34results about How to "Obvious performance advantage" patented technology

The invention relates to the technical field of autonomous robot decision-making systems, in particular to an autonomous robot decision-making system and method based on multi-modal sensing fusion, and the system comprises a multi-modal data collection module which is used for obtaining multi-modal sensing data of an environment; the data preprocessing module is connected with the multi-modal data acquisition module and is used for preprocessing the multi-modal sensing data; the feature extraction module is connected with the data preprocessing module and used for extracting features from the preprocessed multi-modal sensing data; the dynamic space-time diagram construction module is connected with the feature extraction module and is used for constructing a dynamic space-time diagram based on the features; the multi-modal data fusion module is connected with the dynamic space-time diagram construction module and is used for fusing the multi-modal data in the dynamic space-time diagram; through adaptive time alignment and a cross-modal space attention mechanism, the system can effectively process space-time inconsistency of data of different sensors, and centimeter-level positioning precision is realized.

The invention discloses an offline writer identification method and device, electronic equipment and a medium. The method comprises the following steps: acquiring handwritten word image data; converting the handwritten image into a grayscale image, performing binarization processing on the grayscale image, adjusting the size of the processed image to a preset size, and constructing a training set; constructing a handwriting recognition model, and training the handwriting recognition model by adopting the training set; wherein the handwriting recognition model comprises a local modeling module, an identical branch and a global module, the local modeling module is used for capturing local features, the identical branch is used for extracting global features, and the global module is used for integrating the local features and the global features; and inputting a handwritten image to be recognized into the trained handwriting recognition model to obtain a result of a final classification layer of the model, performing softmax probability calculation, and taking a writer with the maximum probability as a recognition result. According to the method, robust feature representation can be effectively extracted from limited handwriting, and the recognition precision far higher than that of other methods is achieved.

The invention discloses a gas sensor for detecting dissolved gas in oil, an array of the gas sensor and a preparation method of the gas sensor. The gas sensor comprises a substrate, an interdigital electrode deposited on the surface of the substrate, and a p-n junction modified TMDCs gas sensitive material attached to the surface of the interdigital electrode, the p-n junction modified TMDCs gas sensitive material is a MoS2-WS2 heterostructure material, and the molar ratio of W to Mo in the MoS2-WS2 heterostructure material is (1.3: 1) to (1.7: 1). The MoS2-WS2 heterostructure is formed by adjusting the ion proportion of W and Mo, and the sensitivity and selectivity of the sensor to gas dissolved in oil are remarkably improved.

The invention relates to a Halton sequence optimization method and system based on RISC-V vector extension and application. The Halton sequence optimization method comprises the following steps: 1) identifying mathematical characteristics generated by a Halton sequence; a Halton sequence is generated; after continuous index values are partitioned and grouped, simultaneous processing of the multiple index values is achieved through a parallelized grouping calculation method in combination with vectorization operation; 2) carrying out vectorization optimization by using an RISC-V vector instruction; in a Halton sequence generation process, carrying out parallel processing on the grouped index values by using an RISC-V vector extension instruction; 3) packaging into a dynamic link library; and 4) integration and calling with a Hadoop framework. According to the Halton sequence generation method, the number of loop iterations is remarkably reduced, and the Halton sequence generation efficiency is greatly improved. According to the method, resource waste in traditional serial computing is avoided, meanwhile, branch jump and condition judgment expenses are reduced, and the utilization rate of hardware computing resources is increased.

The invention relates to a wear-resistant and corrosion-resistant polyurethane film for the front edge of a wind power blade and a preparation method thereof, and belongs to the technical field of films. The dispersion and binding force of graphite-phase carbon nitride in a waterborne polyurethane emulsion are remarkably improved through ultrasonic-assisted solvent stripping, the dispersion and interfacial compatibility are further optimized by introducing carboxymethyl chitosan, and a composite film is formed on the surface of a matrix, so that corrosive ions are effectively shielded, the corrosion resistance is enhanced, and the service life is prolonged. The photocatalytic performance of the graphite-phase carbon nitride can also endow the film with self-cleaning and anti-photoaging capabilities. Besides, the carbon nanotubes and the silicon dioxide are compounded by adopting an in-situ hydrolysis technology, so that the interaction force between the carbon nanotubes and the silicon dioxide is remarkably enhanced, the overall performance of the composite filler is improved, the compatibility between the composite filler and polyurethane is enhanced by introducing a silane coupling agent, and the uniformity and stability of the membrane are ensured; the composite filler combines the impact resistance of the carbon nanotubes and the hardness and wear resistance of silicon dioxide, so that the film shows higher strength and durability under complex stress.

The invention provides a ferronickel-based oxygen evolution electrode and a preparation method thereof based on interface confinement. The preparation method of the ferronickel-based oxygen evolution electrode comprises the following steps: preparing a precursor solution containing ferrous ions and ferric ions; the foamed nickel substrate is cleaned; the cleaned foamed nickel is tightly attached to the surface of the bottom of the reaction container, the foamed nickel is immersed in the precursor solution, under the room-temperature standing condition, divalent and trivalent iron ions are induced to be subjected to a synergistic hydrolysis and coprecipitation reaction through a physical interface confinement space formed by an attached interface, and the nickel-based composite material is obtained. Therefore, a high-activity and high-uniformity nickel-iron-based (oxygen) hydroxide catalyst layer (DIC-NiFeOOH / NF) is grown on the surface of the foamed nickel in situ. The method is simple in process, green and energy-saving, and the prepared nickel-iron-based oxygen evolution electrode shows excellent performance in oxygen evolution reaction and anion exchange membrane water electrolysis.

The present invention discloses an intelligent beam width optimization method based on reinforcement learning. The method suppresses the beam drift effect in millimeter wave communication by dynamically adjusting the beam width. The algorithm models the beam width dynamic optimization problem as a Markov decision process, and selects the optimal beam width for data transmission at each decision moment. Its state design can characterize the severity of the beam drift effect of the current system from multiple angles, and each action corresponds to a different beam width, thereby optimizing and adjusting the width of the data transmission beam by sensing the speed of environmental changes. In each state, the optimal beam width for data transmission is selected according to the AC algorithm, and the policy network and value network are continuously updated during training to improve the rationality and reliability of model selection, while introducing a dynamic network update mechanism to reduce the computational burden of network updates. The throughput of the millimeter wave communication system under the beam drift effect is improved, thereby continuously ensuring the quality of the communication link.

The application belongs to the technical field of biological medical materials, and discloses a polydopamine modified modified nano-graphyne and a preparation method and application thereof. In the method, the polydopamine coating is combined with the nano-graphyne for the first time to construct a GDY@PDA composite system, and the core defects of the blocky graphyne, such as easy agglomeration, poor water solubility and insufficient biocompatibility, are solved. The graphyne-based material is applied to the near-infrared two-region photothermal regulation of neural cell calcium influx for the first time, the calcium ion channel is activated through the local thermal effect of the material, and precise, remote and reversible calcium influx regulation is realized; the limitation that the traditional biocompatibility modification is easy to fall off and difficult to functionalize is broken through, the polydopamine modification makes the material have both photothermal efficiency and biological safety, and the application of the graphyne in the treatment of neural diseases (such as neural cell injury repair) is expanded.

The application discloses a ground-based space target monitoring radar atmospheric refraction error correction method. The ground-based space target monitoring radar atmospheric refraction error correction method comprises the following steps: based on the vertical distribution of atmospheric refraction index and the radar elevation angle measurement value, the signal propagation path is accurately recursively calculated by using the refraction theorem and geometric relationship, the corresponding elevation angle, geocentric angle and target apparent distance of different heights are calculated, and a one-to-one corresponding relationship between the apparent distance and the distance, the elevation angle measurement error is established; then, according to the radar distance measurement value, the radar distance and the elevation angle measurement error are calculated by interpolation, and finally, the distance and the elevation angle measurement error caused by the atmospheric refraction of the space target monitoring radar are corrected. The ground-based space target monitoring radar atmospheric refraction error correction method has higher error calculation precision under the condition of given layer interval, has more stable error correction performance, and can more accurately correct the atmospheric refraction error of the space target monitoring radar.

The application belongs to the field of organic-inorganic heterostructure thin films, and relates to a polymer / inorganic heterostructure composite thin film and a preparation method and application thereof. The composite thin film is formed by compounding PEDOT:PSS and ITO nanocrystals, PEDOT:PSS forms a flexible network skeleton, the growth orientation of ITO nanocrystals is a (222) crystal face, a (400) crystal face or a (211) crystal face, and the ITO nanocrystals are uniformly dispersed and bonded in the PEDOT:PSS network. The flexible organic polymer PEDOT:PSS and the brittle ceramic material ITO are compounded, the surface state and the work function of the two-phase materials are regulated, the heterostructure composite thin film with a molecular-level interface combination is constructed, and the critical mechanical properties, the interface toughness and the electron mobility of the thin film are simultaneously improved, so that the demand of the flexible electronic device for the high-performance and high-reliability thin film material is met.

This invention discloses a semi-supervised remote sensing image change detection method and system based on neighborhood relationship propagation, belonging to the field of computer vision and remote sensing image processing technology. It includes: acquiring dual-temporal remote sensing image pairs and dividing them into labeled and unlabeled datasets; constructing a teacher-student framework semi-supervised model, calculating the squared Euclidean distance between the center pixel and its neighboring pixels, converting it into a relevance weight matrix using Softmax, and performing neighborhood-weighted aggregation on the initial prediction map to obtain a corrected prediction map; calculating the supervised loss using labeled samples, and calculating the unsupervised loss using unlabeled samples through weak perturbation to generate pseudo-labels and strong perturbation for consistency learning; jointly optimizing the student model parameters and updating the teacher model using exponential moving average; inputting the dual-temporal images into the trained student model, outputting a probability map, and generating a binary change detection mask. This invention enhances spatial consistency through neighborhood relationship propagation, improving detection performance when labels are limited.

The invention discloses a production method of mixed alcohol amine without rectification and ammonia recovery. The production method specifically comprises the following steps: S1, a reaction stage of ethylene oxide and ammonia water; s2, a further reaction stage with epoxypropane; the invention relates to the technical field of chemical synthesis. According to the production method of the mixed alcohol amine without rectification and ammonia recovery, a method for producing ethanolamine and isopropanolamine without distillation and free ammonia recovery is adopted, the links of distillation and free amine recovery are omitted, the cost is greatly saved, the method is simple and convenient, a water absorption molecular sieve is used as a catalyst, and the production cost is greatly reduced. According to the method, the catalytic water consumption is greatly reduced, so that the energy consumption is reduced, liquid ammonia is directly used as a raw material, the material cost and the operation cost are reduced, meanwhile, the reaction rate is increased, the generation of impurities is reduced, the decomposition of the product is avoided, the purity of isopropanolamine is increased (99% or above), and the method is suitable for large-scale production and has a wide application prospect. And the production process flow is simplified.

The invention discloses a few-sample named entity recognition method and system based on a neural network model, and the method comprises the following steps: obtaining a to-be-recognized original input text, and retrieving from an external knowledge base to obtain a group of candidate knowledge statements; a dynamic similarity threshold algorithm is adopted to filter the candidate knowledge statements, and enhanced knowledge is screened out; inputting into a first encoder, and generating a context text representation matrix of deep fusion knowledge; inputting the matrix into a sequence labeling classification layer, classifying each lexical element, and identifying a boundary span of a potential entity in a text; and inputting a text corresponding to the identified entity boundary span into a second encoder to generate an entity embedding representation, and determining a final category of the entity by calculating a distance between the entity embedding representation and a predefined category prototype to complete named entity identification. According to the method, the problems caused by data sparseness and semantic fuzziness in a few-sample scene are solved, and the accuracy of named entity recognition is improved.

The invention discloses an unmanned aerial vehicle hovering point and flight path joint optimization scheme (DGQDN) for data collection performance improvement. The DGQDN comprehensively optimizes the hovering point and the flight path of the unmanned aerial vehicle, and aims to maximize the system utility. According to the scheme, the system utility not only covers the total amount of sensor network data acquisition, but also considers optimization of the number of hovering points and the repeated coverage rate of the sensor nodes. In the scheme, after the unmanned aerial vehicle completes data acquisition, the matrix filling technology is adopted to accurately recover the sensing data of an unacquired area. The DGQDN starts from three dimensions of hovering point optimization, flight path planning and data recovery and expansion, the data acquisition efficiency of the unmanned aerial vehicle is systematically improved, and an innovative solution is designed for each dimension. A large number of simulation experiments prove that the DGQDN is obviously superior to an existing scheme in different data acquisition scenes, and excellent performance and wide applicability of the DGQDN are shown.

The utility model relates to the technical field of packaging paper feeding and cutting, and discloses a paper stirring piece separating structure of chain type transmission paper feeding equipment, which comprises a separating plate fixed on a rack, a plurality of air outlet grooves are uniformly arranged at the upper part of the separating plate, a communicating pipe is arranged on one side of each air outlet groove, the communicating pipe is connected with an air flow cleaner, and the air flow cleaner is connected with an air inlet pipe of the chain type transmission paper feeding equipment. And a stripping plate is arranged at the upper part of the communicating pipe and is used for stripping the adsorbed paper. According to the utility model, a'mechanical + pneumatic 'double-acting separation mode is formed through sequential cooperation of stripping elastic sheet mechanical triggering and high-pressure airflow, the problem of electrostatic adsorption of multiple pieces of paper is effectively solved, the separation efficiency is improved by more than 40%, the airflow cleaner directionally sprays compressed air through the air outlet groove, static electricity on the surface of the paper is neutralized while a physical isolation layer is formed, and the separation efficiency is improved. And the electrostatic adsorption force is reduced to 0.5 N or below, so that secondary adhesion is fundamentally avoided.

A quantitative diagnosis method for freshwater fish diseases based on OpenCV weak supervision and ResNet50-ADCA dual-branch model is proposed. The method consists of the following steps: (1) Image acquisition and preprocessing: Based on the publicly available freshwater fish disease dataset, a balanced Salmon++ dataset is constructed by combining samples collected from aquaculture sites; (2) Weakly supervised lesion segmentation and labeling: (2.1) HSV color space targeted segmentation; (2.2) Iterative optimization mechanism; (2.3) Disease grading standard; (3) ResNet50-ADCA dual-branch model construction: (3.1) Backbone network and transfer learning; (3.2) SEA channel attention embedding; (3.3) Dual-branch multi-task output layer design; (3.4) Classification branch; (3.5) Grading branch; (3.6) Loss function design; (4) Model training and optimization. This invention balances labeling efficiency, diagnostic accuracy, and quantitative grading capability, and can effectively solve the problems of high labeling cost, poor adaptability to complex environments, lack of quantitative evaluation function, and imbalance between accuracy and speed in the existing technology.

The invention relates to the technical field of power systems, and particularly discloses an integrated energy system state estimation method based on an LSTM automatic encoder, and the method comprises the following steps: building an IES state estimation model, and describing an energy coupling relation in the IES through a constraint equation; state estimation of the IES is carried out based on a weighted least square method; constructing an LSTM automatic encoder, and training the LSTM automatic encoder by using historical operation data of the IES as a training data set to obtain a trained LSTM automatic encoder model; inputting to-be-estimated IES system time sequence data into the trained LSTM automatic encoder model, and obtaining a state estimation result of the integrated energy system through forward propagation of the model; determining a loss function, wherein the loss function comprises prediction loss and regularization loss; the method has the advantages that the long-term dependency relationship and the nonlinear characteristics in the time sequence data can be effectively learned and captured, so that the accuracy and the reliability of state estimation are improved.

The present invention provides a gas-liquid buffer, which belongs to the technical field of buffer equipment, and comprises an outer cylinder body and an inner cylinder body; one end of the outer cylinder body is an open end, and a guide rod is arranged inside, and a guide circle is fixed to the end of the guide rod close to the open end of the outer cylinder body; one end of the inner cylinder body is inserted into the outer cylinder body from the open end of the outer cylinder body, and the other end is located outside the outer cylinder body; a guide plate is fixed to the insertion end of the inner cylinder body, and the guide plate is provided with a guide hole; the inner diameter of the guide hole is larger than the outer diameter of the guide rod and smaller than the outer diameter of the guide platform; the inner cylinder body is sleeved on the guide platform and the guide rod through the guide plate, and an isolating piston that moves along its axial direction is also arranged in the inner cylinder body. The gas-liquid buffer provided by the present invention can avoid the situation that the traditional buffer is prone to being unable to be completely reset and misaligned, and has the advantages of large energy absorption, long service life, simple structure and automatic and rapid reset.

The present application relates to weak network environment batch user migration management method, device and storage medium, read operation task data from local offline file to execute operation task based on task data; Group operation tasks according to the network address of baseboard management controller, adopt asynchronous I / O concurrent processing mechanism, and execute operation tasks of different network addresses of baseboard management controller concurrently, the present application realizes concurrent operation to user, improves operation efficiency; In the same baseboard management controller, operation tasks are executed in turn according to the definition order of local offline file to avoid concurrent conflict; The business logic of all operation tasks is concentrated in the core module, and the calling module is established as the program entry and dispatcher to call the core module to execute operation tasks, so that the weak network environment batch user migration management is completed. The present application can improve operation efficiency, especially in weak network environment, the performance advantage of the present application is more significant compared with the traditional scheme.

The application discloses a multi-scale energy-phase fusion and adaptive difference domain-invariant underwater acoustic feature extraction method, and particularly relates to the design of four core stages of multi-time-frequency resolution spectrum extraction, multi-scale energy-phase fusion, learnable feature alignment and channel splicing, and adaptive time-frequency difference reconstruction. In view of the performance decline problem of the existing method in cross-sea area deployment, the method has the functions of underwater acoustic target robust recognition and generalization application of resisting amplitude fading and suppressing stationary environment artifacts as a light-weight front-end feature extraction module.

The utility model discloses a digital forest compass, and belongs to the technical field of forest resource investigation. Comprising a measuring device integrated with a single-point ranging radar, an attitude sensor and an industrial camera; and the tablet computer is connected with the measuring device and is used for processing, recording and storing the data extracted by the measuring device. The utility model also discloses a digital forest compass measurement system, which comprises measurement hardware which is the digital forest compass; and the measuring software consists of embedded software and PC (Personal Computer) end software. The whole device is small in size, convenient to carry and capable of being operated by one person; compared with a traditional measuring instrument, data does not need to be manually recorded, input and corrected, real-time images and virtual sight beads presented by the industrial camera on software replace a sighting telescope of a traditional compass for target alignment, and efficiency is greatly improved. In an efficiency measurement test, the measurement efficiency of the instrument is twice that of a traditional instrument.

This invention discloses a composite catalyst with high resistance to coking in methane dry gas reforming. The composite catalyst is composed of three components: a supported nickel-based catalyst, a nitrogen-doped carbon-supported molybdenum catalyst, and an inert support. The invention also discloses its application in methane dry gas reforming. After hydrogen reduction, the composite catalyst exhibits excellent activity and stability under high space velocity reaction conditions in methane dry gas reforming, overcoming the deficiency of Ni-based catalysts in methane dry gas reforming that are prone to coking and deactivation under high-temperature conditions.

The present application discloses a perovskite catalyst and its preparation method and application. The perovskite catalyst is modified by silicon dioxide to form a three-dimensional ordered macroporous perovskite catalyst; wherein the perovskite is AFe x M 1‑x O3, A is selected from at least two of La, Ce, Ca, Sr and Ba, M is selected from at least one of Cr, Mn, Co or Cu, and 0≤x≤1. The present invention regulates the A / B element ratio on the surface of the three-dimensional macroporous perovskite modified with silica by chemical etching and high-temperature atmosphere calcination. The method can significantly increase the proportion of perovskite and the number of active sites in the catalyst, while effectively inhibiting the sintering of the catalyst, so that the catalyst has both high activity and high stability. The catalyst is simple to prepare and has good repeatability. It exhibits excellent reaction performance in the decomposition of high-concentration hydrogen peroxide, the decomposition of methanol and the conversion of chemical chain low-carbon alkanes, and has potential application prospects.

The application discloses a kind of commercial vehicle hybrid system engine start-stop control method, it is related to new energy automobile control technical field, comprising the following steps: after vehicle is in Ready state, obtain vehicle operating state information, judge whether engine satisfies starting condition, when satisfying starting condition and no relevant fault, engine start is executed, and engine starting time is recorded;After engine completes starting, engine continuous operation time is calculated according to the engine starting time recorded in step one, when continuous operation time reaches time, engine shutdown determination condition is opened.The application sets up running time restriction and shutdown delay control, avoids engine frequent start-stop, control logic is simple, can be realized by software calibration, without increasing hardware.Through multi-stage start-stop control, reduce impact load, improve engine and related components reliability, while taking into account power and economy, meet the long life operation demand of commercial vehicle.

The application relates to a Saccharomyces cerevisiae and application thereof in preparation of sensitive skin soothing repair, anti-inflammatory, anti-aging and antioxidant products, and belongs to the technical field of microorganisms. The technical problems to be solved are that the existing Saccharomyces cerevisiae is not subjected to directional breeding for skin care effects, the accumulation amount of intracellular glutathione, active polypeptide and other effective substances is low, the comprehensive performance of fermentation products is poor in terms of antioxidant, anti-inflammatory and anti-aging, meanwhile, the content of traditional strain fermentation ethanol byproducts is high, the raw materials are highly irritating, the safety is insufficient, and the single yeast raw material cannot realize multiple skin care effects, and the natural application is limited. Technical scheme points are to provide a Saccharomyces cerevisiae CS-02, which is preserved in the China General Microbiological Culture Collection Center on March 24, 2026, and the preservation number is CGMCC No. 39164, and application of the Saccharomyces cerevisiae in preparation of sensitive skin repair soothing, anti-inflammatory, antioxidant and anti-aging products.

The invention discloses a work scheduling method for a deep neural network reasoning task. The work scheduling method is applied to a vehicle edge computing system comprising a plurality of roadside units. The method comprises the following steps: firstly, constructing a system model, and modeling an inference task into a directed acyclic graph to distinguish executed and unexecuted layer nodes; establishing a communication time delay and energy consumption model, and defining a system total cost function as a weighted sum of task completion total time delay and intermediate result transmission energy consumption; modeling a scheduling decision process into a multi-agent Markov decision process, and taking each roadside unit as an independent agent; solving by adopting a multi-agent depth deterministic strategy gradient algorithm to generate an optimal scheduling strategy; and finally, determining a target roadside unit according to a strategy, and transmitting an intermediate result for relay calculation. According to the invention, the problems of service interruption caused by high mobility of the vehicle and lack of cooperation of multiple roadside units are effectively solved.

The invention discloses a multi-unmanned aerial vehicle auxiliary wireless emergency communication optimization method, and belongs to the technical field of communication. According to the invention, on the basis of downlink communication energy budget of each unmanned aerial vehicle, a system throughput maximization problem of a multi-unmanned aerial vehicle assisted wireless emergency communication system is modeled into a mathematical optimization problem under multiple constraints; the method comprises the following steps: decomposing an original non-convexity optimization problem into three sub-problems of user association strategy optimization of an unmanned aerial vehicle and a ground user, unmanned aerial vehicle transmitting power strategy optimization and unmanned aerial vehicle trajectory optimization based on a block coordinate descent idea; a non-convex sub-problem is converted into a convex problem which can be solved through a corresponding convex optimization tool by introducing auxiliary variables, continuously convex approximation and other mathematical means, and the three sub-problems are iteratively solved through an alternate optimization mode to obtain an approximate optimal solution of an original problem. According to the invention, accurate management of communication energy can be realized, and the capacity of the system is effectively improved.

The invention discloses an environment-friendly long-acting antifouling paint and a preparation method thereof, and belongs to the technical field of paint materials and antifouling. The environment-friendly long-acting antifouling paint comprises the following components in parts by weight: 30-50 parts of a component A, 25-45 parts of a component B, 8-25 parts of a functional body, 1-5 parts of color paste and 0.3-3 parts of an auxiliary agent, the component A is an aliphatic isocyanate prepolymer which is prepared from aliphatic isocyanate and castor oil through a reaction, and the specific preparation raw materials comprise the following components in parts by weight: 20-40 parts of aliphatic isocyanate and 10-20 parts of castor oil; the component B is polyaspartic acid ester, and the molecular weight of the polyaspartic acid ester is 800-1500 g / mol; the castor oil has good reaction activity with aliphatic isocyanate, can significantly improve the flexibility of a prepolymer (polyurea precursor) and the cohesiveness with a base material, and can optimize the compatibility of the polyurea precursor with polyaspartic acid ester with the molecular weight of 800-1500 g / mol, thereby improving the mechanical stability and weather resistance of the whole polyurea paint coating. The technical problem that an existing coating material cannot give consideration to synergistic improvement of environmental protection property and mechanical property is solved.

The application relates to a multi-modal emotion recognition method and system fusing bidirectional attention and self-distillation, first, a bidirectional cross attention mechanism is designed to realize bidirectional interaction of information flow between modes, so that text, speech and video can be mutually corrected to mine deeper complementary clues. Secondly, a self-distillation type double-strategy learning framework is introduced to realize exploration of a global optimal solution through parallel conservative fusion and aggressive fusion paths, and the adaptability and generalization performance of the model in low-labeled data, noisy labeling and cross-domain migration scenarios are enhanced. Finally, a mode contribution balance mechanism is constructed to dynamically estimate the reliability and importance of different modes in different samples, suppress the interference of low-quality modes, and enhance the contribution of high-quality modes, thereby significantly improving the robustness and stability of the model in complex real scenarios. The application improves the fusion efficiency of text, speech and video modal information and the fine-grained emotion recognition precision.

The invention discloses an X-ray detector based on a sulfur halogen semiconductor Cu2Br2-nInSe6 and a preparation method thereof, the detector takes a Cu2Br2-nInSe6 crystal as a detection semiconductor, and n is more than or equal to 0 and less than or equal to 2; metal electrodes are arranged on the upper and lower sides of the crystal; the preparation method comprises the following steps: (1) preparing a Cu2Br2-nInSe6 crystal; (2) cutting the Cu2Br2-nInSe6 crystal into wafers, and carrying out surface polishing on the Cu2Br2-nInSe6 crystal; and (3) preparing a metal electrode on the wafer. The high-performance X-ray detector based on the Cu2Br2-nInSe6 crystal is prepared through the design of the crystal structure, the device structure is simple, the X-ray detection performance with high sensitivity and low detection limit is achieved, and the detector has outstanding advantages in the field of radiation detection application.