48results about How to "Simplify the algorithm process" patented technology

The invention provides a method for determining a whole-cycle ambiguity of a three-frequency carrier phase of a BeiDou navigation system. The method comprises the steps of by utilizing the advantage of high determination accuracy rate of a whole-cycle ambiguity of a BeiDou ultra-wide lane carrier phase combination, firstly determining whole-cycle ambiguities of two groups of ultra-wide lane carrier phase combinations, and correcting a pseudo-range initial value through the whole-cycle ambiguities of the ultra-wide lane carrier phase combinations; meanwhile, removing influences from an orbit error, a clock error and an ionized layer error in original observed quantity by combining with carrier phase observation quantity of geometry-free de-ionized layer composed of a narrow lane carrier phase combination, so as to enable the resolution of the whole-cycle ambiguity of the narrow lane carrier phase combination to be only affected by random measurement noise; taking an average value of combined observation quantity of multiple epochs to remove random noise error and determine the whole-cycle ambiguity of the narrow lane carrier phase combination; finally, determining the whole-cycle ambiguity of the carrier phases of three frequencies B1, B2 and B3 by resolving a system of linear equations which is composed of the whole-cycle ambiguities of the two ultra-wide lane carrier phase combinations and the whole-cycle ambiguity of the narrow lane carrier phase combination.

The present invention discloses a zone area reasonable line loss prediction method based on a data mining technology. The beneficial effects of the present invention are that: the method of the present invention aims to discover the potential correlation between the zone area characteristic data and the zone area line loss by using the actual sampling data of the zone area line loss and utilizing a data mining means to process the mass data, is reliable in data source by being compared with a conventional theoretical line loss calculation method, and can reflect the on-site actual line loss status better; the algorithm flow is simple and practical, and the efficiency is higher; a multiple linear regression model is compact in structure, can reflect the influence degrees of the characteristic parameters to the zone area line loss visually, and is easy to analyze the status of a zone area of unreasonable line loss; the method is easy for software realization, and can be integrated in a line loss management system more conveniently.

Provided is a high voltage circuit breaker self-diagnosis method based on abnormal sample identification, comprising: (1) collecting high voltage circuit breaker current characteristic data through high voltage circuit breaker manufacturers, Jiangsu province maintenance departments, etc.; (2) initializing a storage area (including a real time database, a distinguishable area and an undistinguishable area) of a self-diagnosis model; (3) setting self-diagnosis model correlated parameters of following modules: a PSO optimization module, a KFCM identification module, etc.; (4) placing an initial training data set in the distinguishable area of the self-diagnosis model, adding KFCM clustering results based on the initial training data set into the real time database, and accordingly the PSO optimization module calculating the initial parameter of an SVDD detection module; and (5) the self-diagnosis model performing circular operation to process test samples, processing types comprising: distinguishable sample processing, undistinguishable sample processing and new fault type appearance processing. The method can effectively clarify and summarize current data of a high voltage circuit breaker, and provide a more scientific basis for management of electric power maintenance departments, and production and improvement of high voltage electrical equipment manufacturers.

The invention belongs to the field of infrared image processing, and provides an infrared polarization image edge detection method. The infrared polarization image edge detection method comprises the following steps: 1) colleting and pre-processing infrared polarization image information; 2) carrying out image shift processing, and separately carrying out a shift operation to the pre-processed polarization images of three channels; 3) carrying out Stokes polarization state resolution; 4) finishing polarization component decomposition; 5) carrying out superposition fusion and threshold value segmentation; and 6) carrying out morphological processing, smoothening a target outline through open operation, weakening a narrow part, removing a thin extruded part, and extracting the edge of a single pixel by morphological refined operation. Infrared polarization characteristic differences are utilized to enhance target edge outline information with transition nature and decompose intensity information ineffective to target identification, and edge information can be effectively highlighted to realize edge detection.

The invention provides a moving target tracking system and tracking method based on random medium surface diffusion light to solve the technical problem of narrow application scope of a conventional moving target tracking system and tracking method. The tracking system comprises a laser, a beam expander, rotary ground glass, a random medium, an aperture diaphragm and a detector array, wherein the light emitted by the laser forms pseudothermal light after passing the beam expander and the rotary ground glass and irradiates on a moving target to be tracked, light carrying target information is irradiated onto the scattering surface of the random medium at an acute angle, the diffusion light, of which the stray light is filtered by the aperture diaphragm, is received and acquired by the detector array, and the tracking method is used to calculate the acquired images to obtain the actual displacement of the moving target to be tracked. Target tracking is realized through diffusion light on the surface of a random medium, and the application scope is wide. The system and the method also have the advantages of simple algorithm, high tracking precision and simple system structure, and are applicable to the fields of biomedicine and earth observation.

The invention belongs to the field of laser scanning surveying and mapping, and provides a geometric correction method for a vehicle-mounted scanning point cloud in a tunnel GNSS denial environment, which can solve the problem of geometric correction of the point cloud of non-rigid body deformation in a tunnel area, thereby meeting the requirement of high-precision point cloud acquisition. The method comprises the following steps: firstly, arranging targets in groups along a tunnel extension direction; secondly, obtaining point cloud data through surveying and mapping of a vehicle-mounted scanning system, and obtaining calibration point coordinates of each target through actual measurement; thirdly, extracting three-dimensional coordinates of homonymy points of the corresponding targets from the point cloud data obtained by the vehicle-mounted scanning system; then, calculating conversion parameters between actual measurement target coordinates and point cloud target coordinates in groups, correcting track positions of the vehicle-mounted scanning system corresponding to each group of targets, and complementing tracks of the mobile measurement system in other track areas to be corrected through adoption of an interpolation method; and finally, re-resolving by using the corrected accurate track of the mobile measurement system to obtain high-precision long tunnel point cloud data.

The present invention provides an image de-noising method and device. The method comprises: establishing a linear shift variant filtering linear function model qi=akIi+bk, and A-inversed I belongs to wi, wherein qi is the brightness value of a pixel point in the image after denoising, and Ii is a brightness value of the pixel point in an original image; performing filtering process of the original image, and obtaining an image to be filtered; employing the principle of infinite approaching of the brightness value qi of the pixel point in the image after denoising and the pixel value pi of the pixel point in the image to be filtered, and calculating the expression of the coefficient ak and the coefficient bk; performing window mean value filtering processing and pixel point operation of the original image and the image to be filtered according to the expression of the coefficient ak and the coefficient bk, and obtaining the concrete values of the coefficient ak and the coefficient bk in a window wk; and substituting the concrete values of the coefficient ak and the coefficient bk in the window wk into the function model qi=akIi+bk, and obtaining a de-noising image. The image de-noising method and device use window mean value filtering and pixel point operation to collect the calculation of the window target value, the algorithm flow is simple and easy to realize, and the linear function model can maintain the image edge characteristics.

The invention provides a white balance method and device. The method includes the following steps of: firstly receiving an RGB data stream, and calculating red gain and blue gain of pixel points in the data stream; then calculating average red gain and average blue gain of all the skin color points; mapping the average red gain and the average blue gain to a skin color characteristic color temperature curve, and determining the color temperature value of a current environment; mapping the color temperature value of the current environment to an ideal white point characteristic temperature color temperature curve, and determining an ideal red gain value and an ideal blue gain value; and performing white balance correction according to the ideal red gain value and the ideal blue gain value. The white balance method in which skin color characteristics are mapped to the white point characteristic color temperature curve is employed, the interference of the skin color to the white balance can be fully removed, and the problem that large area skin color photograph pictures have color bias is solved.

The invention relates to the technical field of unsupervised image anomaly detection and segmentation, in particular to a deep learning algorithm for detecting anomaly and segmenting an abnormal region in an image. The method comprises the following steps: S1, firstly, performing positive sample image feature extraction, designing a deep learning algorithm for detecting anomalies and segmenting abnormal regions in an image, enabling the algorithm to be high in accuracy, and obtaining top1 classification precision on a public data set test list. Reasoning time is short, only model parameters need to be stored, and positive sample features or statistics thereof do not need to be stored; compared with an algorithm for performing mahalanobis distance measurement according to feature statistics, the algorithm is wider in applicability, and the image does not need to meet the constraint condition of alignment; the algorithm process is simple, only the optimal single-level feature representation needs to be selected, and multi-level feature fusion is not needed. The segmentation threshold value can be selected in different modes according to actual conditions. And the algorithm is good in generalization ability and easy to reuse.

The invention relates to a non-invasive load monitoring method based on equipment transfer learning, and the method comprises the steps: collecting the load amplitude data of household electrical appliances and the household total load amplitude data, and building an original data set; normalizing the data of the original data set; constructing an initialization sequence-to-point CNN model to construct a transfer learning equipment selection model, and selecting initial data and target data of transfer learning; performing a training process and a verification process on the initialization sequence-to-point CNN model by adopting the initial data to form a transfer learning initial model; finely adjusting a full connection layer of the initial model by adopting the target data to form a transfer learning target model; and inputting the normalized household total load amplitude data into the transfer learning target model to obtain the power amplitude of the target equipment, thereby realizing non-intrusive load monitoring of the target electrical equipment. According to the invention, the model training time is reduced, and the economic cost and time cost of non-intrusive load monitoring are reduced.

The invention discloses a fluorescence image registration method, and belongs to the field of gene sequencing. And carrying out image registration through a preset positioning line. According to the fluorescent image registration method, image registration can be completed by identifying fluorescent image characteristic positioning marks and comparing position differences of the fluorescent image characteristic positioning marks in different sequencing cycles depending on the positioning lines which are preset on the gene sequencing chip and distributed at equal intervals.

The invention discloses a millimeter wave sparse array remote monitoring imaging method and system, and belongs to the technical field of millimeter wave three-dimensional holographic imaging. The method comprises the following steps: S1, obtaining an echo signal; S2, performing interpolation operation; S3, performing fast Fourier transform; S4, selecting a distance plane; S5, performing frequencydomain matched filtering; S6, performing fast Fourier inverse transform; S7, performing time domain matched filtering; S8, performing coherent accumulation; and S9, obtaining a three-dimensional complex image. According to the millimeter wave sparse array remote monitoring imaging method and system disclosed by the invention, a millimeter wave sparse array full-electronic scanning imaging systemis adopted, and compared with a traditional optical machine scanning far-field imaging system, the millimeter wave sparse array full-electronic scanning imaging system has the characteristics of smallelectromagnetic wave flicker effect, high image signal-to-noise ratio and large imaging field range; meanwhile, compared with a traditional time domain type imaging method, the adopted signal processing imaging method has the advantages of less hardware resources and less storage space, the algorithm process is simple and easy to understand, the algorithm main body only comprises the main body process of fast Fourier transform, matched filtering and coherent accumulation, and the calculation efficiency is higher than that of the traditional time domain type algorithm.

The invention provides a robot teaching recording system which is efficient, high in applicability, speedy and low in cost, and has few parts, provides a teaching method based on the system, and belongs to the field of robot teaching. The system comprises optical beacons, an IMU sensor, a camera, a data fusion calculation unit and a robot control cabinet, wherein the robot control cabinet is a control driving unit for a robot and receives data or files output by the IMU sensor and a visual fusion calculation unit; the camera is used for acquiring the positions of the beacons; IMU data are fused by the data fusion calculation unit; and the positions, speeds and accelerated speeds of the optical beacons and process actions at corresponding positions are output in a fused manner finally by the data fusion calculation unit. The provided teaching system and method can be applied to following occasions: teaching actions and process can be finished by teaching workers, robot sensors do not require to be added or a mechanical structure does not need to be changed, and a robot system only needs to receive external commands or external files.

The invention belongs to the technical field of the inverter, and particularly relates to an SHEPWM control system and method for a multi-level inverter. The SHEPWM technology of the multi-level inverter is realized by adopting a particle swarm-bacteria foraging optimization algorithm (PSO-BFO), so that the output waveform is ensured while the switching angle is worked out accurately, and the purpose of eliminating special harmonics, simplifying the operational process and shortening the calculation time is achieved. The SHEPWM control system for the multi-level inverter comprises a DPS processing circuit, a voltage sampling circuit, an isolation driving circuit, a switching power supply circuit and a voltage sensor, wherein the voltage sensor, the voltage sampling circuit, the DPS processing circuit and the isolation driving circuit are connected in sequence; and the output port of the isolation driving circuit is connected with the control port of the switching device of a T-type three-level photovoltaic grid-connected inverter.

The method is used in an electronic control system composed of the key-press, the display, the video cameral, the DSP, the micro phone, the speaker, the audio codes, the MPU and the memory. It includes following steps: the video cameral module collects image data and at same time sends a instruction to the audio collection module for starting audio collection; in the process of collection the image data positioning label and the audio data positioning label are set at fixed time, when the collection of image is finished the video cameral module sends a instruction to the audio module for stopping the collection of audio, the image data and audio data are separately saved in the memory; the mobile phone display the audio and video file, the video camera reads the image and audio data from memory and sends them into the image buffer and audio buffer; the image is displayed on the display screen and the sound is outputted by the speaker through decode and D / A conversion.

The invention is suitable for the technical field of computers, and provides a vehicle positioning method and device, computer equipment and a storage medium. The method comprises the steps of obtaining the position information of a vehicle at a plurality of moments in a preset sampling period; judging whether the vehicle state is a static state or not; if yes, determining accurate positioning position information of the vehicle based on a preset static state vehicle positioning algorithm; and if not, mapping and processing the position information of the vehicle at the multiple moments in the preset sampling period into multiple pieces of mapping position information of the vehicle at the initial moment on the basis of a preset dynamic mapping rule, and then determining accurate positioning position information of the vehicle at the initial moment. According to the vehicle positioning method provided by the invention, different positioning algorithms are provided for different vehicle states, the accuracy is higher, a problem in a motion state is converted into a positioning problem in a static state, the algorithm process is simplified, and the solving efficiency is improved.

The invention relates to a single robotic fish behavioral control strategy based on a three-dimensional fuzzy PTP algorithm and belongs to the field of intelligent control technology. The strategy relates to robotic fish ball heading movements, provides a PTP algorithm which faces to division and improvement for ball heading points in different regions, designs behavioral strategies corresponding to different regions and provides a basic behavioral sequence in robotic fish ball heading process. A control algorithm which faces to different regions is adopted to realize a simple task which is to let a robotic fish to head a ball to reach a target point; pool regions are divided specifically according to ball heading points, and ball heading movement behaviors of a robotic fish in corresponding regions are designed. According to the invention, the strategy is used for the robotic fish in complex underwater environments with barriers, noise and the like; three-dimensional fuzzy processing is performed on speed, angle and sampling time, the mode turns to control mode, it is convenient to directly use the robotic fish control mode, algorithm flow is greatly simplified, algorithm complexity is reduced, and the strategy also has advantages of fast data update speed, high precision and good real-time capability.

The invention discloses a data consistency method for a distributed service discovery cluster in a micro-service architecture. The method comprises the following steps: when a service instance of a client is started, any one service discovery node is selected from the service discovery node cluster as a Leader node for connection; sending the own service information to the Leader node; other service discovery nodes in the service discovery node cluster are used as Flower nodes; wherein the service information comprises a service name, an IP address, a port number, a service instance current state, a service instance data type, a Leader node ID, a data version number, time for writing data into a certain service discovery node, time for writing the data into a service discovery node clusterand a state whether the data is submitted or not; the Leader node stores the service information in a tree structure; and the Leader node copies the service information to other Folower nodes.

The invention discloses a macaque emotion recognition method and system based on time-frequency double-branch features, and the method comprises the steps: inputting collected macaque sounds into a pre-established and trained macaque emotion recognition model, and obtaining corresponding accurate emotion types; wherein the accurate emotion types comprise friendly emotion, attack or threat emotion, falling or compliance emotion and non-emotion; and the macaque emotion recognition model is used for extracting intermediate features of macaque sounds, performing separation and extraction on time domain features and frequency domain features based on time-frequency double branches, and determining corresponding accurate emotion categories through fusion and classification. According to the method and the system, original sound data is used as network input, emotion-related intermediate features are extracted from sound, voice preprocessing is not needed, the algorithm process is simplified, then time domain and frequency domain features are separated and extracted, compression features of the sound are obtained from different angles, the time domain and frequency domain features are fused, the expressive power of the features is enhanced, and the identification accuracy is effectively improved.

The invention belongs to the technical field of function optimization. In order to solve the technical problems that an existing function optimization algorithm is low in parallel efficiency and the optimal solution position is prone to loss, a multi-scale quantum harmonic oscillator optimization method with the inherent parallel ability comprises the steps that firstly, the value of sigmamin is set, the value of sigmas is set, and K first random numbers are decimated randomly in the domain of definition of f(x); secondly, K second random numbers are generated through sigmas and the first random numbers according to normal distribution; thirdly, the K first random numbers are substituted into f(x) to obtain K first function values, and the K second random numbers are substituted into f(x) to obtain K second function values; fourthly, for any xi and xi', if f(xi') is smaller than f(xi), xi is substituted with xi', if substitution exists, the second step is executed, and if no substitution exists, the fifth step is executed; fifthly, Xworst is substituted with Xmean, and the sixth step is executed; sixthly, if sigmak is larger than sigmas, the second step is executed, otherwise, the value of sigmas is reduced, and then the seventh step is executed; seventhly, if sigmas is larger than sigmamin, the second step is executed, and otherwise, f(xbest) and xbest serve as results to be output.

The invention provides a novel multi-process image compression encryption algorithm which is based on a super Lorenz system and is simultaneously combined with a two-dimensional compressed sensing algorithm, an image bit plane decomposition method and DNA coding. The method mainly comprises the following steps of: sampling a plain text image by adopting a compressed sensing algorithm according to a compression ratio, obtaining four groups of pseudo-random sequences with specified lengths through a super Lorenz system based on the compressed image, decomposing the compressed image to obtain eight bit planes, scrambling the eight bit planes by combining integers of the pseudo-random sequences respectively, performing DNA coding on the scrambled image and the integer pseudo-random sequence, and performing exclusive or operation based on DNA coding on the scrambled image and the integer pseudo-random sequence to obtain a cipher text image. The encryption algorithm provided by the invention has very strong plain text correlation, and the execution efficiency of the algorithm is also improved while the security is ensured.

The invention discloses a depth-of-field grade estimation method based on a single frame image, which is carried out according to the following steps: step 1, calculating a dark channel image of the image according to the dark channel theory; step 2, calculating that gray scale image of the air transmittance of the image according to the prior theory of the dark channel; step 3, statistic that airtransmissivity map obtained in the step 2 to obtain a gray level histogram; step 4, performing threshold analysis on that gray level histogram to select an appropriate threshold; step 5, applying a threshold value to that transmissivity image to obtain the segmented depth-of-field image. The invention is a simple and easy-to-operate depth-of-field estimation method for daytime scene, which does not need to rely on complex samples and pixel classifiers, and is an end-to-end image processing algorithm. At the same time, it can greatly reduce the amount of computation, which is suitable for embedded platform, UAV and other application scenarios. Based on the theory of dark channel, a simple and easy-to-implement method for estimating the depth of field is derived.

The invention discloses a vehicle detection method and device, a storage medium and electronic equipment, and the method comprises the steps: inputting a to-be-detected target image into a feature extraction layer of a vehicle detection model during vehicle detection, and extracting a first feature map of the target image; then, detecting a vehicle detection frame, a license plate detection frame and a vehicle type in the first feature map through a target detection layer, and performing up-sampling on the first feature map to obtain a second feature map; and then, determining a license plate region feature map from the second feature map based on a license plate detection frame, and recognizing the license plate number of the license plate region feature map through a number recognition layer. And finally, according to the vehicle detection frame, the license plate detection frame, the identified vehicle type and the license plate number, determining a vehicle detection result of the target image. The vehicle position, the vehicle type, the license plate position and the license plate number in the target image are detected through the vehicle detection model, manpower resources are saved, and the detection efficiency is improved.

The invention provides an image beautifying processing method and device, a storage medium and electronic equipment, and relates to the technical field of image and video processing. The image beautification processing method comprises the following steps: acquiring a face image to be beautified; performing first facial beautification processing on the to-be-beautified face image through a pre-trained deep neural network to obtain a middle face image, and determining a mask of the to-be-beautified face image through the deep neural network; and based on the mask of the to-be-beautified face image, performing second beautification processing on the middle face image to obtain a target image. The problem of facial beautification error processing is improved, and the facial beautification effect is improved.

The invention discloses a high-voltage line detection method combining deep semantic segmentation with Hough transform, and the method employs a deep learning semantic segmentation technology to makethreshold judgment of a target through combining with the Hough transform, achieves high-performance detection of a high-voltage line, can achieve further confirmation of a detection target through the threshold after the Hough transform, and improves detection accuracy. The false target of an original image is suppressed, and the dual effects of target detection and false alarm reduction are achieved. According to the technical scheme disclosed by the invention, an algorithm flow is concise, detection rate is higher than that of a traditional parameter adjustment method, the false alarm rateis further reduced through Hough transform, and the problem of excessive false alarms of the high-voltage line caused by semantic segmentation is avoided.

The invention relates to an OCR-based character segmentation method, which comprises the following steps: 1, acquiring a template character library based on an OCR technology, the template character library comprising standard characters and characteristic data of the standard characters; 2, identifying a part of characters in the same batch as the to-be-identified characters by using a character identification model in the OCR technology to obtain a character segmentation result, manually marking error items in the segmentation result, and updating the character identification model; 3, performing line scanning on a character to be recognized, performing initial recognition on the character to be recognized based on the updated character recognition model, and when the score of the recognition result of a certain character is smaller than a first threshold value, performing forced segmentation on the character; step 4, carrying out normalization processing; and 5, matching a certain character with a standard character according to the normalized feature data, calculating to obtain the standard character with the highest score, and determining the segmentation position of the current character based on the standard character with the highest score. The accuracy of character segmentation can be improved.

The invention provides a method for determining a whole-cycle ambiguity of a three-frequency carrier phase of a BeiDou navigation system. The method comprises the steps of by utilizing the advantage of high determination accuracy rate of a whole-cycle ambiguity of a BeiDou ultra-wide lane carrier phase combination, firstly determining whole-cycle ambiguities of two groups of ultra-wide lane carrier phase combinations, and correcting a pseudo-range initial value through the whole-cycle ambiguities of the ultra-wide lane carrier phase combinations; meanwhile, removing influences from an orbit error, a clock error and an ionized layer error in original observed quantity by combining with carrier phase observation quantity of geometry-free de-ionized layer composed of a narrow lane carrier phase combination, so as to enable the resolution of the whole-cycle ambiguity of the narrow lane carrier phase combination to be only affected by random measurement noise; taking an average value of combined observation quantity of multiple epochs to remove random noise error and determine the whole-cycle ambiguity of the narrow lane carrier phase combination; finally, determining the whole-cycle ambiguity of the carrier phases of three frequencies B1, B2 and B3 by resolving a system of linear equations which is composed of the whole-cycle ambiguities of the two ultra-wide lane carrier phase combinations and the whole-cycle ambiguity of the narrow lane carrier phase combination.

The invention discloses a method for efficiently solving a steady-state microdynamics equation set based on automatic combination of time integration and a Newton method. The method comprises the following steps of carrying out time integration on the initial value by adopting a time integration method based on a low-precision numerical value, automatically integrating to obtain a coarse-precisionsteady-state coverage degree by combining coverage degree sensitivity, and obtaining the steady-state coverage degree by combining with a high-precision numerical value Newton method; if single solving fails, randomly initializing the coverage in the exponential space to carry out a new round of solving, which is used for automatically solving the microcosmic kinetic equation set of the catalystsurface chemical reaction network. The method provided by the invention has higher success rate, speed and precision for solving a steady-state microdynamics equation set; meanwhile, the method is high in automation degree, automatic iterative solving can be achieved only by setting initial parameters, man-machine interaction is not needed in the process, and the algorithm process is simple and easy to master.