1185 results about "Reference sample" patented technology

The invention discloses a sewage plant accurate dosing method based on multi-mode automatic machine learning, and belongs to the technical field of sewage treatment. The method comprises the following steps: acquiring historical sensing data and image data of a sewage plant; extracting key features in the image by using a YOLO visual model, fusing the key features with sensor data, and constructing a multi-modal sample data set; carrying out modeling training on the fused data by adopting an automatic machine learning framework, and constructing a'water inlet-dosing-water outlet 'prediction model; generating a reference sample data set according to an existing dosing rule; inputting the sample data into the trained prediction model to obtain predicted effluent quality data, and adjusting the dosage until the predicted effluent quality of all samples meets a water quality standard condition; and finally outputting an optimized dosing sample data set. According to the invention, precision and real-time dosing of the sewage plant can be realized, chemical waste can be effectively reduced, the treatment efficiency is improved, the cost is saved, and the water quality is ensured to stably reach the standard.

The invention discloses an intelligent event studying and judging method based on machine learning, which comprises the following steps of: extracting features from multi-source event data, constructing an event fragment set with a uniform structure according to a preset time window, identifying fragments with stable features in the event fragment set as normal samples, constructing a reference model through an MCD algorithm, and analyzing the normal samples. The method comprises the following steps: extracting feature distribution in a normal state, comparing a to-be-analyzed event with a reference model, calculating feature offset, forming a behavior trajectory, determining an anomaly judgment range by adopting an elliptical envelope algorithm, and finally evaluating an event risk level and dynamically adjusting reference sample composition according to a continuous anomaly condition to realize intelligent identification and risk judgment of the event. According to the invention, dynamic perception and intelligent identification of complex events can be realized.

A method comprising: receiving an image block unit of a frame, the image block unit comprising samples in color channels comprising at least one chrominance channel and one luminance channel (900); reconstructing samples of said luminance channel of the image block unit (902); determining a reference area for predicting samples of at least one color channel of the image block unit, wherein said reference area comprises one or more of reference samples in a neighboring block in current color channel / frame, in the neighboring of a co-located block in reference color channel / frame; and / or inside the co-located block in reference color channel / frame (904); determining filter coefficients of a filter for said predicting based on the reference samples and a shape of the filter (906); and predicting samples of at least one color channel of the image block unit by performing a convolution between the samples of said luminance channel and the filter coefficients (908).

A method comprising: receiving an image block unit of a frame, the image block unit comprising samples in color channels comprising at least one chrominance channel and one luminance channel; reconstructing samples of said luminance channel of the image block unit; determining a reference area for predicting target samples of at least one color channel of the image block unit, wherein said reference area comprises one or more reference samples in a neighboring block in current color channel / frame, in the neighboring of a co-located block in reference color channel / frame; and / or inside the co-located block in reference color channel / frame; determining weights for predicting said target samples based on a ratio between a normalized luminance histogram in said reference area and a normalized luminance histogram co-locating said target samples; determining filter coefficients of a filter for said predicting based on the weights, the reference samples and a shape of the filter; and predicting samples of at least one color channel of the image block unit based on the samples of said luminance channel and the filter coefficients.

The invention relates to the technical field of remote sensing recognition, in particular to a garlic recognition system and method based on remote sensing deep learning. According to the method, visible light and red edge wave band information in a multi-temporal remote sensing image for the phenological period of garlic are combined, the alignment precision of image data in the time sequence is improved through spectrum normalization and spatial calibration, and a multi-source classification sample set is constructed by superposing vegetation spectrum indexes and spatial sample positioning labels; a deep learning structure of semantic segmentation and an attention mechanism is introduced to enhance fine expression of spatial coherence and feature response of region recognition, the discrimination definition of a garlic region in an image is improved through channel spatial feature fusion, and a garlic pattern spot spatial boundary is subjected to continuous recombination and comparison verification with a reference sample. And the pattern spot precision and the sample consistency of the final identification area are ensured, so that the classification accuracy and the identification stability of the garlic planting area are remarkably improved under the complex background.

A video encoding or decoding apparatus and method perform existing inter prediction on a current block to generate a motion vector and predicted samples. The video encoding or decoding apparatus and method generate enhanced predicted samples for the current block using a deep learning-based video prediction network (VPN) on the basis of the motion vector, the reference samples, the predicted samples, and the like to improve encoding efficiency.

Embodiments of the present disclosure provide a solution for video processing. A method for video processing is proposed. In the method, for a conversion between a current video block of a video and a bitstream of the video, a process is applied to the current video block based on template matching. At least one reference sample of a current template of the current video block is determined based on a block vector (BV) of the current video block during the process. The conversion is performed based on the applying.

An intra frame prediction method and apparatus based on interpolation filtering, a video encoding method and apparatus, a video decoding method and apparatus, and a system. The intra prediction method based on interpolation filtering comprises: determining, based on a determined filtering window and a reconstructed value of a sample in a reconstruction area neighboring to a current block, a filtering coefficient used to perform interpolation filtering on the current block; and performing interpolation filtering based on a sample value of a reference sample neighboring to a current sample and the filtering coefficient, to determine a predicted value of the current sample.

The invention discloses a zero sample / small sample industrial defect detection method and system, and the method comprises the steps: freezing the parameters of a CLIP image encoder in a segmentation branch, extracting image features, projecting the corrected features, carrying out the one-by-one product operation of the image features of different layers and corresponding text features, and splicing to generate an image-text similarity graph; in the classification branch, the extracted final image features are multiplied by the text features, and a positive / abnormal classification result of the image is obtained after processing through a learnable adapter; the parameters of the adapter are optimized through the difference between the output result of the minimization segmentation branch and the classification branch and the real label, and the zero sample detection mode directly generates an abnormal graph through the matrix multiplication of the image and the text features among the layers; a reference sample is introduced in a small sample detection mode, the cosine similarity between each layer of features and input sample features is calculated, and an obtained similarity graph and an anomaly graph generated by a zero sample are added to form a final anomaly thermodynamic diagram.

The invention provides a rice planting area identification method and device based on multi-temporal remote sensing data, equipment and a storage medium. The method comprises the following steps: acquiring the multi-temporal remote sensing data according to a rice phenological period; determining a spectral index sequence of the multi-temporal remote sensing data to construct a spectral index time sequence curve, and generating planting characteristic data in the target area according to a spectral change curve; determining a spectrum time sequence template matched with the rice growth rule according to the planting characteristic data, performing matching analysis on the planting mode of each land block, and marking a land block set of rice planting mode characteristics; performing cross comparison on the plot set and the rice planting reference sample to generate a rice planting mode sample set, and constructing a rice planting area identification model based on the rice planting mode sample set; and analyzing the remote sensing data of the land parcels to be identified in the research area according to the rice planting area identification model, and determining a rice planting area identification result. According to the scheme, the accuracy of rice identification can be effectively improved.

In some embodiments, a decoder may receive, from a bit stream, a residual of a block and an indication that the residual was determined based on a refined partition edge. The decoder may determine the refined partition edge based on the indication and reference samples neighboring the block. The decoder may determine based on at least one prediction mode, a prediction for two parts of the block defined by the refined partition edge. The decoder may determine a reconstruction for the block based on the residual and the prediction of the block.

The invention provides an underground water environment monitoring system and method, and the method comprises the steps: carrying out the three-dimensional fluorescence spectrum full-wave band scanning of a known type of pollution source reference sample in a target monitoring region, and obtaining a three-dimensional fluorescence spectrum data set; performing characteristic analysis on the three-dimensional fluorescence spectrum data set to obtain fluorescence characteristic components, and determining a characteristic excitation wavelength corresponding to each fluorescence characteristic component; the classification contribution degree of each characteristic excitation wavelength is evaluated, and a characteristic excitation wavelength combination during source tracing of the pollution source is obtained through screening; performing fixed wavelength scanning on the to-be-detected underground water sample in the target monitoring area according to the characteristic excitation wavelength combination, and determining the initial type of the pollution source; and determining the relative contribution degree of the type of the pollution source of the to-be-detected underground water sample, simulating the migration trend of the pollution plume according to the underground water migration and diffusion model, and generating the position of the pollution source of the to-be-detected underground water sample. By adopting the scheme of the application, the pollution source of the underground water can be quickly identified and positioned at low cost.

The invention belongs to the technical field of image processing, and particularly relates to a coded disc surface tiny scratch recognition method based on micro texture feature analysis, which comprises the following steps: acquiring a coded disc grating texture image and converting the coded disc grating texture image into a rectangular coordinate system to obtain an expanded grey-scale map; the optimal main direction of the grating texture is determined through rotation projection analysis; for each pixel point in the grey-scale map, sampling along two sides of the optimal main direction to form a reference sample; calculating a weighted average value of the gray values of the sample points in the reference sample to obtain a trend prediction value, calculating a residual error between the gray values of the pixel points and the trend prediction value, and determining a texture deviation index in combination with the change condition of the gray values of all neighborhood pixel points of the pixel points; and finally, identifying the scratches through threshold segmentation. According to the method, the grating edge interference can be effectively inhibited under the strong periodic texture background, and the tiny scratch features can be accurately extracted.

The invention discloses an image coding prediction mode selection method, which solves the problems that in the prior art, image redundant information is not efficiently utilized in conventional image coding, and complex video coding is relatively high in application delay for low hardware cost, and the like. The absolute reference sample value and the recovery sample value of the reference image are used as reference sample values, so that the image redundant information is effectively utilized, and the compression efficiency is improved; a fixed number of coding prediction modes are set in a selection set, different coding prediction mode selection sets are determined for different scenes, and coding code streams are reduced; when the optimal coding prediction mode is selected, resource consumption and delay are reduced by means of source image information, so that high coding efficiency and good subjective effect can be realized under the condition of low cost and low delay.

A method for decoding a picture, a method for encoding a picture, an encoder, and a decoder are provided. The method for encoding a picture includes (i) determining a width and a height of a coding block in the picture; (ii) if the width and the height are equal to N, where N is a positive integer power of 2, determining a matrix-based intra prediction (MIP) size identifier indicating that an MIP prediction size equal to N; (iii) deriving a group of reference samples of the coding block; and (iv) deriving an MIP prediction of the coding block based on the group of reference samples and an MIP matrix corresponding to the MIP size identifier.

The embodiment of the invention provides a dynamic assessment method and device for a large language model, equipment and a medium. The method comprises the steps that an original reference sample set of the large language model to be assessed is acquired; performing sample dynamic optimization processing on the original reference sample set to obtain a dynamic reference sample set of the to-be-evaluated large language model, the sample dynamic optimization processing including sample screening, sample enhancement and negative sample creation; and based on at least one set large language model and the dynamic reference sample set, verifying and evaluating the to-be-evaluated large language model to obtain a verification and evaluation result of the to-be-evaluated large language model. Different from a traditional static method, the method is used for dynamically evaluating the to-be-evaluated large language model by adopting a dynamic evaluation framework to generate a more complex and novel evaluation sample, so that more detailed and comprehensive evaluation on the model performance is provided.

The invention discloses a text recognition model training method and device, electronic equipment and a storage medium. The method comprises the steps of performing text prediction on a first training sample set of a current training round based on a first text recognition model of the current training round to determine a first prediction result and a first confidence coefficient of the first prediction result; selecting a first reference sample from each first training sample based on each first confidence coefficient; adjusting parameters of the first text recognition model based on a first loss value determined by the first prediction result to obtain a second text recognition model of the next training round; and determining a second reference sample based on the first recognition model and the first reference sample, adding the first reference sample and the second reference sample into the first training sample set to obtain a second training sample set of the next training round, and training a second text recognition model based on the second training sample set. According to the method, the problems of low text recognition accuracy and insufficient generalization ability of a text recognition model in a specific scene are solved.

The invention belongs to the technical field of traditional Chinese medicine preparation, and particularly relates to a peach kernel and safflower four-ingredient decoction compound preparation and a preparation method and a quality control method thereof. The invention provides a preparation method of a peach kernel and safflower four-component decoction compound preparation by screening and optimizing an enrichment method of volatile oil in extraction, concentration, inclusion and drying processes on the basis of an ancient water decoction process. According to the peach kernel and safflower four-drug decoction compound preparation prepared by the invention, the quality consistency with a reference sample is realized, including the consistency of each content index in the compound preparation with the reference sample and the consistency of a fingerprint spectrum with the reference sample; especially, characteristic peaks of volatile components-ligustilide, senkyunolide I and senkyunolide A are presented in the fingerprint spectrum, and the consistency of the content of the volatile oil is determined. Moreover, the quality control method provided by the invention is expected to be used for quality control of industrial large-scale production of a classic famous prescription, namely the peach kernel and safflower four-ingredient decoction compound preparation, and has a good application prospect.

A method for in-loop sample offset filtering in a video decoder is disclosed. The method includes obtaining at least one statistical property associated with reconstructed samples of at least a first color component in a current reconstructed data block of a video stream, selecting a target sample offset filter among a plurality of sample offset filters based on the at least one statistical property, the target sample offset filter comprising a nonlinear mapping between sample delta measures and sample offset values, and filtering a current sample in a second color component of the current reconstructed data block using the target sample offset filter and reference samples in a third color component of the current reconstructed data block to generate a filtered reconstructed sample of the current sample.

The invention relates to methods for determining the concentration of an analyte in a sample obtained from processing of blood-derived plasma, the method comprising applying a light source in the near-infrared spectrum to a test sample obtained from processing of blood-derived plasma; measuring reflectance, transmission or transflectance of the test sample over a range of near-infrared wavelengths, thereby generating test wavelength spectra, comparing the test wavelength spectra with reference wavelength spectra obtained from reference samples having known concentrations of the analyte, to determine the concentration of the analyte in the sample. The invention also relates to the development of a multivariate model for use in determining the concentration of an analyte, for example, total protein or alcohol (e.g. ethanol).

Innovations in intra-picture prediction with multiple candidate reference lines available are described herein. For example, intra-picture prediction for a current block uses a non-adjacent reference line of sample values to predict the sample values of the current block. This can improve the effectiveness of the intra-picture prediction when the reference line of sample values that is adjacent the current block includes significant capture noise, significant quantization error, or significantly different values (compared to the current block) due to an occlusion. Innovations described herein include, but are not limited to, the following: intra-picture prediction with multiple candidate reference lines available; encoding / decoding of reference line indices using prediction; filtering of reference sample values; residue compensation; weighted prediction; mode-dependent padding to replace unavailable reference sample values; using in-loop-filtered reference sample values; encoder-side decisions for selecting reference lines; and post-filtering of predicted sample values.

An aspect of the disclosure provides a method of video decoding. For example, a coded video bitstream is received. The coded video bitstream includes coded information of a plurality of pictures. Based on coded information of a current block in a current picture, it is determined that the current block is coded using an intra prediction that generates prediction samples of the current block based on reference samples in the current picture. At least a first transform kernel set is determined based on a histogram of occurrence (HoC) of intra prediction modes in neighboring samples of the current block. Based on the coded information of the current block, a residual block of the current block is calculated according to at least the first transform kernel set. The current block is reconstructed based on the residual block and the intra prediction of the current block.

The invention discloses a motor bearing fault detection system and method based on robust deep learning, and belongs to the technical field of mechanical fault detection and intelligent perception. Feature extraction is carried out on an original vibration signal with a label based on a supervised learning branch network, and the original vibration signal is used as a reference sample; the samples with the same fault category and different fault categories as the reference samples are positive samples and negative samples, and inter-class separation and intra-class aggregation relations in a triple loss optimization embedding feature space are introduced to generate embedding representation; based on an unsupervised learning branch network, encoding the original vibration signal after time domain and frequency domain artificial feature extraction, and introducing triple loss to carry out unsupervised embedding learning to generate high-level feature embedding representation; and the embedded representations output by the two branch networks are fused, dual loss of triple loss and center loss is introduced for training, and a bearing fault detection model after training is completed is used for bearing fault detection.

A method of decoding coding units of a coding tree for an image frame from a video bitstream comprising splitting a region of the coding tree into a plurality of coding blocks, each of the coding blocks including a prediction block. Determining matrix intra prediction flags for the prediction block of each of the coding blocks, the determination based upon (i) an area of the region if the region meets a threshold, or (ii) a budget for the region if the area of the region does not meet the threshold. The method further comprises reading matrix coefficients from a memory for each prediction block determined to use matrix intra prediction according to the determined flag; and decoding the coding units using prediction blocks generated for each coding unit in the region using reference samples of each prediction block and the matrix coefficients.

Implementations of the disclosure provide a video processing method for performing intra prediction on a video block. The video processing method may include receiving reconstructed samples from a video frame in the bitstream including the video block. The video processing method may further include determining reference samples in neighboring video blocks of the video block according to a first slant line and a second slant line through a sample of the video block. The first slant line is along an interpolation direction and the second slant line is perpendicular to the first slant line. The video processing method may also include interpolating the sample of the video block based on the reference samples.

The invention relates to an advertisement click-through rate prediction method and prediction system based on multi-scale feature fusion. The method comprises the following steps: analyzing a feature tag of an advertisement object; matching a reference advertisement object with the advertisement object by using the feature tag; calculating a predicted click rate of the advertisement object and comparing the predicted click rate with an actual click rate of the reference advertisement object to obtain a comparison result; calculating the error degree of the comparison result, replacing the reference advertisement object according to the error degree, and repeatedly calculating the comparison result; drawing a comparison result curve by using the comparison result and the comparison result ranking; and using the last comparison result in the sequential sequence as the predicted click rate of the advertisement object. According to the advertisement click-through rate prediction method and prediction system based on multi-scale feature fusion disclosed by the invention, the advertisement click-through rate is calculated by using a reference mode, and the reference content is corrected according to the calculation result, so that a more accurate reference sample can be obtained, and more accurate click-through rate data can be obtained based on the reference sample.

The invention provides a refrigerator fault diagnosis method and system based on multi-model integration and adaptive learning, and the method comprises the steps: obtaining historical refrigerator operation parameters, taking the historical refrigerator operation parameters as a reference sample, and obtaining to-be-judged refrigerator operation parameters, and taking the to-be-judged refrigerator operation parameters as a judgment sample; constructing a working condition box based on the current reference sample, judging whether the current judgment sample is within the range of the current working condition box, and if not, deleting the current judgment sample; otherwise, reserving; constructing a refrigerator fault recognizer, and performing fault recognition based on the judgment sample by using the constructed refrigerator fault recognizer; constructing a cold machine power physical grey bin model, and judging the sample to predict the power of the refrigerating machine by utilizing the constructed cold machine power physical grey bin model; a power prediction range is constructed based on the refrigerator power obtained through current prediction; comparing the refrigerator power in the judgment sample with the power prediction range; and whether the current refrigerator is abnormal or not is judged by combining the recognition result of the refrigerator fault recognizer.

The invention discloses a cross-sample interpolation attack and defense method for a depth code model, and relates to a cross-sample interpolation attack and defense method. The invention aims to solve the problems of low adversarial sample generation efficiency and insufficient diversity in the prior art. The method comprises the following steps of 1, extracting multi-category function code samples from a source code corpus, constructing an initial data set and performing preprocessing; 2, randomly selecting a reference sample and a target sample; step 3, extracting different types of interpolation elements from the reference code sample, and adaptively inserting the interpolation elements into a target code sample to generate candidate adversarial samples; 4, proposing an iterative attack algorithm, dynamically optimizing an interpolation strategy, and constructing an efficient adversarial sample of the misleading model through multiple rounds of structural disturbance and confidence judgment; and 5, injecting the generated adversarial sample into the training set, and carrying out model retraining to enhance robustness. The invention belongs to the technical field of information security and artificial intelligence.

An encoding method includes: decoding a bitstream to determine prediction parameters of a current block; determining a first prediction block of the current block according to the prediction parameters; determining model parameters according to a first color component value and a second color component value of the first prediction block; determining a reference sample value of a first color component sample of the current block; determining a prediction value of a second color component sample of the current block according to the reference sample value of the first color component sample of the current block and the model parameters; and determining a reconstructed value of the second color component sample of the current block according to the prediction value of the second color component sample of the current block.

The invention discloses a microscopic imaging system for measuring a second harmonic phase based on polarization interference. In order to overcome the problems of insufficient precision and limited application scene of the traditional interference type second harmonic technology, a material (such as a two-dimensional material single crystal meeting D3h symmetry) with a specific second harmonic polarization response characteristic is creatively adopted as a reference sample, and the reference sample and a sample to be tested construct a coaxial interference light path. The system changes the polarization direction of the reference second harmonic by regulating and controlling the orientation angle of the reference sample, so that a controllable polarization angle difference is formed between the reference second harmonic and a signal to be detected, and implicit phase information is converted into an imaging contrast ratio capable of being quantitatively analyzed. According to the system, in-situ dynamic measurement of the phase can be realized with submicron resolution and high imaging contrast. The method has important application value in the field of characterization of novel semiconductor photoelectric materials and quantum devices, and a standardized technical scheme is provided for defect detection, intrinsic symmetry analysis and domain structure dynamics research of crystals.