205results about How to "Accurate representation" patented technology

The invention discloses a contract evaluation method, system and device based on a large model and a medium, and the method comprises the steps: obtaining a target contract, carrying out the data preprocessing, obtaining the preprocessed contract data, and inputting the preprocessed contract data into a preset contract large language model; the preset contract large language model stores a vector knowledge base including coastal bulk cargo transportation knowledge, legal clauses and industry regulations; in a preset contract big language model, performing contract classification and contract summary extraction on the current preprocessed contract data by using a first model agent to obtain contract summary information; performing contract risk assessment on the current preprocessed contract data to obtain risk assessment information and / or risk avoidance suggestion information; performing benefit measurement and calculation on the current preprocessed contract by using a third model agent to obtain benefit measurement and calculation information; according to the method, the examination and evaluation of the large model in the coastal bulk cargo transportation contract are realized, and the purposes of saving the labor cost and improving the business turnover rate are achieved.

The invention belongs to the technical field of computer vision, provides a real-time semantic perception system running on equipment, performs real-time instance segmentation and semantic mapping on a working environment, and provides a perception method based on lightweight real-time instance segmentation and semantic mapping. In the first part, firstly, a color image containing color and texture information is used for object classification and bounding box regression; cutting the original depth image into a depth image with an accurate geometric edge, and inputting the depth image into the truncation network to generate a mask of the instance; a second part: (1) calculating the attitude and size of a target object in the view of the camera through principal component analysis; (2) obtaining the category of the object in the view obtained by segmentation; (3) creating a point cloud of the depth image projection of the target object in a map; and then cross-frame tracking and fusion are carried out by using data association to realize the purpose of sensing. The method is mainly applied to automatic sensing occasions.

The invention discloses a TCBL wind field high-resolution LES simulation method facing gray area parameterization, a medium and a program product. According to the method, a TCBL idealized initial gradient wind field is constructed based on a modified Rankine vortex model, controllability and stability of initial conditions are ensured by specifying key kinetic parameters, and thermodynamic profiles under different wind speed conditions are obtained based on synthetic analysis of tropical cyclone sounding data. And correcting the bottom of the boundary layer by adopting a quadratic and logarithmic function fitting method. Through high-resolution grid configuration and dynamic boundary conditions, secondary grid turbulence is processed in combination with an SGS-TKE closed model, and adaptability of the model to strong rotation and high wind shear environments is optimized by adopting a dynamic correction mechanism. Besides, the stability of thermodynamic variables is maintained through large-scale tendency correction, and high-resolution three-dimensional wind field and turbulence statistical reference data are output and used for verifying and improving a turbulence parameterization scheme of the TCBL under gray region resolution.

The invention discloses a daily-scale corn water stress diagnosis method based on an unmanned aerial vehicle technology, and relates to the technical field of agriculture, the method is based on the definition of a water stress coefficient, introduces a Jarvis stomatal conductance model, combines canopy and plant height data obtained by an unmanned aerial vehicle, accurately estimates the aerodynamic impedance and the canopy impedance of a canopy, and improves the accuracy of the canopy stress. The actual evapotranspiration and the potential evapotranspiration are obtained through a multispectral leaf area index inversion model, then a daily scale moisture stress index is calculated, a plurality of groups of unmanned aerial vehicle thermal images are obtained according to four important growth periods of corn, and the moisture stress index is calculated and counted. And comparing the characterization capabilities of the daily scale moisture stress index and the moisture stress index in different time periods in one day on the actually measured moisture index, so as to improve the problems that the instantaneous moisture stress coefficient is insufficient in time representativeness and is easily influenced by the environment, and more accurately characterize the crop moisture stress. A monitoring effect close to a daily-scale moisture stress index is obtained by a easier method, and the practicability of the index is improved.

The invention belongs to the field of engineering technology risk management, and provides a tunnel and underground engineering safety risk multi-identification intelligent sensing method. The implementation of the method comprises the following steps: firstly, acquiring multi-modal data on a tunnel and an underground engineering structure by using a multi-identification method; constructing and training an engineering safety risk prediction model; and performing dynamic quantitative evaluation on the tunnel and underground engineering safety risk by using the trained engineering safety risk prediction model. According to the method, the multi-recognition perception technology and the artificial intelligence technology are combined, safety risk information in the tunnel and underground engineering can be observed, the method can be applied to safety risk management in the tunnel and underground engineering, the safety risk level of the tunnel in the construction period and the operation period is evaluated, and the safety risk management efficiency is improved. And the safe state of the tunnel and the underground engineering can be expressed quickly and accurately.

The invention belongs to the technical field of deep learning, and discloses a drug and protein interaction prediction model construction method based on multi-modal fusion. Due to the lack of protein three-dimensional data, the three-dimensional data of the protein is subjected to high-accuracy complementation by using a prediction model AlphaFold 2, and for the modal fusion problem, the drug molecules are fused at the atomic level with finer granularity, so that the multi-modal fusion effect is further enhanced, and the multi-modal fusion effect is improved. Protein 1D data and protein 3D data are processed by using Cross-Attention, the processed protein 1D data and protein 3D data are subsequently fused by using Cross-Attention, multi-modal fusion is performed by using Cross-Attention, so that the interaction relationship between modals can be carefully captured, and in a word, the DTI task effect is improved by designing a multi-modal fusion mode.

The invention discloses a method and system for representing shale oil sample occurrence state based on nuclear magnetic resonance, and relates to the technical field of offshore shale oil exploration, and the method comprises the following steps: respectively obtaining a regular plunger shale sample and an irregular shale sample at the same sampling position in a target stratum; performing a nuclear magnetic resonance experiment on the irregular shale sample to obtain a nuclear magnetic signal of the irregular shale sample, performing inversion to obtain a T2 spectrum of the irregular shale sample, and determining an initial nuclear magnetic porosity; carrying out different equal cutting on the regular plunger shale sample, and sequentially carrying out nuclear magnetic resonance experiments to obtain a target fitting relationship between different core surface areas and nuclear magnetic porosity; and determining a porosity correction coefficient based on the target fitting relationship and the initial nuclear magnetic porosity, and determining the porosity of the irregular shale sample to be measured based on the porosity correction coefficient. According to the method, the technical problem that the fluid occurrence state of the irregular shale sample is difficult to characterize due to scarcity of the shale oil sample on the sea in the prior art is solved.

The invention discloses a rock pore three-dimensional reconstruction method and system based on an adaptive hierarchical cube algorithm, and a storage medium, relates to the technical field of rock structure measurement, and aims to solve the problems of low extraction execution efficiency, ambiguity and poor reconstruction precision of an existing moving cube algorithm. According to an image segmentation model based on RefineNet, the model adopts three 3 * 3 convolution cascading to replace 7 * 7 convolution, and a coordinate attention mechanism is introduced behind each block of an encoder to improve the segmentation precision and efficiency of a key area; dividing the cube into tetrahedrons by adopting a saddle point-adaptive layered downward subdivision method, solving intersection point coordinates and normal vectors, and constructing a triangular patch index; determining the contour surface shape penetrating through the cube by adopting a cube lookup table, and merging triangles of all tetrahedrons to form a complete contour surface; and creating a new triangular patch by using the vertex of the edge of the hole and the newly generated vertex, and repairing the hole to obtain the accurately reconstructed three-dimensional hole of the rock.

The invention relates to the technical field of graph neural network and heterogeneous graph representation learning, in particular to a heterogeneous graph node classification method based on relation perception label propagation, which comprises the following steps of: firstly, according to a given heterogeneous network, carrying out type-specific linear transformation on different types of nodes, and then carrying out type-specific linear transformation on different types of nodes; the method comprises the following steps of: designing a relation sensing label propagation algorithm for target nodes in a heterogeneous network, generating a pseudo label through a relation sub-graph, performing information transmission by adopting a double-layer aggregation strategy based on a type attention mechanism, and effectively fusing multi-layer information of the nodes, so as to obtain a multi-layer information fusion model; and the node feature representation is richer and more accurate. According to the method, the pseudo labels are obtained by designing a relation perception label propagation method, heterogeneous message passing is carried out by utilizing double-layer aggregation based on a type attention mechanism, and the classification performance and robustness are remarkably improved in combination with a multi-objective optimization strategy.

The invention belongs to the technical field of fuel cells, and particularly relates to a fuel cell centrifugal air compressor irreversible loss prediction method and system, and the method comprises the steps: collecting performance data based on a fuel cell centrifugal air compressor comprehensive performance test platform, and building and verifying a multi-physics coupling CFD simulation model; defining an irreversible loss type and related structure parameters, constructing an integrated CFD simulation framework, generating a parameter combination table through an orthogonal test, and performing simulation calculation; an RL-GHEM method is provided, the weight and combination strategy of a base model are optimized, linear and nonlinear combination modes are dynamically switched, and irreversible loss prediction models under different working conditions are constructed. The hybrid integrated model based on reinforcement learning guidance is provided, the model combination strategy is dynamically optimized through the reward mechanism, the base model with the optimal performance is optimized, the prediction precision, stability and robustness are improved, precise prediction of the irreversible loss of the air compressor is achieved, and the problem that in the prior art, the prediction precision is insufficient under variable working conditions is solved.

The invention relates to the technical field of medical endoscope imaging, and discloses an endoscope image correction method and electronic equipment, and the image correction method comprises the steps: obtaining an original image collected by an imaging unit, the original image comprises a first color channel for representing first human tissue characteristics and a second color channel for representing second human tissue characteristics; performing edge attribute marking on pixels in the original image according to the first color channel and the second color channel to obtain a plurality of to-be-processed pixels with edge attribute marks; performing color contrast correction on the to-be-processed pixels according to the edge attribute mark corresponding to each to-be-processed pixel and the relationship between the edge attribute marks to obtain a corrected image; according to the invention, in the specific illumination mode, the colors corresponding to the to-be-processed pixels representing the first human body tissue characteristics and the second human body tissue characteristics are enhanced, so that the color differentiation between different tissues in the specific illumination mode can be highlighted.

The invention discloses a multi-state-based power grid information analyzing and checking processing device and method. According to the ratio of unreasonable amplitudes of reference time points on the front side and the rear side of each lost time point, the voltage value of an urban power grid at the front target reference time point and the voltage value of the urban power grid at the rear target reference time point, the unreasonable amplitudes of the reference time points are calculated; obtaining a complementation value of the voltage value of the initial urban power grid at each missing time point; acquiring a complementation value of the voltage value of the last urban power grid at each missing time point, and performing complementation analysis on the voltage value of the urban power grid at the missing time point by combining the variation attribute of the voltage value of the urban power grid at the sampling time point around each missing time point with the variation attribute of the current value of the urban power grid; according to the method, a complete urban power grid information queue is obtained, so that when the Savitzky-Golay method is used for performing clutter value removal on the urban power grid information queue, disturbance of a missing time point is avoided, and the urban power grid information after clutter value removal can accurately represent the urban power grid information.

The invention relates to a point cloud quality assessment method based on depth ordinal regression. The method comprises a feature fusion step based on a symmetric cross-modal attention mechanism and an ordered regression step of point cloud quality assessment. In the feature fusion step, point cloud features are extracted through farthest point sampling and a K nearest neighbor algorithm, point cloud projection is performed to generate an image, image features are extracted, the two features are fused by using a cross-modal symmetric attention mechanism, and global feature representation is generated; in the step of ordered regression, a depth ordinal number loss function is designed to optimize network parameters, the prediction probability is converted into a continuous point cloud quality score by using ordered information of the quality score, a step effect is prevented, and an evaluation result is closer to a true value. Through feature fusion and ordered regression, the accuracy and stability of point cloud quality evaluation are effectively improved, and the evaluation result is closer to a true value.

The invention relates to the field of image clustering analysis, in particular to an image clustering method and device based on a particle multi-kernel k-mean value and electronic equipment. The method comprises the following steps: acquiring a discrete data point set of a plurality of to-be-clustered images; performing granular ball division on the image data to obtain an image granular ball set; calculating to obtain an image pellet center, an image pellet distance matrix and an average distance between image pellets; calculating to obtain an image pellet kernel matrix; according to the image pellet kernel matrix and a preset clustering division matrix, constructing a multi-kernel k-means clustering objective function based on pellets; calculating to obtain a corresponding target clustering division matrix when the target function value is minimum; discretizing the target clustering division matrix according to a multi-kernel k-mean algorithm to obtain a clustering result of image pellets; and distributing the clustering result of the image pellets to all discrete data points contained in the image pellets to obtain the clustering result of the image data. According to the invention, the efficiency and accuracy of image clustering are improved.

The invention discloses an unmanned aerial vehicle navigation method, device and equipment and a storage medium, and the method constructs an LIEKF algorithm on a world coordinate system (local coordinate system), enables the value to be small, and guarantees the stability of a linear error kinematics matrix and an observation matrix. A left invariant error on a matrix Lie group is defined under a world coordinate system, and the error is kept unchanged on a left group. According to the method, state quantities such as attitude, position and speed and errors are defined in a group space, and a linear state space model independent of the state is deduced through multiplication closure of the Lie group and affine characteristics between Lie algebra and corresponding nonlinear error quantities constructed by the multiplication closure and the Lie algebra. Because the redefined speed error and position error both comprise attitude items, the real error can be represented more accurately.

The invention discloses a multi-dialect speech recognition method, which relates to the technical field of language processing and is used for solving the problem of difficulty in multi-dialect recognition in the prior art. Extracting features of the dialect data, and representing the features as voice representation; inputting the voice representation into a Transform encoder based on a sparse bias calculation mode to extract a dialect representation; performing dialect classification according to the dialect representation information to obtain a dialect classification result; and inputting the classification result and the voice pointer into a multi-expert integrated system for prediction to obtain a voice recognition result. The invention further discloses a multi-dialect speech recognition device, electronic equipment and a computer storage medium. According to the method, accurate and efficient identification of multiple dialects is realized by combining a multi-expert integrated system based on a sparse bias calculation mode encoder.

The invention relates to the technical field of airplane windshield wiper control, in particular to an airplane windshield wiper rotation angle control device and method, and the method comprises the steps: obtaining an initial windshield wiper vibration signal of a to-be-controlled airplane in a current windshield wiper cleaning process, decomposing the initial windshield wiper vibration signal, and screening out a target current vibration signal; determining a rotation error factor corresponding to the complete cleaning process of each windshield wiper; determining vibration abnormal factors corresponding to each current sampling moment and each historical sampling moment; determining a current error accumulation factor; according to the vibration abnormal factors corresponding to all the current sampling moments and the current error accumulation factor, the PID control quantity at the current moment is adjusted, a target PID control quantity is obtained, and turning angle control is conducted on a windshield wiper on the to-be-controlled aircraft. According to the invention, the rotation angle control of the windshield wiper is realized, and the reasonability of PID control quantity adjustment and the reasonability of rotation angle control of the wiper are improved.

The invention discloses a structural plane network sequence modeling method and system suitable for layered rock mass, and relates to the technical field of rock mass engineering, and the method comprises the steps: obtaining layered rock mass outcrop data as basic data, and obtaining basic data interpretation representation after processing; determining a rock mass outcrop homogeneous region division result according to the type attribute of the basic data; according to the interpreted occurrence data in combination with a rock mass outcrop homogeneous region division result, correcting the occurrence data in each homogeneous region by using a weighting function, and grouping the structural surfaces by using a K-mean particle swarm algorithm to obtain grouped structural surfaces; determining the density of the distribution space of each group of structural surfaces according to the spacing, occurrence, size and survey line trend data of each group of structural surfaces; determining the position and size of a layer in the layer structure space according to the determined distribution position of the structural surface; and randomly combining parameters of each group of structural surfaces, and constructing a stratified rock mass structure spatial network model in combination with sequence modeling and Monte Carlo simulation methods.

The invention discloses a time-frequency analysis method for processing a multi-component fast-changing time-frequency cross signal, relates to the technical field of signal processing, is suitable for processing and analyzing the fast-changing time-frequency cross signal, and can realize accurate characterization and processing of the fast-changing time-frequency cross signal under a strong noise condition. According to the technical scheme, the method comprises the steps that S1, an EMCT is used for processing collected time-frequency cross signals, and an EMCT result of the signals is obtained; and S2, establishing a new frequency redistribution operator, processing an EMCT result of the signal, obtaining an intermediate frequency track of the signal, and carrying out adaptive time-frequency cross ridge extraction. And S3, establishing a time-frequency coefficient compression method for a time-frequency cross signal, and processing an EMCT result of the signal in combination with the obtained intermediate-frequency trajectory to obtain a time-frequency result. And S4, in the generated time-frequency result, extracting a target component of the signal by using an intermediate-frequency track, and representing and reconstructing a time-frequency cross signal.

The embodiment of the invention discloses a business type determination method and device, electronic equipment, a medium and a program product, and can be applied to the technical field of artificial intelligence, such as the field of machine learning. According to the embodiment of the invention, subject business data of an operation subject is acquired, the subject business data comprises subject attribute data and transaction data, and the transaction data comprises behavior data and position data; according to an association relationship between the behavior data and the position data, classifying the behavior data and the position data to obtain target behavior data corresponding to the specified transaction mode and transaction position distribution; determining a transaction distribution rule according to the target behavior data and the transaction position distribution; and determining the business type of the operation subject according to the transaction distribution rule and the subject attribute data. According to the scheme, the real business type of the business subject can be determined based on the multi-dimensional business data, so that a business platform can conveniently manage and supervise the business subject.

The invention discloses a path planning method, system and device for a complex terrain and a storage medium. The method comprises the following steps: acquiring topographic data of a complex terrain through a laser radar, and constructing a map model based on an irregular triangular network; calculating the topographic cost of each triangle according to the topographic features of the map model, and converting the topographic cost of each triangle into the total topographic cost; and inputting the total terrain cost of each triangle into a Double DQN algorithm, and planning an optimal path. By combining the three-dimensional terrain modeling technology and the reinforcement learning algorithm, efficient, accurate and real-time path planning can be realized in a complex three-dimensional environment, the limitation of the prior art is overcome, and the development of the autonomous navigation technology of the robot is promoted.

The invention belongs to the technical field of open domain multi-modal classification, and provides an open domain multi-modal identification method for aviation safety hot events. According to the method, the text features and the image features of the aviation safety hot events are extracted, meanwhile, a memory enhancement module is introduced, a scene guide pseudo event filter is designed, final classification is obtained by combining source events and pseudo events, and more effective help is provided for coping with and managing the aviation safety hot events. Compared with the prior art, the multi-modal aviation security hotspot event classification task in a complex scene can be effectively solved. Meanwhile, the influence of unknown events on classification is considered, known events are distinguished from the unknown events, and the problem that the classification precision is reduced by the unknown events is avoided.

The invention discloses a method for representing low-temperature reduction degradation performance of pellets. The method comprises the following steps: (1) carrying out low-temperature reduction degradation reaction on the pellets at 400-600 DEG C; (2) the reduction expansion rate of the pellets after the reaction is detected, and the crack morphology of the pellets after the reaction is analyzed; and (3) the detected reduction expansion rate and crack morphology are jointly used as characterization of the low-temperature reduction degradation performance of the pellets. According to the method, the reduction expansion rate of the pellets at 400-600 DEG C is adopted, so that the method is more suitable for representing the low-temperature reduction degradation performance of the pellets with regular shapes; meanwhile, the pulverization process and mechanism of the pellets after the reaction can be intuitively reflected by combining different section crack shapes of the pellets after the reaction; the pellet low-temperature reduction degradation performance can be accurately represented. The method has the characteristics of simplicity in operation, easiness in implementation and high accuracy, and can better replace the traditional low-temperature reduction degradation performance index.

The invention relates to the technical field of lithium metal battery health monitoring, and particularly discloses a lithium metal battery pack health monitoring system and method based on thin film sensors, the system comprises a plurality of thin film sensors distributed among lithium metal battery pack cells, and each thin film sensor comprises a temperature measuring unit, a pressure measuring unit, a flexible circuit board and silica gel foam; the adapter plate is electrically connected with the plurality of thin film sensors; the signal acquisition unit is electrically connected with the adapter plate and comprises a constant current source circuit, a signal conditioning circuit, a filter circuit and an analog-to-digital conversion circuit; the input end of the data transmission unit is electrically connected with the output of the analog-to-digital conversion circuit; the input end of the digital signal processor is electrically connected with the output end of the data transmission unit; and the input end of the battery management unit is electrically connected with the output end of the digital signal processor. The mechanical information and the temperature information between the battery monomers can be monitored at the same time, the battery management can evaluate the health state of the battery more comprehensively based on the information, and the abnormal problem can be warned in advance.

The invention relates to an ionosphere TEC prediction method in a near space, and belongs to the technical field of ionospheres. The method comprises the following steps: acquiring ionosphere environment data at three consecutive moments; wherein the ionosphere environment data comprises ionosphere TEC data and earth magnetic field data; inputting the ionosphere environment data at the three moments into the trained ionosphere TEC prediction model to obtain ionosphere TEC data at the next moment; wherein the ionosphere TEC prediction model comprises an encoder, a condition generator and a decoder; the encoder is used for encoding the ionized layer TEC data at the current moment to obtain potential representation of the ionized layer TEC data; the condition generator is used for performing inversion coding on the ionosphere environment data at the three moments to obtain ionosphere electron density information; and the decoder is used for reconstructing the ionospheric TEC data at the next moment from the potential representation of the ionospheric TEC data at the current moment by using the ionospheric electron density information. The state of the ionosphere is predicted through inversion of the electron density information of the ionosphere.

The invention discloses a streamline clustering-based water-drive reservoir flow field regulation and control method, which comprises the following steps of: 1, establishing a block numerical simulation model, and carrying out streamline numerical simulation; 2, connecting a Python program with a streamline numerical simulator, extracting streamline attributes, and obtaining a streamline field; 3, flow field water drive capability characterization indexes are constructed, wherein the indexes comprise the production potential coefficient and the water content; 4, calculating the production potential coefficient and the water content of each streamline, carrying out streamline clustering analysis, and identifying a dominant seepage channel and a low-efficiency water injection region as constraint conditions; and 5, optimizing the injection-production quantity of each well by taking the maximum streamline average oil displacement efficiency as a target function in combination with a clustering result. According to the method, the flow field can be identified and regulated more quickly and accurately, the calculation efficiency is higher, and the method has wide application value in complex water-drive oil reservoirs.

The invention relates to a thin-wall ceramic tube for static pressure strength testing and a preparation method thereof. The ceramic tube comprises a prefabricated body, and grinding sections are arranged at the two ends of the prefabricated body; and densifying the formed preform to obtain a ceramic tube, and grinding the surfaces of the two ends of the ceramic tube to obtain the thin-wall ceramic tube for static pressure strength test. The coaxiality of the two ground ends is not larger than 0.2 mm, and the surface smoothness is smaller than 0.2. Compared with the prior art, accurate characterization of the circumferential strength of the ceramic tube is realized.

The invention relates to an object point cloud reconstruction method based on a guided cross-modal robot, and the method comprises the steps: obtaining a depth image and a corresponding RGB image of an object under a single visual angle of the robot, carrying out the preprocessing, and obtaining the incomplete data of the point cloud of the object; a point cloud classification encoder is adopted to encode the point cloud classification task, a point cloud classification decoder is adopted to decode the point cloud classification task, and a classification result of the object is obtained; based on the object point cloud incomplete data, a point cloud data encoder is adopted to carry out position embedded coding, and point cloud features containing position codes are obtained; acquiring tactile data of an object at a single-view blind area position of the robot, and encoding the tactile data by a tactile space encoder to obtain tactile features; a visual touch cross-modal fusion module is adopted to fuse the point cloud features and the touch features, and a classification result is utilized to guide in the fusion process; and decoding the fused features by adopting a pyramid decoder to obtain a complete point cloud of the object. Compared with the prior art, the method has the advantages of high precision and the like.

The invention provides a geological disaster risk evaluation method and device fusing ascending and descending orbit time sequence InSAR deformation, and the method comprises the steps: carrying out the deformation inversion based on ascending and descending orbit time sequence InSAR data, obtaining an ascending orbit surface deformation quantity and a descending orbit surface deformation quantity, and obtaining a dynamic deformation factor corresponding to each grid in a research region after fusion; determining a multi-dimensional static environment factor corresponding to each grid in the research area based on the optical remote sensing data and the environment data; traversing each grid in the research area, and screening the dynamic deformation factors and the multi-dimensional static environment factors corresponding to the grids to determine a plurality of target evaluation factors corresponding to the grids; and based on the contribution degree of the target evaluation factor corresponding to each grid to the geological disaster event occurring in the research area, generating a geological disaster risk evaluation result corresponding to the research area. According to the method, geological disaster deformation can be represented more comprehensively and accurately, so that the geological disaster risk evaluation precision is improved.

The invention discloses a pavement crack extraction method based on a neural radiation field and a semantic segmentation network, and particularly relates to the technical field of road engineering. Pavement image data of a research area are acquired through an unmanned aerial vehicle and are preprocessed to form a required image data set; and performing three-dimensional reconstruction on the preprocessed image data set through combination of a motion recovery structure multi-view stereoscopic vision algorithm and a neural radiation field algorithm to obtain a road surface point cloud model and a point cloud data set, constructing a semantic segmentation network and training the point cloud data set, and completing extraction and calculation of road surface crack information according to a semantic segmentation result. Compared with the prior art, the method has the advantages that the detection cost is reduced, the real-time detection requirement is met, the current situations of poor automation, poor expressivity, insufficient application and the like of existing point cloud processing are obviously improved, in addition, a road scene with cracks can be visually and accurately represented, and therefore a bottom layer platform and data support are provided for digital representation of road surface information.