1219 results about "Linear transform" patented technology

The invention provides a predictive maintenance management method and system for an industrial robot, and relates to the technical field of equipment maintenance management, and the method comprises the steps: obtaining multi-modal heterogeneous data of the industrial robot, carrying out the intelligent preprocessing, obtaining first preprocessing data, carrying out the cross-modal correlation fusion of the first preprocessing data, generating fusion data, and carrying out the prediction of the industrial robot; extracting complementary redundant information and constructing a semantic feature map; constructing a health state prediction model, adding the semantic feature map into the health state prediction model, mapping the semantic feature map to an embedding space through linear transformation, extracting time sequence information, carrying out feature interaction to obtain interaction features, and carrying out transformation and feature enhancement through a feedforward neural network layer to obtain health state probability distribution; comprehensive health state distribution is generated through Bayesian reasoning fusion, an optimal maintenance strategy is generated in combination with a detection model, an abnormal mode and a key turning point, collaborative maintenance is conducted on the industrial robot, a maintenance result is uploaded to a server, and the current state is stored.

The invention relates to the technical field of data encryption, and discloses a data encryption protection system and method for intelligent controller hardware, and the method comprises the steps: generating a random confusion factor based on a real-time operation state through constructing a dynamic confusion algorithm, and carrying out the nonlinear transformation of original data; implementing a fragment encryption strategy, dividing fragments according to data sensitivity levels, and adopting a heterogeneous encryption algorithm; generating a dynamic binding key in combination with the environment parameters and the hardware unique identifier; hardware-level integrity check is executed, and data and hardware are verified through cyclic redundancy check codes and hardware fingerprints; establishing a self-adaptive key updating mechanism; performing multi-layer defense linkage; and implementing anti-side channel protection, and constructing a security audit log. According to the method, the hardware data security of the intelligent controller is comprehensively guaranteed, various security threats are effectively resisted, and the data confidentiality, integrity and availability are improved.

The invention provides a structured data modeling analysis method based on Transform, and relates to the technical field of data analysis, and the method comprises the steps: carrying out the piecewise linear transformation and adaptive quantum coding of numeric features and category features through a double-flow feature encoder, and carrying out the fusion through a gating mechanism to obtain fusion features; inputting the enhanced training sample into a feature extraction network comprising a hierarchical attention module and a self-calibration module, and extracting global feature representation through a feature affinity graph and a causal attention mechanism; and constructing an information bottleneck feature filter to select a key feature subset, performing prediction and reconstruction by adopting a double-branch network, and analyzing the feature contribution degree by using a structured gradient method. According to the method, heterogeneous features can be effectively processed, the model performance is improved, and meanwhile an interpretable feature importance evaluation mechanism is provided.

The invention relates to the technical field of voice processing, can be applied to business scenes of financial science and technology, medical health and the like, and discloses a voice feature processing method, device, equipment and medium. Performing time resolution analysis based on the fused Mel band energy to generate a multi-scale Mel spectrum amplitude value, and performing nonlinear transformation on the multi-scale Mel spectrum amplitude value according to the noise intensity parameter to generate a noise suppression Mel component; and generating a perception weighting coefficient according to an auditory perception model, and executing frequency domain energy adjustment on the noise suppression Mel component to generate Mel spectrum representation. On the basis of frequency resolution self-adaption, time resolution dynamic adjustment and auditory perception modeling, nonlinear transformation and perception weighting processing are applied to the multi-scale Mel spectrum amplitude value, the influence of noise interference on voice features can be effectively reduced, and the key information retention capacity of voice signals is enhanced.

The invention discloses a coal mine industry multi-dimensional data real-time monitoring and abnormity positioning method and system, and the method comprises the steps: collecting real-time data in coal mine operation, and carrying out the preprocessing; and carrying out dimension construction on the preprocessed data, and fusing the data of different dimensions to form multi-dimensional data. And inputting the multi-dimensional data into the anomaly detection model, and carrying out backtracking analysis on the abnormal data. And performing exception handling and response according to the detected exception and backtracking analysis. According to the coal mining industry multi-dimensional data real-time monitoring and anomaly positioning method and system provided by the invention, correlation analysis and fusion are performed on different-dimensional data, key features are extracted by using principal component analysis, and efficient fusion of the multi-dimensional data is ensured in combination with nonlinear transformation. The main information is reserved, and the accuracy of anomaly detection is improved by enhancing the representativeness of the features. The abnormal response time is shortened, and the emergency processing accuracy is improved.

The invention provides a power load long time sequence prediction method and a related device, and relates to the technical field of power load prediction. The method comprises the steps of collecting power load time sequence data; preprocessing the power load time sequence data; decomposing the preprocessed power load time sequence data into a seasonal component and a trend component through a mobile decomposition method; inputting the trend components into a trend clustering enhancement network, performing channel clustering on the trend components through a K-means clustering algorithm, applying independent linear transformation to each clustering group, and extracting long-term trend features; inputting the seasonal components into a seasonal causal attention network, and extracting short-term seasonal features through a causal mask mechanism and a multi-head attention mechanism; fusing the long-term trend features and the short-term seasonal features to generate fused features; and inputting the fusion features into a decoder, and outputting a power load prediction result through a full connection layer. The change trend of the power load data is fully considered, and the prediction precision of the power load long sequence is improved.

The invention discloses an electric power system short-term load prediction method based on TCN-BIGRU-CBAM, and the method comprises the steps: obtaining a standard load time series data set as input data, and carrying out the preprocessing of the input data; establishing a time convolutional network TCN model, and extracting a long-term dependency relationship in the time sequence; establishing a two-way gating cycle unit network BIGRU, and outputting two-way time characteristics; constructing a convolutional block attention module (CBAM), and obtaining a four-dimensional feature map subjected to channel and space attention enhancement; and establishing a full connection layer, mapping the time sequence characteristics to a scalar through linear transformation, and outputting a power load prediction value of a target time step. According to the method, the problems of information loss, difficulty in high-dimensional feature extraction and the like of a traditional short-term load prediction model in complex time sequence modeling are effectively solved, the precision and the stability of short-term load prediction are remarkably improved, and the method has relatively high practical value and popularization potential.

The embodiment of the invention provides a multi-modal knowledge graph construction method in cross-media retrieval. The method comprises the steps that extracted multi-modal features are mapped to a multi-modal feature space through a linear transformation layer; in the multi-modal feature space, the intra-modal attention weight of each modal feature is calculated according to a self-attention mechanism, the cross-modal attention weight of different modal features is calculated according to a cross attention mechanism, the two weights are fused to obtain a final fusion weight, each modal feature is weighted and input into a graph attention network, and the multi-modal feature is obtained. Obtaining a multi-modal fusion graph structure; performing semantic analysis on each modal feature, matching with a preset multi-modal semantic knowledge base, determining potential semantic association, performing semantic alignment on the multi-modal fusion graph structure according to a preset graph matching algorithm and the potential semantic association to obtain a multi-modal knowledge graph, performing cross-modal data retrieval according to the multi-modal knowledge graph, and performing cross-modal data retrieval according to the multi-modal knowledge graph. The multi-modal data cross-media retrieval method and device can improve the efficiency and accuracy of multi-modal data cross-media retrieval.

The invention discloses an RGB data and event data fusion target tracking method based on visual Mama, relates to the technical field of computer vision, and solves the technical problem that efficient and accurate target tracking of RGB data and event data is difficult to realize. The method comprises the following steps: performing frame-level synchronization and conversion on an original RGB image of an RGB camera and an original event stream of an event camera to generate RGB input data and event input data; obtaining an RGB feature map and an event feature map through a feature extraction module of the Mamba network; performing linear transformation and deep convolution operation through an interaction module to obtain RGB features and event features; dynamically fusing the RGB features and the event features through a fusion module to obtain cross-modal fusion features; and the template features and the search features are aligned through an alignment module to generate an alignment feature graph, and target positioning and tracking are completed. According to the method, through efficient modal fusion and the state space model, the target tracking precision and robustness in a dynamic scene are remarkably improved.

The invention discloses a word embedding vector extraction method and system based on a large language model, and belongs to the field of natural language processing, and the word embedding vector extraction method based on the large language model comprises the following steps: S1, embedding a global context attention module, generating a Q / K / V vector through linear transformation, and calculating a global attention weight in a cross-position manner; s2, splicing coding features of adjacent layers, and performing gating fusion to generate a global context dynamic semantic state; s3, multi-head attention joint coding is carried out on the dynamic semantic state of the current layer and the global state; s4, performing average pooling to generate a global semantic vector, and calculating a dynamic attention weight; s5, performing weighted fusion on the dynamic weight and the global vector to generate an enhanced code; s6, outputting a final word vector by a GELU nonlinear transformation full-connection layer; the method has the beneficial effects that the global context semantic capture capability is enhanced, and the expression effect of the word embedding vector is improved.

The invention provides a cellular network flow space-time prediction method and system based on a hybrid neural network, and belongs to the technical field of intelligent communication. The method adopts a layered deep neural network architecture, and comprises a data embedding layer, a space-time coding layer, a feature fusion layer and an output layer. The data embedding layer maps a historical traffic sequence, cross-domain external data and metadata into high-dimensional features; the space-time coding layer is used for respectively fusing one-dimensional causal convolution and a Mama neural network to extract multi-scale time features and densely connecting convolution and a multi-head attention mechanism to capture multi-scale space features through time and space modeling branches; the feature fusion layer realizes adaptive weighted fusion of spatial-temporal features, cross-domain features and metadata features by using a gating fusion mechanism; and the output layer performs linear transformation on the fusion features to generate a final prediction result. According to the method, the spatial-temporal dynamic capture of the service traffic is accurate, the prediction curve is highly fit with the true value, and the accurate prediction of the multi-service traffic of the cellular network is realized.

A method and system for optimizing compute runtime and memory footprint of a linear transformation process are provided. The method includes determining a set of optimal rotation parameters, wherein the optimal rotation parameters provide an optimal tradeoff between runtime compute resources and a memory footprint for a runtime execution of the linear transformation process; initializing the linear transformation process to run a boosting technique with the determined set of optimal rotation parameters, wherein the boosting technique, when executed at runtime as part of the linear transformation process, performs at least one iteration that yields rotated ciphertexts, and wherein the at least one iteration is based on the determined set optimal rotation parameters and at least one key switching key (KSK); and loading the initialized linear transformation process to an internal memory of a hardware accelerator.

The invention discloses a dynamic feature retrieval generation and management method based on a cross attention mechanism, which belongs to the technical field of natural language processing, and comprises the following steps: step 1, converting knowledge base document fragments into atomic knowledge units, each atomic knowledge unit comprising question and answer pairs, codes as key value pairs, and adding dynamic priority weights; self-attention query is replaced with a double-channel query structure, one channel is used for cross attention, a cross attention query vector is generated through linear transformation, and key value pairs of atomic knowledge units are used; and step 3, training a cross attention adapter, freezing the weight of the language model, optimizing parameters of the adapter, and dynamically adjusting a loss function by using pre-judgment parameters based on input sequence context complexity. By means of the method, deep correlation description of user input and knowledge fragments can be achieved, semantic ambiguity is eliminated, knowledge injection and context information are balanced, distortion is avoided, and the strict requirement for semantic precision in the professional field is met.

The invention provides a fault detection method based on the combination of a Transform model of VMD and a BiGRU, and the method comprises the steps: firstly collecting voltage and current signals, and carrying out the preprocessing of the voltage and current signals; then, decomposing the preprocessed signal into a plurality of modal components by using VMD, and performing normalized coding to form a feature vector; the feature vectors are then input into a Transform model, deep features are extracted, and residual connection and layer normalization processing are carried out; the output of the Transform model is input into the BiGRU network, time sequence change characteristics are captured, and a characteristic vector with a fixed length is output; and finally, inputting the output of the BiGRU network into a softmax classifier, outputting probability distribution through linear transformation and a softmax function, and selecting a fault type with the maximum probability as a final result. And if the fault type is not a normal state, generating early warning information and giving an alarm. The method has the advantages of high-precision fault detection, strong robustness, strong adaptability and good real-time performance, and can effectively improve the accuracy and efficiency of fault diagnosis of the power system.

According to the motor anomaly detection model training method and the motor anomaly detection method provided by the invention, a plurality of audio samples and category labels thereof during motor operation are acquired, a plurality of disturbance audios are generated through speed disturbance for each audio sample, the disturbance audios are input into a frequency domain feature extraction sub-model and a time domain feature extraction sub-model, and the frequency domain feature extraction sub-model and the time domain feature extraction sub-model are extracted. Performing linear transformation and weighted summation on the enhanced wavelet time-frequency spectrum and the enhanced time-domain features through a full connection layer to generate fusion enhanced features, performing feature transformation on the fusion enhanced features through a deep neural network, inputting the fusion enhanced features into a classifier to obtain a classification probability, and calculating triple loss according to a category label and the classification probability; and updating parameters of the frequency domain feature extraction sub-model, the time domain feature extraction sub-model, the full connection layer, the deep neural network and the classifier through loss back propagation and a stochastic gradient descent algorithm to obtain a trained motor anomaly detection model. Therefore, the accuracy of motor state recognition of the model in a high-noise environment is improved.

The invention discloses an acute pancreatitis complication assessment method based on artificial intelligence, and belongs to the technical field of medical information, and the method specifically comprises the steps: firstly, obtaining a clinical data set containing real-time clinical indexes and basic disease historical records; performing time sequence analysis on the historical records of the basic diseases, extracting long-term influence characteristics of the historical records of the basic diseases and generating basic disease influence factors; carrying out standardization processing on the real-time clinical indexes and the basic disease influence factors, and generating a comprehensive feature vector after calculating association weights among features; inputting the vector into a trained artificial intelligence evaluation model, and outputting a complication risk probability through layer-by-layer nonlinear transformation; and finally, mapping the risk probability value into a risk level, and integrating patient information to generate a structured risk assessment report. And an acute pancreatitis complication evaluation scheme integrating the acute stage index and the chronic basic disease influence is established.

The invention provides a machine vision-based panoramic look-around auxiliary driving method and system, and the method comprises the steps: employing a plurality of fisheye cameras to obtain a plurality of first images around a special vehicle, obtaining a distortion model of the fisheye cameras, employing a polynomial model correction method to correct the first images through the distortion model to obtain second images, and obtaining the second images; a direct linear transformation algorithm is used to transform the pitch map of the second image to obtain third images, a weighted average algorithm is used to eliminate splicing gaps of all the third images to obtain a panoramic look-around image, a YOLOv5s-based multi-task perception algorithm is used to detect the target of the panoramic look-around image, the target comprises a drivable area segmentation result and a lane line segmentation result, and the target comprises a driving area segmentation result and a lane line segmentation result. And displaying the panoramic look-around image and the target in the panoramic look-around image. The method can completely display the fusion image of the virtual vehicle, the vehicle bottom and the surrounding road surface, intelligently recognizes the road scene information, eliminates the view blind area, and enhances the view of the driver of the special vehicle.

Disclosed in the present invention is a deep learning-based power distribution network load prediction method, comprising: acquiring regional load data and renewable energy power generation data to form a data set, and preprocessing the data set to obtain a first data set; using a convolutional neural network to extract a time series feature in the first data set, and converting the time series feature into a data form of a deep learning model by means of an embedding layer to obtain one-dimensional time series data; performing fast Fourier transform on the one-dimensional time series data to obtain a frequency curve, extracting amplitude values on the frequency curve to calculate corresponding periods, and selecting the corresponding periods to slice the one-dimensional time series data and form same into two-dimensional matrixes; using a two-dimensional convolutional network to perform feature extraction, reshaping the two-dimensional matrixes that have undergone feature extraction into one-dimensional arrays, and performing adaptive fusion on the one-dimensional arrays to obtain a first time series feature; and inputting the first time series feature into a fully connected layer for weight calculation and linear transformation to obtain a power distribution network load prediction result, thereby improving the stability and reliability of electric power supply.

The invention discloses an OFDM-LFM (Orthogonal Frequency Division Multiplexing-Local Frequency Modulation) integrated signal waveform receiving method and equipment based on fractional Fourier transform, LFM and OFDM are combined and improved, an OFDM-LFM integrated radar communication signal taking an orthogonal LFM signal as a subcarrier is adopted, and as the LFM subcarrier has larger Doppler tolerance, compared with a normal subcarrier, the OFDM-LFM integrated signal waveform receiving method and equipment based on fractional Fourier transform have higher Doppler tolerance, the radar communication signal can be received more accurately. The method not only can improve radar distance resolution and spectral efficiency, but also has a strong capability of avoiding Doppler effect, and can reliably transmit information, and when OFDM-LFM signals are received, in order to solve the problems of subcarrier frequency linear transformation, radar parameter extraction and the like, based on the sensitivity of fractional Fourier transform (FRFT) to LFM signals, the method can be used for solving the problems of subcarrier frequency linear transformation, radar parameter extraction and the like. According to the invention, the FRFT is used as the integrated signal receiver, the separation of the integrated signal communication receiver and the radar receiver is realized, and the system performance is improved while the complexity of the system is reduced.

The invention discloses a recommendation method fusing a knowledge graph and sentiment analysis, which comprises the following steps of: constructing a data set, processing comment text data in the data set to obtain a triple, processing the triple by adopting linear transformation to obtain a user embedded vector and a project embedded vector, and constructing the user embedded vector and the project embedded vector. Coding the comment text data through a pre-training language learning model to obtain a probability distribution value of sentiment analysis, fusing the user embedding vector and the item embedding vector, and connecting the obtained context representation of the comment text data with the corresponding probability distribution value of sentiment analysis to obtain a matching score of the user and the item, and performing matching prediction verification on the matching score of the user and the item to obtain the accuracy of the matching score of the user and the item. According to the method, sentiment analysis is introduced, and the sentiment analysis can help to identify the sentiment attitude of the user to the item or the content, so that the preference of the user is more accurately captured, and the matching degree of recommendation and the satisfaction degree of the user are improved.

The invention discloses a model training method and device based on a reattention mechanism. In the scheme, an embedded vector sequence is input into a multi-head attention layer of a pre-training language model to extract global semantic features, and a global context vector is generated based on the global semantic features; based on a sparse Softmax function and the global context vector, a sparse gating network calculates dynamic weights corresponding to multiple attention heads, and constructs an attention weight vector; performing weighted fusion on the context vectors output by the attention heads and the corresponding attention weight vectors by a multi-head attention layer, and performing linear transformation on the fused context representation based on a fixed weight matrix to obtain a self-attention value; and adjusting parameters of the pre-training language model according to the self-attention value through a preset loss function to obtain a target pre-training language model. According to the method, the sparse gating mechanism is introduced to dynamically adjust the weight of the multiple attention heads, and the model prediction precision is remarkably improved.

The invention discloses an attention mechanism optimization method and device, a storage medium and a processor. In the scheme, rotation position coding is carried out on a Query vector and a Key vector which are obtained through linear transformation of an input sequence, and an original attention value is obtained; based on the relative distance of each position in the input sequence, constructing a bias item; the bias item comprises a sine bias item; and carrying out dynamic fusion on the original attention value and the offset item, and outputting an optimized attention value. Compared with the position coding technology in the prior art, the method has the obvious advantages that the performance of the model is obviously reduced when the model is used for processing the super-long sequence, and the extrapolation is insufficient.

The invention discloses a lithium ion battery capacity inflection point prediction method and system based on a multi-parameter data fusion decision. The method comprises the following steps: collecting multi-parameter data of a lithium battery and preprocessing the multi-parameter data; constructing an inflection point prediction model based on deep learning, and extracting time sequence data characteristics of current, voltage and temperature; based on an attention-enhanced graph convolutional neural network AGCN, an attention mechanism is introduced into a graph convolutional neural network GCN to dynamically learn the association weight of a multi-parameter feature matrix, and multi-parameter data fusion features are obtained; dynamic decision making is carried out on the battery multi-parameter data fusion features, linear transformation is carried out on a dynamic decision making result to obtain a predicted value of an inflection point, and construction of an inflection point prediction model is completed; carrying out training optimization on the whole model, and predicting the residual cycle period of the battery to the inflection point; according to the method, inflection point high-precision prediction of any stage of the battery can be realized by depending on relatively short cycle period data.

The invention provides a feature fusion-based machine-generated text traceability method and a feature fusion-based machine-generated text traceability system aiming at the problem that a single feature in the existing machine-generated text traceability technology is difficult to accurately trace the source of a machine-generated text, a new machine-generated text traceability model is constructed, and the traceability of the machine-generated text is improved by combining statistical features and semantic features of the text. And richer feature representation is provided for the model. The method comprises the following steps of: firstly, converting an input text into a semantic vector by using a word embedding technology, and calculating a prediction probability, a cumulative probability and an information entropy of the text by using a GPT-2 model to serve as a statistical vector of the text; then, further extracting features through GRU and Transform encoders, performing linear transformation on the features through a linear layer to obtain semantic features and statistical features of the text, and integrating the semantic features and the statistical features through a feature fusion strategy; and finally, after the fused feature vectors are analyzed by a classifier, outputting a specific source of the text.

The invention relates to the technical field of image recognition, in particular to a remote sensing image segmentation method and system, a storage medium and equipment. Taking the preprocessed remote sensing image as a real sample, and obtaining a generated sample by using a conditional diffusion model; respectively converting the real sample and the generated sample into a set color space, calculating the color value of each channel pixel by pixel, determining the color distribution of the real sample, aligning the color distribution of the generated sample with the color distribution of the real sample through linear transformation, and obtaining the generated sample after color correction; and performing similarity screening on the generated samples after color correction, mixing the generated samples with real samples according to a set proportion, taking a formed mixed sample set as a training set of a semantic segmentation model, and performing training to realize remote sensing image segmentation. By combining the conditional diffusion model and color correction, the spectrum consistency of the generated samples is improved, the quality of the generated samples is optimized, and by means of mixed sample training, the requirement for manual annotation is reduced, and the precision of remote sensing image interpretation is improved.

The invention relates to a ginkgo leaf canopy near-earth remote sensing image super-resolution reconstruction method, which comprises the following steps: (1) collecting ginkgo multispectral remote sensing images at different heights, and establishing a data set by aligning real LR images and HR images; (2) recovering a high-resolution image from the low-resolution remote sensing image by adopting a super-resolution reconstruction model; the super-resolution reconstruction model RTGAN takes an SRGAN model as a baseline model; the RTGAN is composed of a generator network and a discriminator network, an MDRB composed of a plurality of continuous residual blocks is integrated in the generator network, and a U-Trans discriminator is used as a discrimination network; the generator network maps the LR image into an SR image through multilayer nonlinear transformation and feature extraction; the discriminator network evaluates the quality and authenticity of the generated image, and distinguishes a real HR image and a generated SR image; finally, the discriminator network cannot distinguish the two images, and a high-quality SR image is obtained.

The invention relates to an image recognition method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: acquiring a to-be-detected image and an image recognition model; the image recognition model comprises an extraction unit, a gating unit, a plurality of expert units and a feature fusion unit; performing feature extraction on the to-be-detected image based on the extraction unit to obtain an optimized feature map, and determining a scheduling coefficient based on the gating unit and the optimized feature map; performing linear transformation on the optimized feature map according to the plurality of expert units to obtain a mapping feature representation corresponding to each expert unit; determining prompt features corresponding to the optimized feature map according to the scheduling coefficient and the mapping feature representation, and performing feature fusion on the prompt features and the optimized feature map based on a feature fusion unit to obtain feature lexical elements; and determining an identification result of the to-be-detected image based on the feature lexical elements. By adopting the method, the accuracy of image recognition can be improved.

The invention discloses a model reasoning method and device. The method comprises the following steps: preprocessing an input text to convert the input text into a semantic vector sequence; calculating a compression key and a compression value matrix corresponding to the semantic vector sequence through dimension reduction mapping and key-value mapping, and storing the compression key and the compression value matrix into a preset cache space; in response to the fact that the loop end condition is not met, the following steps are executed in a loop mode: based on the target semantic vector and first linear transformation of a compression key and a compression value matrix stored in a preset cache space, current attention representation is calculated; generating a new semantic vector based on the current attention representation and updating a semantic vector sequence; and through dimension reduction mapping and key-value mapping, calculating a compression key vector and a compression value vector corresponding to the new semantic vector, and updating a compression key matrix and a compression value matrix. According to the model reasoning method, the reasoning efficiency can be remarkably improved while the model reasoning effect and performance are ensured.

The invention provides a building identification method and system based on three-dimensional point cloud, and particularly relates to the technical field of computer vision and three-dimensional target identification, and the method comprises the steps: carrying out the undersampling, oversampling and data enhancement preprocessing of inputted building roof point cloud data, and carrying out the T-net network space alignment to output the aligned data. Then, an NEWPCT model is obtained by improving a PCT model, aligned data are input into the NEWPCT model, a point cloud feature matrix is generated through a linear coding layer, query, key and value matrixes are obtained through linear transformation, multi-head attention output is calculated by introducing an offset matrix, intermediate correlation features are obtained after splicing and fusion, and a global feature vector is generated through global maximum pooling; and finally, inputting the vector into a classifier, and outputting a building roof type, and the method solves the technical problems of geometric transformation sensitivity and local feature loss while improving the precision and robustness of three-dimensional point cloud building identification, thereby effectively improving the identification precision and robustness.