1070results about How to "Good application effect" patented technology

The invention provides a multi-level low-altitude air route network construction method in a complex urban environment, and relates to the technical field of low-altitude air route construction, and the method comprises the steps: building a low-altitude flight rule base; performing three-dimensional discretization on the urban space, and dividing the airspace into three-dimensional grids; designing a risk assessment function R; dividing a plurality of height layers, and distributing the height layers for different types of aircrafts; constructing a multi-stage air route network; constructing an airspace three-dimensional airway model; based on the airspace three-dimensional air route model, adopting a hybrid iterative algorithm to optimize the air route planning of each height layer; performing path smoothing on the discrete path points generated by the hybrid iterative algorithm by adopting a B spline curve, and performing flight performance verification on the smoothed path to obtain a continuous flight path; and generating a low-altitude air route network according to the continuous flight path. According to the method, the defects in the aspects of complex urban environment processing, multi-source data integration, multi-level airspace division, path optimization and the like in the prior art can be overcome, so that safe, efficient and flexible low-altitude flight management is realized.

The invention relates to a closed-loop multi-guide multi-mode time domain interference electrical stimulation system and method. The method comprises the steps that a software part determines stimulation parameters; the automatic electrode position adjusting device adjusts the spatial position of the multi-conductive-electrode device, and the multi-conductive-electrode device stimulates the position of a target spot; the current output precision and impedance monitoring module monitors the precision and impedance information of electrode output current, and the core main control and functional circuit module adjusts a control signal according to the precision and impedance information; when the multi-conducting-electrode device conducts stimulation, the electroencephalogram device synchronously collects electroencephalogram signals and uploads the electroencephalogram signals to the electroencephalogram analysis algorithm module of the software part, the electroencephalogram analysis algorithm module evaluates and feeds back the effectiveness and the effect of stimulation, and the time domain interference electrical stimulation function module and the target area dynamic adjustment algorithm module optimize stimulation parameters according to the effectiveness and the effect of stimulation. Through the arrangement, a real-time closed-loop feedback control system is formed, continuous monitoring, real-time optimization and adjustment of electrical stimulation parameters and dynamic adjustment of a stimulated brain target region are realized, and the accuracy and safety of current output are ensured.

The invention provides a multi-modal large model processing method and device, a storage medium and a program product. The method comprises the steps of obtaining an original image set of a monitoring scene; identifying feature information of a target object in the original image set; according to the feature information, generating question and answer pairs corresponding to the original image set; and taking the original image set and the corresponding question and answer pair as training data, and performing fine tuning training on a pre-trained multi-modal large model to obtain a target multi-modal large model adaptive to the monitoring scene. According to the method and the device, the multi-modal large model can better adapt to the requirements of a specific monitoring scene, the performance and the application effect of the multi-modal large model in the monitoring scene are improved, and the intelligent level of a monitoring system can be improved.

The invention relates to the field of nutrition and health management, and discloses a nutrition evaluation and prediction method and system based on big data, and the method comprises the following steps: firstly collecting the physiological data, diet record and activity data of a user, and carrying out the preprocessing of the data, and guaranteeing the data quality; dimensionality reduction is carried out through feature extraction and principal component analysis, and key features are extracted; the method comprises the following steps: evaluating daily nutritional requirements of a user by using a multivariable linear regression model, a Bayesian network and causal inference, predicting a nutritional requirement change trend of the user in a future period of time based on a long-short-term memory network, and optimizing daily diet selection of the user by using a Markov decision process; and finally, generating a personalized nutrition scheme by adopting linear programming and a genetic algorithm, and ensuring that the nutritional requirements and health targets of the user are met. According to the method, personalized and scientific diet suggestions can be provided, and the quality and effectiveness of individual nutrition management are improved.

The invention relates to the technical field of quality traceability management, in particular to an organic fertilizer quality safety traceability management method based on multi-source data. The method comprises the following steps: collecting organic fertilizer source data in real time from links of organic fertilizer production, circulation, storage and use, performing standardization processing, and performing production batch synchronous set processing and traceability association connection file establishment to generate an organic fertilizer multi-source data traceability association connection file; and carrying out production batch tracing extraction processing on the organic fertilizer multi-source data traceability association connection file, carrying out organic fertilizer quality index evaluation analysis and quality index error limitation analysis, carrying out organic fertilizer quality safety abnormity traceability processing and abnormity decision support processing, and generating an organic fertilizer quality safety abnormity problem decision support report. Therefore, the corresponding organic fertilizer quality safety abnormity traceability problem management work is executed. The scientificity and accuracy of organic fertilizer quality safety management can be improved.

The invention discloses a remote sensing surveying and mapping image enhancement processing method for an unmanned aerial vehicle, and relates to the technical field of image enhancement processing. During operation of a system, flight attitude and position information are recorded through GNSS and IMU equipment, the flight path of the unmanned aerial vehicle is controlled, and collected original images are primarily processed; calculating and obtaining a flight stability coefficient Fs, an image acquisition quality coefficient Iq and a topographic feature retention coefficient Tf, removing random noise in a remote sensing image by using an improved non-local mean NLM algorithm, optimizing and processing an edge region, improving edge definition and texture details of a topographic image through a Laplacian pyramid algorithm, strengthening landform features of mountains and hills, and obtaining a remote sensing image; according to the method, color distortion caused by complex mountain illumination conditions is corrected, image colors are enabled to truly restore topographic features, important landform information is prevented from being lost due to over-brightness or over-darkness, topographic feature extraction is optimized, a final index Zfinal is generated, and quality evaluation is performed on an enhanced and corrected image.

The invention discloses a method for analyzing mineral trace elements by utilizing a micro-Raman spectrum, which comprises the following steps of: fixing a mineral sample on a rotary sample table of a micro-Raman spectrometer, selecting an area to be detected by combining a micro-imaging technology, collecting a Raman background signal under the pre-illumination of low-power laser, and analyzing the trace elements in the area to be detected according to the Raman background signal. The laser focal length, the spot size and the incident angle are adjusted accordingly, the laser is focused on a target area, under the calibrated laser condition, through a multi-angle excitation micro-area, after an algorithm is applied to optimize signals, reduce impurity interference and integrate multi-angle Raman signals, the characteristic spectrum information of the microelements can be extracted, and the microelements can be accurately detected. High-resolution analysis of sample components is realized, and a three-dimensional distribution model of microelements in a microcell is constructed in combination with microscopic imaging data. The method has the characteristics of non-destructiveness, high resolution and rapid detection, and is suitable for mineral component analysis in the fields of geology, material science and the like.

The invention relates to the technical field of safety evaluation, in particular to an ultra-deep soft soil foundation pit safety evaluation method and system, and the method comprises the following steps: obtaining the geological data of an ultra-deep soft soil foundation pit, recording the soil type, soil density, underground water level and temperature data, carrying out the data arrangement and formatting, and generating a geological data record sheet. According to the method, the detailed geological data of the ultra-deep soft soil foundation pit are acquired and arranged, the accuracy and the applicability of the basic data are enhanced, so that subsequent analysis is carried out according to clear actual conditions, and the timeliness and the accuracy of risk prediction are ensured by directly associating prediction of mechanical behaviors with risk assessment; the risk assessment model can be dynamically adjusted, the effectiveness and accuracy of early warning are improved, the risk dependency relationship model and the monitoring scheme are integrated, theoretical and actual operations are effectively linked, the systematicness and adaptability of construction safety management are effectively improved, and the accident prevention capability is remarkably improved.

The invention belongs to the technical field of automobile information safety fuzzy testing, discloses an automobile information safety fuzzy testing method based on a large model, and develops optimization application research of LLMs in an automobile software fuzzy testing scene by utilizing the advantages of an artificial intelligence technology in the aspects of data processing, natural language interaction, generalization ability and the like. The concealment and effectiveness of the test set generated by the fuzz test are optimized, and the fuzz test process efficiency is improved. Through the implementation of the invention, related key technologies are applied to the fuzz test of the automobile, and other related products and enterprises in China are radiated by writing technical standards, so that the fuzz test system can be used as an application demonstration and sample plate of the fuzz test of the intelligent driving automobile. According to market analysis, fuzzy test calculation is carried out on 100 thousand automobiles every year according to the method, and the data safety and the system safety are improved.

The invention provides an open vocabulary 3D segmentation method based on a three-dimensional Gaussian sputtering technology, and the method comprises the steps: S1, obtaining a multi-view 2D image set, and carrying out the initialization three-dimensional reconstruction; s2, segmenting the multiple views by using SAM to obtain a 2D mask; s3, associating masks of the same instance by using a video tracking segmentation model (SAM2); s4, extracting open vocabulary features of each mask by using a visual language model (CLIP); s5, performing average pooling and dimension reduction on the CLIP feature of the mask of each instance to form a representative open vocabulary feature of the instance; s7, training the model, and supervising the embedding of a low-dimensional embedding feature # imgabs0 #; and S8, comparing the CLIP feature of the query text with the semantic feature of the Gaussian body to obtain the Gaussian body with high correlation, and generating a mask of the query object through rasterization rendering. According to the method, only a multi-view 2D image is needed, and the three-dimensional object and the scene are segmented by combining the three-dimensional Gaussian sputtering technology and the open vocabulary technology, so that the problems of difficulty in 3D data acquisition and high cost are solved.

The invention provides a negative pressure suction dust removal device based on a Venturi effect, and belongs to the field of coal mine dust removal. The problems that an existing dust removal device suitable for the field of coal mine dust removal is too large in size and mass, difficult to move, large in noise, high in daily maintenance cost and the like are solved. The device comprises a box body, a water power fan, a Venturi tube, a desilting assembly, a counter-rotating fan, a filtering cabin and a water tank, wherein an air inlet pipeline, a settling cavity, an air duct and an air outlet pipeline are sequentially arranged in the box body; the hydrodynamic fan is arranged at the inlet end of the air inlet pipeline; the input end of the Venturi tube is communicated with the air inlet pipeline, and the output end is communicated with the settling cavity; the desilting assembly is arranged below the air duct; the counter-rotating fan is arranged in the air duct and is positioned between the settling cavity and the filtering cabin, and the filtering cabin is communicated with the air outlet pipeline; and the water tank is communicated with the filtering cabin and is also communicated with a water inlet of the hydrodynamic fan. The dust removal device is simple in structure, small in size and mass, convenient to move, low in noise and low in daily maintenance cost.

The invention provides an operator precision verification method and device, a server and a medium, and the method comprises the steps: carrying out the reasoning of a model based on model input, and obtaining the standard output of each operator in the model in reasoning; aiming at each operator, taking the standard output of the previous operator of the operator in the model as the test input of the operator, and obtaining the test output of each operator under the test input; and based on the standard output and the test output of each operator, carrying out precision verification on each operator. According to the method, the device, the server and the medium provided by the invention, the error of each operator is determined based on the standard output and the test output in the comparison reasoning process, and the accuracy and efficiency of operator precision verification can be effectively improved. The reliable standard output is generated and used as the test input, so that the data transmission continuity in the verification process is ensured, the problem of accumulative errors is avoided, and automatic high-efficiency precision verification is realized.

The invention discloses an aquatic product ultra-low temperature fresh-keeping processing control method, and relates to the technical field of aquatic product fresh-keeping processing, and the method specifically comprises the following steps: evaluating a fat local oxidation risk grade caused by oxygen exposure on the surface of a high-fat aquatic product in each region in a freezing chamber; dividing each area in the freezing chamber into a low risk area, a medium risk area and a high risk area according to an evaluation result; constructing an oxidation risk adjustment mechanism according to the division result of each region in the freezing chamber; in the implementation process of the oxidation risk adjustment mechanism, whether the adjustment effect of the oxidation risk adjustment mechanism on each region can reach the expectation or not is evaluated, and the oxidation risk adjustment mechanism is optimized according to the evaluation result. According to the method, the problem of fat oxidation caused by oxygen exposure on the surface of the high-fat aquatic product in the ultralow-temperature freezing process is solved, and the preservation effect in the freezing process is effectively improved through real-time monitoring and dynamic optimization of an adjustment mechanism.

The invention discloses a high-flexibility structure nonlinear aerodynamic damping estimation method based on a Sage-Husa algorithm and square root cubature Kalman filtering. According to the method, aerodynamic damping with nonlinear characteristics can be effectively identified from free vibration data, forced vibration data, steady-state vibration data and random vibration data of high-flexibility structures such as high-rise buildings and fan towers under crosswind excitation, and process noise can be automatically adjusted according to the current state; the interested parameters of the high-dimensional system are accurately predicted, the noise level does not need to be known sufficiently, and a theoretical basis can be provided for evaluating the structural response by calculating the aerodynamic damping of the high-flexibility structure more conveniently and accurately in actual engineering. The invention further discloses a high-flexibility structure nonlinear aerodynamic damping estimation system based on the Sage-Husa algorithm and the square root cubature Kalman filter and a storage medium.

The invention discloses a three-dimensional reconstruction method and system based on a line shift coding pattern, and the method comprises the steps: designing an original fringe pattern in a line shift manner to obtain a plurality of coding patterns, and carrying out the projection and collection of a to-be-measured object through the plurality of coding patterns and a preset Gray code pattern, respectively obtaining a plurality of fringe projection drawings and a plurality of Gray code projection drawings; synthesizing M phase shift fringe patterns according to the collected multiple fringe projection patterns; performing phase calculation on the M phase shift fringe diagrams to obtain a folded phase diagram; and combining the folded phase diagram with the plurality of Gray code projection drawings for phase unwrapping to obtain an unwrapped phase diagram, and reconstructing the to-be-measured object by using a three-dimensional reconstruction principle. According to the method, fringe synthesis is firstly carried out on the plurality of coding patterns, and then phase calculation is carried out on the M phase shift fringe patterns, so that three-dimensional reconstruction is carried out on the to-be-measured object by using the phase, relatively high reconstruction precision is achieved, the robustness is relatively high, and the method has a relatively good application effect in a high dynamic range scene.

The invention discloses a model processing method, a chip assembly and equipment. The method comprises the following steps: an in-memory computing element performs first operation processing on input data to obtain first processing data; wherein the first operation processing comprises fixed weight calculation of an operand; the near memory computing element performs second operation processing on the first processing data to obtain second processing data; wherein the second operation processing at least comprises attention calculation of operand change. Thus, heterogeneous deployment of model parameters of the model can be realized based on cooperation of the in-memory computing element and the near-memory computing element, the in-memory computing element can realize weight calculation of fixed operands, and the near-memory computing element can realize attention calculation of operand changes. The characteristics of high computing power efficiency and low power consumption of the in-memory computing element can be utilized to the greatest extent, the erasing operation on the in-memory computing element can be reduced by means of the near-memory computing element, and the application performance of a model deployed on the equipment side can be effectively improved.

The invention discloses a coal power multi-path transformation and new energy combination decision-making method under low-carbon transformation. The method comprises the following steps: firstly, constructing a system comprising first-level effect indexes such as environmental influence, economic benefit, energy safety and technical innovation and corresponding second-level indexes; then, alternative schemes of various coal power transformation technology and new energy project type combinations are generated; then quantitatively evaluating the effect index of each scheme, and calculating a comprehensive utility value by using multi-criterion decision analysis; identifying potential risk factors, and performing uncertainty analysis through Monte Carlo simulation; and finally, according to the low-carbon transformation target and policy evaluation scheme conformity, selecting a scheme which most conforms to the policy target, and continuously monitoring and adjusting. The coal-electricity multi-path transformation and new energy combination decision-making method under low-carbon transformation is beneficial to promoting energy low-carbon transformation, optimizing resource allocation, improving economic benefits, promoting technical innovation, enhancing risk management capability and ensuring that a project is consistent with policy orientation.

The invention relates to the technical field of sodium ion battery research and development, and particularly discloses a strong-oxidizing acidified asphalt-based hard carbon negative electrode material and a preparation method and application thereof, concentrated nitric acid or concentrated sulfuric acid is introduced into asphalt as a cross-linking agent, so that the cross-linking reaction among molecules can be promoted under a low-temperature condition, and the structural stability of the material is greatly improved. Compared with a traditional method, the method has the advantages that the technological process is simplified, only one-step pretreatment is needed before carbonization, and energy consumption and pollution caused by high temperature can be avoided. According to the liquid phase cross-linking strategy, through reconstruction of chemical bonds, the low-softening-point asphalt can form a more compact carbon precursor network structure at low temperature, and formation of a high-performance hard carbon material in the subsequent carbonization process is facilitated.

The invention discloses a processing system and method for real-time self-adaptive adjustment of galvanometer jitter based on movement speed, and the method comprises the steps: obtaining the current instantaneous speed of a galvanometer relative to a foil in real time in the foil cutting process, and automatically building a relation model between the instantaneous speed of the galvanometer and jitter parameters according to a machine learning method adopted in advance, dynamically adjusting jitter parameters of the galvanometer by using the trained neural network model; a new cutting path is generated according to the shaking parameters of the galvanometer after real-time adjustment, and the scanning speed of the galvanometer is adjusted according to the adjusted shaking parameters so as to meet the requirement of the new cutting path; and the galvanometer is adjusted according to the new cutting path and the new scanning speed, the output of the laser is controlled, and the galvanometer continues to run to complete the cutting operation of the foil. Based on the instantaneous movement speed of the galvanometer, jitter parameters of the galvanometer are adjusted in a real-time and self-adaptive mode, the flight cutting track is corrected, the stability and consistency of the cutting shape are ensured, and therefore the overall quality of a final product is improved.

The invention relates to the technical field of network security, and discloses a deception defense system design method based on heterogeneous fusion, and the method comprises the steps: receiving a network access request sent by a client, carrying out the recognition of the network access request through an attack recognition model, and obtaining a recognition result, wherein the attack recognition model comprises a data processing module and an attack recognition module; when the network access request in the identification result is identified as a normal request, forwarding the network access request to the simulation model; when the network access request in the recognition result is recognized as an attack operation request for service target defect utilization, key parameters and loads are extracted from attack operation to automatically initialize a remote control execution environment, and then an attack behavior is redirected to a high-interaction remote control execution environment unit to be executed; according to the method, automatic and intelligent simulation of a complex service system is realized, deep trapping of highly-hidden unknown attacks is formed, and the problem that highly-hidden unknown network attack threats are difficult to find and capture can be effectively solved.

The invention belongs to but is not limited to the technical field of seabed monitoring, and particularly relates to a strain test method and system for subsurface buoy shell strength, and the method comprises the steps: S1, selecting points on the surface of a shell, and pasting strain gauges; s2, putting into a sealed kettle, and injecting water and pressurizing; s3, testing the strain condition of the shell by using a strain tester; and S4, the pressure tolerance degree of the shell is analyzed through calculation. The buoyancy provided by the floating body shell and the overall weight of the subsurface buoy are calculated in detail, and the needed ballast balance weight is accurately measured and calculated. Buoyancy and gravity balance is verified through dynamic simulation, it is ensured that the subsurface buoy can stably suspend or settle in water, and the design feasibility and the actual application effect are improved.

The invention discloses a topological optimization method for an optical storage integrated AC / DC system based on an adaptive algorithm, and the method comprises the following steps: obtaining an initial topological structure and operation parameters of the optical storage integrated AC / DC system, including photovoltaic power generation, configuration of an energy storage system, operation state parameters of a microgrid, and external environment parameters; optimizing and adjusting the topological structure, and dynamically generating a plurality of candidate topological structures according to system operation data and environment change which are acquired in real time; carrying out analogue simulation on the candidate topological structure, and evaluating the stability, power distribution and energy utilization efficiency of the system under different operation conditions; performing reliability analysis on the optimized topological structure, and monitoring and adjusting the topological structure of the system in real time; and performing simulation verification on the adjusted topological structure to obtain an optimal topological structure. According to the invention, the system stability, reliability and fault processing capability can be improved.

The invention belongs to the technical field of underwater positioning, and discloses an underwater single beacon virtual long baseline positioning method and system capable of estimating an unknown sound velocity. The method comprises the following steps: constructing a virtual long baseline array by combining stored dead reckoning data; constructing a virtual long baseline positioning problem by combining the relationship between the underwater acoustic signal propagation time and the position and sound velocity of the AUV; auxiliary variables are introduced, a pseudo-linear positioning equation is obtained based on algebraic conversion, and an AUV position initial solution is obtained based on a weighted least square method; and constructing a second-step linear estimation problem by exploring the relationship between the auxiliary variable and the AUV position, and solving by using weighted least squares to obtain a more accurate AUV position estimation solution. Compared with a single-beacon positioning method which does not consider sound velocity unknown, the method provided by the invention can obtain more ideal position estimation performance, and the practical application capability of an underwater single-beacon positioning system is enhanced.

The invention relates to an efficient catalytic composite nano-capture deep deodorization method for waste gas with complex components, which is applied to the field of purification of waste gas with complex components, and comprises the steps of pretreatment of gas inlet, catalytic enrichment, nano-liquid capture and the like, and through the synergistic effect of a catalytic enrichment coupling reactor, an atomizing nozzle, nano-liquid and an intelligent deodorization control system, the deodorization effect of the waste gas with complex components is improved. VOCs and odor molecules in the malodorous gas can be efficiently and rapidly removed, the human body is free of feeling, the whole deodorization process is carried out under the normal temperature condition, the investment and operation cost for malodorous gas treatment is remarkably reduced, and the deodorization method has the advantages of being high in safety, low in operation cost, small in occupied space, low in secondary pollution and the like and is suitable for popularization and application. According to the method, the practical application effect of malodorous gas treatment is effectively improved, the safety of discharged gas is ensured by evaluating the treated gas, and the application effect and quality of the catalytic enrichment coupling reactor and the nano liquid can be monitored and adjusted in real time, so that the intelligent level of the malodorous gas treatment process is enhanced.

The utility model discloses a liquid fertilizer spraying device for garden planting, which relates to the technical field of spraying devices and comprises a bearing plate, rollers are fixedly connected to four corners of the bottom end of the bearing plate, a handrail is fixedly connected to the upper end of the rear side of the bearing plate, and a storage box is arranged at the upper end of the rear side of the bearing plate. Under the combined action of the storage box, the motor I, the rotating shaft, the stirring rod and the sewer pipe, liquid fertilizer for garden planting can be fully stirred and mixed during use, and under the combined action of a water suction pump, a water suction pipe, a water supply pipe, a hose and a spraying seat, the mixed liquid fertilizer can be conveniently conveyed to a spraying device, so that the spraying device is convenient to use. The garden liquid fertilizer spraying device has the advantages that liquid fertilizer can be conveniently sprayed, the height and angle of the spraying device can be adjusted according to use requirements under the combined action of a bearing seat, a first gear, a second motor, a second gear, an adjusting rod and an adjusting knob, and trees in gardens can be conveniently and quickly sprayed.

The embodiment of the invention provides a retrieval processing method and related equipment. The method comprises the steps of obtaining a to-be-retrieved first query statement; identifying a target word belonging to the target field from the first query statement, and performing semantic extension processing on the target word to obtain at least one extension word belonging to the target field; performing optimization processing on the first query statement based on the at least one expansion word to obtain at least one optimized query statement; and performing retrieval processing by adopting at least one optimized query statement. Through the embodiment of the invention, the optimization quality of the query statement can be improved from the source, and the retrieval accuracy is improved.

The invention provides a flexible production plan prediction system based on a knowledge graph and a language model, and the system comprises a feature network graph building module which is used for aggregating the historical production data of a flexible factory to obtain feature network graph data; the flexible production large language model establishment module is used for training the natural language model by using the feature network graph data to obtain a flexible production large language model; the flexible production plan model establishing module is used for obtaining a flexible production plan model by adopting a stacking algorithm and training; the knowledge graph establishing module is used for obtaining a knowledge graph according to the feature network graph; and the flexible production plan prediction module is used for calling the flexible production large language model, the flexible production plan model and the knowledge graph by utilizing the language model development framework, and inputting the current customer demand to obtain a comprehensive flexible production plan. And predicting the flexible production plan through the synergistic effect of the two models and the knowledge graph.

The invention relates to the technical field of meteorological element forecasting, in particular to a meteorological element forecasting product intelligent generation method and system. The method comprises the following steps: acquiring meteorological element original data, and performing data assimilation on the meteorological element original data to obtain assimilated product data; combining the meteorological element original data and assimilation product data to obtain a forecast sensitive factor and site feature data; establishing a meteorological element product data set according to the meteorological element original data, the assimilation product data, the forecast sensitive factors and the site feature data; constructing a meteorological element forecast product intelligent generation model, and training and verifying the meteorological element forecast product intelligent generation model by using the meteorological element product data set; and generating a meteorological element forecasting product through the trained meteorological element forecasting product intelligent generation model. According to the method, the meteorological element original data is utilized, and the problem of efficient, accurate and intelligent generation of the meteorological element forecast product of the meteorological station is solved.

The invention belongs to the field of natural language processing, and particularly relates to the field of large language models, in particular to an education large model training method under a fine-grained weighted reward mechanism. The method comprises the steps of collecting vocational education field data training strategy models and cleaning data; after completing the training of the base large model, training a sentence-level fine-grained reward model by using supervision data, and designing a context-associated scoring mechanism; after scoring model training and large model supervision fine tuning are completed, large model alignment is carried out, and the strategy is updated according to the total reward value. According to the method, the context factor and the dimension interaction factor are introduced, so that sentence scoring is more reasonable and effective, the safety and reliability of a large language model in the vocational education field are improved, and the method has wide application prospects.

The invention discloses a visual question and answer method based on environmental perception and a knowledge base, and aims to realize an accurate visual question and answer function through a reasoning technology of integrating field specific environmental information and knowledge base information. The method comprises the following steps: firstly, extracting image features by utilizing a ResNet model in visual feature extraction, extracting text features by utilizing a BERT model in text feature extraction, and comprehensively processing the features and context information acquired by an environment perception module through multi-modal fusion; and then, on the basis of the fused multi-modal features, through combination of a Wikdata knowledge base and Seq2Seq, an answer is output in a text form, and visual feedback can be provided. Through environment perception, the system can adjust answers in real time to ensure that the system highly adapts to the current scene, and is applied to the fields of intelligent assistants, aided driving, household robots and the like.