viscosity by utilization of a capillary and a device used for the method. The method is as follows: at a certain temperature, two ends of a capillary liquid are applied with pressure, a liquid interfacial movement rate is measured, a v(delta P) relation straight line is constructed, a gradient of the straight line is obtained, and the liquid viscosity is calculated according to a known capillary radius and a known liquid column length. Through the method, a viscosity of a single liquid can be measured. When two immiscible liquids are coexistent and the viscosity of one liquid is known, the viscosity of the other liquid can be measured through the method. The capillary employed for measurement in the method is in a micrometer level, and the viscosity scope of a liquid to be measured is 5*10<-4>-10Pa.s. The measuring device in the measuring method is simple in structure and is reasonable in design, parameters are convenient to control and results are reliable.

The invention discloses a photon number resolution enhanced laser communication system and method. The system comprises a controller, a waveform generator, a pulse laser source, an adjustable attenuator, an array superconducting nanowire single-photon detector, a power amplifier, a power synthesizer, an analog-to-digital converter and an oscilloscope, wherein the controller carries out code modulation on data needing to be output and inputs the data to a waveform generator, the waveform generator converts the coded information into an electric pulse signal and outputs the electric pulse signalto the pulse laser source, the pulse laser source generates a corresponding optical pulse signal according to the input electric pulse signal, then the optical pulse signal is input into the array superconducting nanowire single-photon detector through the adjustable attenuator, a signal with photon number resolution is output through the power amplifier and the power synthesizer, and finally thesignal is received and collected by the oscilloscope after being shaped by the analog-to-digital converter. Error code interference caused by external optical space noise and dark noise of the superconducting detector is greatly reduced, so that the error rate is enabled to be close to the self-owned light quantum error code limit of optical communication.

The invention relates to a satellite measurement and control task scheduling method, device and equipment based on hierarchical optimization, and the method comprises the steps: building a medium and high orbit satellite resource demand object set, an equipment capability object set and a satellite-to-equipment visible forecast set of a unified model based on a resource use application; performing preprocessing and decoupling association constraint on the demand object set to generate a new demand object set of each scheduling hierarchy; according to the new demand object set, the equipment capability object set and the visible forecast set, designed sub-scheduling layers are sequentially and serially called to carry out resource hierarchical scheduling, and a hierarchical scheduling initial solution set is generated; calling a designed optimal scheduling layer, and performing association constraint reconstruction and optimal scheduling processing on the hierarchical scheduling initial solution set according to the demand object set to generate a hierarchical scheduling optimal solution set; the hierarchical scheduling optimal solution set is used for indicating a medium and high orbit satellite resource scheduling result corresponding to the resource use application. And the scheduling efficiency of medium and high orbit satellite routine measurement and control tasks is greatly improved.

The invention discloses a service location method and provides a semantic-oriented distributed service discovery system. The method comprises the following steps: dividing a main body graph into a plurality of CGs through a main body graph division algorithm by the distributed service discovery system, wherein each CG has similar concept semantics; converting semantic description information of the service into a CG sequence, that is to say, generating codes of a length-variable concept region combination sequence; forming a semantic P2P network with a structure by corresponding the nodes and the concept region sequence in the distributed service discovery system; achieving service location according to strategy of a prefix-based semantic route supporting the concept region sequence; and carrying out effective distributed storage and discovery of large-scale service described by a large-scale main body concept. The proposed scheme has high versatility, and can extend to description, management, location and discovery of universal resources on the network besides the discovery and management of semantic-based Web service.

The embodiment of the invention relates to the computer field, and discloses a memory leakage detection method and a device thereof. The method comprises the following steps: when a Java application program runs in a virtual machine, calling a Java virtual machine tool interface JVMTI to monitor whether an object which is recycled exists; wherein the object is associated with an identifier; removing an identifier associated with the object in a preset set in the event that an object which is recycled is monitored; wherein the preset set comprises respective associated identifiers of a plurality of objects; an object associated with one or more of the remaining identifiers in the preset set is determined to be a memory leak object. The embodiment of the invention can improve the efficiencyof memory leak detection.

The invention relates to a discrete method for signals with complicated and chanceful sampling step length, which can be used in a signal generator; the invention is characterized in that the method includes the following steps: the sampling step length (see formula 1) required for signal dispersing is acquired according to the period Ti of each subwave and required signal discrete point number Ni; the calculation of all sampling step length dTi is in line with a reference time t which is equal to Zdt in the requirement that dTi is equal to Xit; controlling parameters (see formula 2) take the reference time t as the step length to carry out equal step length signal dispersion to each subwave to obtain a discrete signal sequence, that is, yij is equal to fi(j-1)f, j is equal to 1,2,..., Ni+1...; in the discrete signal sequence obtained in the step 3, Xi discrete points are taken as a group and are uniformly evaluated to the dereference of a first point in the group; discrete signals which meet the requirements of complicated and chanceful sampling step length are formed. The method has the advantages that the dispersion of the signals with complicated and chanceful sampling step length of the signal generator can be realized; compared with the existing signal generator which has the same function, the cost can be reduced by one order of magnitude.

The invention discloses an adaptive screening method for similar images of a capsule endoscope. The method comprises: all original images are divided into a plurality of sub image sequences according to a time sequence; and sampling is carried out on the sub image sequences, a similarity degree of has two adjacent images in a sampling sequence is calculated, similar images are screened out, and then the rest of image as a screened image is outputted. Compared with the prior art, the provided method has the following beneficial effects: temporal correlation and structural correlation of similar images of the capsule endoscope are considered fully and a structural similarity degree is used as a measure for an image similarity degree of the capsule endoscope; and thus compared with other methods, the provided method has high accuracy and the magnitude order of the image of the capsule endoscope is reduced, so that a doctor can carry out diagnosis and analysis conveniently.

The invention discloses an image superpixel meshing method based on fusion. The method comprises the following steps of partitioning an input image by a plurality of superpixel partition methods to obtain a plurality of superpixel partition results, carrying out edge extraction, linear fusion and normalization processing in sequence, and performing distance conversion on a processed result to generate a soft edge graphic; setting mesh points by a Hungary algorithm; calculating an edge between adjacent mesh points according to an active contour to update the soft edge graphic; endowing each image region which is surrounded by four mesh points and the edges among the mesh points with a corresponding label which uniquely determines the image region, thus generating mesh superpixels. According to the method, the computational order of magnitudes is reduced, the operation speed is improved, the partitioning effect of the mesh superpixels is good on the edge of an object, the detection precision is improved, the need in the actual application is met, and according to an experiment result, the method provided by the invention can obtain more accurate detection result within shorter computation time.

The invention provides an Android application network behavior classification method and system based on deep learning, and the method comprises the steps of firstly constructing a user-defined eventexecution sequence combination, driving an Android application to operate, and obtaining a dynamic network behavior data packet of the Android application; bicycling, preprocessing the dynamic behavior data packet of the Android application obtained in the step 1, and converting binary data transmitted in the network into a gray value so as to convert the data packet into a gray image; then constructing a convolutional neural network classification model; and finally converting the preprocessed grayscale image into a matrix vector, and inputting the matrix vector into a convolutional neural network classification model for learning to obtain a classification model of the Android application sample. According to the present invention, the Android application classification problem is converted into the image classification problem, the features are learned by using the multilayer convolutional neural network, and the accuracy is higher.

The invention provides a heat pipe with a primary conductive circuit. The heat pipe comprises a heat pipe body, a heat-conducting insulating layer, a circuit layer and an electronic device, wherein the heat-conducting insulating layer covers one end of the heat pipe body or cover a plurality of positions of the heat pipe body; the circuit layer covers the heat-conducting insulating layer; the electronic device is fixed on the circuit layer; heat emitted when the electronic device works is transmitted to the heat pipe body from the bottom through the circuit layer and the heat-conducting insulating layer; the heat can be rapidly transmitted to other heat dissipation devices connected with the heat pipe body with low resistance through the heat pipe body, so as to be dissipated, while electric connection is realized through the circuit layer; and the invention further provides a preparation process of the heat pipe with the primary conductive circuit.

The invention provides an audio processing device, an audio crosstalk processing method and an audio crosstalk processing device. The audio processing device comprises an audio processing chip, a control switch and an audio output interface, the audio processing chip comprises a first power amplifier and a second power amplifier. The first power amplifier is used for outputting a left channel signal; the second power amplifier is used for outputting a right channel signal; the control switch is coupled with the common negative end of the first power amplifier and the second power amplifier andused for feeding back a reference feedback signal to the common negative end and connecting an earphone ground signal of the audio output interface with the main board ground, and the reference feedback signal is obtained by conducting voltage sampling on the earphone ground signal. By means of the audio processing method and device, the problems that crosstalk between sound channels is likely tobe too large and the listening experience of a user is affected by an audio processing device provided in the related technology are solved, and the effect of improving the crosstalk index so that when the earphone is used for listening to high-quality HiFi music, the stronger space sense can be obtained can be achieved.

The invention discloses a time synchronization method and electronic equipment, and the method comprises that: a first module transmits a clock synchronization signal to a second module, enables RTC information to be coded on the clock synchronization signal, and transmits the clock synchronization signal to the second module, wherein the clock synchronization signal is used for measuring the timedelay between the first module and at least one second module, the time delay is used for carrying out phase compensation on a clock synchronization signal received by the second module side, and theclock synchronization signal after phase compensation is used for triggering the second module to update local second RTC information to the first RTC information.

High performance non-volatile memory devices have the programming voltages trimmed for individual types of memory pages and word lines. A group of word lines within each erasable block of memory are tested in successive program loops to minimize the problem of incurring excessive number of erase / program cycles. An optimum programming voltage for a given type of memory pages is derived from statistical results of a sample of similar of memory pages.

The invention relates to an improved scheme for a thrust device of a space rocket. The improved scheme is that groups of eruption tubes E are additionally installed on the exterior walls of a carrier rocket and a spaceship A of a spacecraft for adjusting the deviation degree of the flying path of the spacecraft at any time. On the one hand, energy consumption required to adjust the course after launch of the spacecraft can be saved and on the other hand, the flying state of the spaceship A being separated from the carrier rocket C can be controlled.

The invention provides a cloud traffic management system. The system comprises an image acquisition module, an image processing module, an image transmission module, a cloud server and a traffic management module, wherein the image acquisition module is used for acquiring a road image; the image processing module processes the road image on the basis of a superpixel model; the image transmission module is used for transmitting the processed image to the cloud server; the cloud server recognizes and analyzes the road image to obtain current traffic jam conditions; the traffic management moduleis used for performing traffic dispersion according to the traffic jam conditions. The system has the benefits as follows: real-time traffic monitoring and effective management are realized, traffic jam is relieved, the order of magnitudes of image representation is reduced by establishing the superpixel model, the complexity of a following image processing algorithm is further reduced, and the probability is provided for real timeliness of the image processing algorithm.

The invention discloses a production process of a soft package instant fish block based on a microwave fast sterilization technology. The method comprises the following steps of (1) performing raw material pretreatment, and performing cleaning, skin removal and bone removal on fresh fish; (2) performing split charging; cutting fish meat into blocks; performing weighing; performing split charging into steaming and boiling bags; (3) performing vacuum sealing packaging; sealing openings of the steaming and boiling bags filled with the fish; (4) performing microwave fast heating, wherein the heating time is shorter than 5min; performing microwave temperature rise to 60 to 90 DEG C so that the coldest point temperature of the fish block reaches 90 DEG C to 100 DEG C; (5) after the microwave heating, performing heat insulation on the fish block for 3min to 6min at 95 + / - 2 DEG C; (6) performing water bath cooling for 10 min; lowering the temperature of the coldest point of the fish block toa value below 25 DEG C; (7) finally, performing storage at normal temperature or at 0 to 4 DEG C in a refrigeration house. An alternating electric field is used for performing integral heating on thematerials during microwave heating; the heating efficiency is higher; the speed is higher; the production of high-quality low-salt and low-addition health products can be realized.

Theembodiment of that present application provide a neural network system, Among them, the neural network system, includes a clustering module for clustering a plurality of simulation results to determine a plurality of simulation clustering results, and determining a corresponding plurality of action sets according to the plurality of simulation clustering results, wherein each action set includes at least one action sample, and the simulation result is a simulation result of a target entity after the action sample is executed; The reinforcement learning module is configured to select an action set to be executed by the target entity from the plurality of action sets according to a state set of the target entity and output the selected action set to be executed by the target entity. The scheme provided by the embodiment adopts the reinforcement learning module of the action set training neural network system, and the order of magnitude of the action set is greatly reduced compared with the action sample, so that the training time of the neural network system can be shortened, the speed of finding the optimal solution in the training process can be increased, and the accuracy of the training result can be improved.

The invention provides a method for detecting aeromonas hydrophila B11 through lutetium-europium co-luminescence time-resolved fluoroimmunoassay, and relates to co-luminescence time-resolved fluoroimmunoassay. The method comprises the steps that an aeromonas hydrophila antiserum is prepared; a B11 antibody is purified; a co-luminescence system is established; and lutetium-europium co-luminescenceTRFIA of the B11 is detected. In the presence of excessive Lu<3+>, the fluorescence intensity of a Eu<3+>-TTA-phen-Tritonx-100 system is enhanced. Condition optimization is conducted on the co-luminescence system, and a result shows that when the concentration of Lu3+, the concentration of TTA, the concentration of phen and the concentration of Triton X-100 is 7.5*10<-6> mol / L, 5*10<-5> mol / L, 4*10<-4> mol / L and 0.075% respectively, the fluorescence intensity of the co-luminescence system is the highest. By applying the direct co-luminescence TRFIA method to detect the B11, the bacterium withthe concentration of 1*10<3> cfu / mL is detected and is two magnitude order lower than a detection result obtained by using direct TRFIA.

The invention belongs to the technical field of lithium ion battery anode material comprehensive utilization and particularly relates to a lithium battery anode recycled material dry purification separation and regeneration method and an obtained lithium battery anode recycled material. The invention provides a low-temperature dry heat treatment, water heating impurity removing and lithium supplementing and high-temperature solid phase regeneration combination technical scheme. The obtained lithium battery anode repaired material is recovered in terms of shape and crystal form, the performanceof the material is excellent, the purity is high, the material can be directly used for lithium battery production, and the performance of an obtained lithium battery is good.

The embodiment of the invention provides a display screen state control method and apparatus, a storage medium and a terminal. The method comprises the following steps: in a closed state of a signal transmitter, obtaining, by a signal receiver, a first signal; controlling the signal transmitter to send a detection signal, and obtaining, by the signal receiver, a second signal; obtaining a first intermediate signal according to the first signal and the second signal, and converting the first intermediate signal into a digital signal obtain a first digital signal; controlling to close the signaltransmitter and controlling the signal receiver to obtain a third signal; obtaining a second intermediate signal according to the first signal and the third signal, and converting the second intermediate signal into a digital signal obtain a second digital signal; calculating a proximity signal according to the first digital signal and the second digital signal, and controlling the state of a display screen according to the signal strength of the proximity signal. The magnitudes of the first digital signal and the second digital signal are reduced and the accuracy of the data is improved, thereby improving the accuracy of controlling the state of the display screen of the terminal.