52results about How to "Achieve capture" patented technology

The invention provides a post-retinal detachment reduction body position management method and system and wearable equipment, and relates to the technical field of medical monitoring. The method comprises the following steps: continuously acquiring head motion data through an inertial measurement unit integrated on head-mounted equipment, resolving the data through an embedded processor to obtain a real-time head posture angle, and comparing the real-time head posture angle with a preset rehabilitation position angle range to generate a posture comparison result; meanwhile, neck strain signals are collected through a flexible strain sensor array, and biomechanical characteristic signals are obtained through signal conditioning; and finally, the two types of data are fused, and a judgment result of the consistency of the overall body position of the user is generated through a classification algorithm, so that multi-modal data fusion analysis is realized, the postoperative body position state can be accurately monitored in real time, immediate feedback is provided, and the accuracy of body position management and patient compliance are effectively improved.

The application provides a method and device for simulating a water delivery system of an impulse water turbine and electronic equipment, and the method comprises the following steps: establishing a mathematical model of the water delivery system of the impulse water turbine with variable step grids; dividing the water delivery pipe upstream of the distribution ring and the side pipe of the distribution ring into grids with the variable step grids; determining the boundary conditions of the inlet and outlet of the pipe based on the calculation working condition, and determining the initial flow and pressure values of each grid node according to the constant flow theory; combining the initial flow and pressure values of the grid nodes, iteratively calculating the corresponding flow and pressure values of each grid node at the current time, and determining the unit side parameters at the current time until the termination condition is met. Through the application, certain theoretical basis and technical support are provided for the safe and stable operation of the impulse unit, and the problem that the existing related technologies are difficult to balance the pipe calculation efficiency and the calculation accuracy is solved.

The application discloses a multi-lamp-bar type lamp strip structure and relates to the field of mechanical equipment for textile industry, which comprises a long-strip-shaped mounting plate and an LED lamp bar, the LED lamp bar comprises a strip-shaped conductive substrate and a plurality of lamp beads arranged on the conductive substrate, the conductive substrate is arranged on one side of the mounting plate and detachably fixedly connected with the mounting plate through a threaded part, a plurality of the lamp beads are arranged in multiple rows along the width direction of the conductive substrate; and the lamp beads comprise white light lamp beads and purple light lamp beads, a plurality of the purple light lamp beads are uniformly arranged along the length direction of the conductive substrate and form a purple light lamp strip, and a plurality of the white light lamp beads are uniformly arranged along the length direction of the conductive substrate and form a white light lamp strip. Through the dual-light-source space cooperation, the end brightness compensation, the modular heat dissipation and the quick maintenance architecture, the application precisely matches the strict requirements of the cotton foreign fiber removing equipment on the light source stability and the imaging uniformity, finally realizes efficient and durable lighting support, and improves the comprehensiveness and sensitivity of the foreign fiber machine in detecting foreign fibers.

This invention provides a method and device for anti-pinch control of electric seats. The method includes real-time acquisition of current, speed, and physical angle during motor operation; when the current motor current exceeds a dynamic current threshold or the transient collision energy exceeds a collision energy threshold, and the current speed is less than a expected reference speed, the motor is controlled to stop or reverse. The dynamic current threshold is determined comprehensively based on a preset angle reference ammeter, load mass compensation, ambient temperature compensation, and a mechanical wear self-learning factor updated with the number of opening and closing cycles; the transient collision energy is calculated only when the transient current change rate exceeds a dynamic slope threshold, at which point an integral observation window is activated. The electric seat anti-pinch control method and device provided by this invention achieve anti-pinch functionality during the operation of the electric seat.

The application provides a mercury adsorbent based on sulfur abundance regulation and a preparation method and application thereof. The mercury adsorbent is a sulfur-poor type or a sulfur-rich type, and is prepared from precursors with a molybdenum-sulfur molar ratio of 1:1-8:1 or 1:3-1:30 through a hydrothermal reaction. The sulfur-poor type has an amorphous structure with sulfur vacancy defects, and the sulfur-rich type has a three-dimensional flower-like morphology. The application realizes the synergistic unity of high capacity, strong acid adaptability and high selectivity through sulfur abundance regulation, and is particularly suitable for deep mercury removal of non-ferrous metal smelting waste acid wastewater.

The invention belongs to the technical field of inter-satellite laser link establishment pointing, and provides an inter-satellite laser link establishment pointing method and system suitable for a low-orbit mobile communication constellation in order to solve the problem that an existing azimuth angle and pitch angle pointing technology based on an east-north-sky coordinate system has defects. According to an inter-satellite laser communication ephemeris mutual transmission mechanism, a mathematical model of inter-satellite laser link establishment angle pointing requirements is established, and azimuth angle and pitch angle calculation formulas of laser link establishment of two adjacent satellites in a constellation orbital plane, satellites in the orbital plane and satellites in two adjacent orbital planes are deduced. On one hand, by introducing an orbital coordinate system, an inter-satellite azimuth angle and a pitch angle are defined; on the other hand, a mathematical model of an inter-satellite laser link establishment pointing angle is obtained based on a positioning orbit determination ephemeris data source of an on-satellite GPS. And the inter-satellite laser link establishment pointing technology can be realized through a two-dimensional rotating mechanism of the laser. The method can be widely applied to an inter-satellite laser link establishment pointing scene of a low-orbit mobile communication constellation.

The application discloses an Internet of Things anomaly detection method and system based on quaternion state space diffusion enhancement, and belongs to the technical field of network security and artificial intelligence. The method comprises the following steps: mapping traffic time sequence features into a quaternion tensor to maintain the internal coupling relationship of multi-dimensional features; designing a double-flow encoder, extracting continuous flow features by using a quaternion selective state space model, and modeling discrete protocol features by using a dynamic hypergraph neural network; generating an adversarial pseudo-anomaly sample by using a latent diffusion model to perform self-supervised pre-training, and combining a quaternion cepstrum distance loss to optimize the representation; optimizing the injected learnable prompt vector in the small sample fine-tuning stage, and correcting the class prototype by using a semi-supervised expectation maximization algorithm; calculating sample anomaly scores based on an energy model to realize known attack classification and unknown anomaly determination. The application significantly improves the generalization ability of the model under the condition of small samples and the detection ability of unknown threats.

The present application relates to the field of biological identification, and provides a multi-modal impression identification method based on causality, which comprises obtaining first multi-modal data of a speaker and second multi-modal data of a listener; performing causality identification on the first multi-modal data and the second multi-modal data through a preset causality model to obtain first features; performing feature extraction on the second multi-modal data through a preset first BLSTM network to obtain second features; performing cross-domain fusion on the first features and the second features through a preset attention network to obtain first fusion features; and performing impression identification on the first fusion features through a preset full connection network to obtain a first impression identification result. The preset causality model is set to capture the causality between the speaker and the listener. The preset attention network is set to realize cross-domain fusion of the related information between the two, improve the accuracy of impression identification, and improve the performance of identifying impressions from a single aspect.

The invention provides a sedentariness reminding method and device, computer equipment and a computer readable storage medium. The method comprises the following steps: collecting working state associated information of a user within a preset duration; determining a first user behavior feature of the user within a predetermined time length based on the working state association information; according to the acquisition time of the working state associated information corresponding to each first user behavior feature, constructing all the first user behavior features into a user behavior sequence arranged according to the acquisition time; extracting a second user behavior feature from the user behavior sequence through a predetermined feature extraction mode; taking the second user behavior characteristics as input information of a preset fatigue degree prediction model, and outputting the fatigue degree of the user through the fatigue degree prediction model; and if the fatigue degree is greater than a preset fatigue threshold value, executing a sedentariness reminding strategy. According to the scheme, real-time accurate prediction of the sedentariness fatigue state of the user is realized, and the accurate opportunity suitable for sedentariness reminding can be captured.

The invention provides a fault diagnosis method and device based on causal state space prompt, a medium and electronic equipment, and relates to the field of data processing, and the method comprises the steps: mapping each IT operation and maintenance event into an event feature vector; inputting the online state space model according to a time sequence, and outputting a potential hidden state corresponding to each time step; in response to the received IT operation and maintenance diagnosis request, mapping the potential hidden state corresponding to the current time step into a plurality of discrete state lexical elements; and splicing the problem description corresponding to the IT operation and maintenance service interruption, the IT operation and maintenance event occurring in the target time window and the plurality of discrete state lexical elements into a target prompt, and inputting the target prompt into the frozen large language model to obtain a fault diagnosis result. The method not only has the advantage of natural language reasoning of the large language model, but also avoids the illusion problem caused by limited context windows and missing of key causal information, and meanwhile, the use mode of freezing the large language model greatly reduces the training cost and the computing power consumption.

This invention discloses a surface-source blackbody structure with a multi-level composite cone structure and its fabrication method, belonging to the field of microstructure optical design and fabrication technology. This invention forms a multi-level composite absorbing unit by setting a periodically arranged first-level main cone structure on a basic support structure, and further constructing a second-level secondary cone structure array with gradient characteristics on its top platform. This not only increases the optical path and multiple internal reflections, but also matches different wavelengths through gradient dimensions. Since the secondary cone array contains multiple scales, energy is absorbed step by step during the transmission process from the outside to the inside; simultaneously, it can achieve efficient absorption over a wide wavelength range and large angles, thus structurally realizing a near-ideal planar blackbody target surface, which is beneficial for promoting its integrated application in infrared calibration, thermal imaging system calibration, and space remote sensing.

The invention discloses application of a molecular genetic marker in simultaneous prediction of intestinal immunity, egg weight at first laying and body weight at first laying. The molecular genetic marker disclosed by the invention can be used for simultaneously marking the intestinal immune function, the egg weight at the first laying time and the body weight at the first laying time; by means of the detection method, one-time detection can be achieved, and early, accurate and efficient prediction of the intestinal immune function, the first laying weight and the first laying egg weight of the chickens can be completed. The'one-point three-detection 'technology accords with the mainstream goal of laying hen breeding, and can save the detection cost: taking a single sample detection condition as an example, under the detection method of the same type of PCR principle, corresponding sites are respectively detected for three times if the three characters are detected, however, the prediction of the three characters can be realized by only detecting for one time in the invention, and the detection cost is greatly reduced. The detection cost is reduced by 2 / 3, and meanwhile, the important characters of a plurality of chickens can be predicted.

This invention discloses a wavelength division multiplexing (WDM) trapping device and a method for measuring extremely weak forces in a single-beam vacuum optical trap. A fiber laser emits a laser beam, which passes sequentially along the fiber through a fiber acousto-optic modulator, a fiber beam splitter, and a 1064nm fiber collimator to obtain beam A. Beam A is transmitted through a dichroic mirror into a vacuum cavity, trapping and detecting micro / nano particles. Beams B, C, and D, obtained through multiple fiber beam splitters, a PPLN waveguide frequency doubling module, a fiber electro-optic modulator, and a 532nm fiber collimator, sequentially enter the vacuum cavity to triaxially cool the micro / nano particles in a vacuum environment. At this point, the micro / nano particles are in a stable state. A voltage is applied to the electrode plate, reducing the intensity of beam A, ultimately achieving a balance between electric field force and gravity. The electric field force is then used as the measurement value for the gravity of the micro / nano particles. This invention employs a wavelength division multiplexing method, achieving triaxial cooling, trapping, and detection with only one light source, and using optical fiber to replace most optical components, resulting in greater integration.

The invention relates to the technical field of coal mine underground dust prevention and control, and discloses a working face negative pressure ventilation dust falling method, system and device. The method comprises the following steps that an intelligent control center obtains the real-time traction speed and the real-time coordinate position of a coal mining machine through a dust source tracking and positioning sensing unit; establishing coordinate mapping logic according to the real-time coordinate position so as to calculate the number of the hydraulic support currently corresponding to the coal mining machine; judging the running direction of the coal mining machine according to the sign sign of the real-time traction speed, and generating a support number set in an activated state at the current moment by combining a preset advanced dust pre-suction support number and a preset lagging purification support number; and sending an instruction to the distributed negative pressure dust collection device array according to the support number set in the activated state, and controlling the gas circuit control units in the support number set in the activated state to be started, so that a negative pressure suction and capture area is formed around the roller of the coal mining machine. According to the invention, accurate capture of the mobile dust production source and on-demand adjustment of the air volume are realized.

The application discloses a wire-driven multi-mode folding and unfolding gripper with a flexible joint, which has a folding and unfolding gripping part composed of three X-shaped scissors units in a circumferential direction, and adjacent scissors units are connected through connecting assemblies. The bottom of the folding and unfolding gripping part is connected with the flexible joint, and in an initial state, the whole is in a folded state under the action of a torsional spring. Motion control of the application is realized through three driving wires in the circumferential direction, including bending motion control that can pass through a narrow and curved pipeline, capture motion control after reaching a target position, and locking control after successful capture. Meanwhile, the application proposes a driving wire control strategy to realize the control and switching of various motion modes of the gripper. In summary, the application has a small size in the folded state to pass through a narrow cavity environment, and the operation task is completed by virtue of the reconstruction of the mechanism motion mode and the change of the driving wire control law.

This invention discloses a medical image segmentation method based on dynamic deformable convolution and a sliding window adaptive complementary attention mechanism, which improves the perception of small lesions and highly deformed targets in medical images, as well as the ability to distinguish between segmented targets and the background. Dynamic deformable convolution, through task-adaptive learning, can flexibly change weight coefficients and deformation biases, enhancing the ability to express local image features and achieving adaptive extraction of spatial features. The sliding window adaptive complementary attention mechanism achieves cross-dimensional global modeling of medical images through a self-attention branch of adaptively learned weight coefficients. This mechanism overcomes the shortcomings of conventional methods in modeling cross-dimensional relationships between space and channels, and can capture long-distance cross-dimensional correlation features in images. Furthermore, the parallel interactive approach combines local and global features at different resolutions to enhance representation learning, maximizing the preservation of both local and global features in medical images.

The application provides a computer vision defect detection system and method on an electric control board production line, and relates to the technical field of data processing.The method comprises the following steps: according to collected multi-angle images, combining illumination conditions and visual angle information, the overall area of the electric control board is partitioned; image block processing is performed to generate a plurality of image sub-areas, and texture features, brightness distribution features and geometric edge features are extracted in each image sub-area; rare defect feature enhancement processing is performed, low-frequency abnormal textures are repeatedly superimposed in multiple scales, and the edge fracture is directionally extended; the corresponding normal area combined features are compared and differenced, and normalized correction is performed in combination with the illumination conditions and the visual angle information; multi-angle recurrence degree analysis is performed, when the same suspected defect is detected under different angles, a reliable defect area is formed, otherwise, an interference area is removed, and an electric control board defect detection result is generated; and the application improves the accuracy of defect detection.

The utility model discloses a system for reforming hydrogen from natural gas. The system comprises a desulfurizing tower, a steam mixer, a preheater, a reforming reactor, a water-gas conversion tower, a cooling tower, a gas-liquid separator, a carbon dioxide absorption tower and an adsorption tower which are sequentially connected through gas pipelines, a booster pump, a pressure control valve and an emergency cut-off valve are arranged on the upstream gas pipeline of the desulfurization tower; a regulating valve is arranged between the desulfurizing tower and the steam mixer; a temperature control valve and a high-temperature switching valve are arranged between the reforming reactor and the water-gas conversion tower; a purified gas control valve is arranged between the carbon dioxide absorption tower and the adsorption tower, and a product hydrogen outlet valve is arranged at the downstream of the adsorption tower; the system further comprises a carbon dioxide capturing pipeline connected with the carbon dioxide absorption tower, the carbon dioxide capturing pipeline is provided with a regeneration tower and a carbon dioxide compressor, and one end of the carbon dioxide capturing pipeline is connected with a liquid carbon dioxide storage tank. The whole system is simple and reasonable, the production cost is relatively low, carbon capture can be achieved, and the system is more environmentally friendly.

This invention provides a multimodal trajectory prediction method based on a conditional variational autoencoder, belonging to the fields of computer vision and machine learning. The technical solution includes the following steps: S1, data preprocessing and trajectory sample extraction; S2, multimodal feature extraction module; S3, cross-modal attention fusion; S4, conditional variational autoencoder generation; S5, loss function optimization; S6, model evaluation, calculating standard trajectory prediction metrics on a test set to evaluate model performance. This invention is applicable to trajectory prediction tasks in autonomous driving, robot navigation, and intelligent transportation systems. It can effectively integrate multimodal information such as historical trajectories and lane features to generate diverse trajectory prediction results with uncertainties, improving the accuracy and robustness of trajectory prediction.

The application discloses an intelligent operation and maintenance management system based on artificial intelligence, which comprises a multi-source data collection module, a data optimization module, a power plant equipment maintenance prediction model establishing module and an intelligent operation and maintenance management module. The application relates to the technical field of data processing, and particularly relates to an intelligent operation and maintenance management system based on artificial intelligence. The application innovatively proposes that equipment is taken as a network node, equipment process dependence is taken as a directed edge, an equipment maintenance network graph structure is constructed, the risk propagation chain among different equipment can be effectively identified, and power plant operation and maintenance management is changed from local prediction to global correlation modeling; a direct and indirect double-layer space dependence joint modeling mechanism is adopted, direct space dependence models the explicit process relationship among adjacent equipment through a graph attention network, indirect space dependence models the implicit correlation of non-adjacent equipment through a space attention mechanism, the prediction precision is improved, and the accuracy of intelligent operation and maintenance operation decision is improved.

The application relates to the technical field of construction safety management, and discloses a power engineering construction safety management system, which comprises a real-time risk assessment scheduling unit, a forward-looking human factor risk perception unit and a risk simulation mapping unit. The real-time risk assessment scheduling unit is used for generating a safety state characteristic map based on heterogeneous safety data of a construction site, identifying a risk event of the construction site through the safety state characteristic map, and generating an intervention instruction and a resource scheduling scheme to optimize the construction site. The application can comprehensively capture multi-dimensional information of the construction site by collecting personnel biological characteristics, equipment operation logs, environmental indexes and work video streams and other heterogeneous safety data through a multi-modal perception unit, breaks the limitation of single and one-sided data collection in the traditional management mode, and enables management personnel to obtain more rich and comprehensive on-site safety information, so that the overall safety condition of the construction site can be more clearly understood.

The present application belongs to the technical field of material chemistry and environmental governance, and discloses a plant decomposition type deodorant formula and a configuration method. In view of the defects of the existing deodorization technology, such as secondary pollution, slow effect, only masking odor, insufficient product stability and the like, the present application uses tea polyphenol extract, yucca extract, grape seed extract, various plant essential oils and other pure natural plant extracts as raw materials, does not contain chemical synthetic additives, and is prepared through a standardized process of precise control of constant temperature environment, grading stirring speed, step-by-step feeding and reaction time. The deodorant can target decomposition of various foul-smelling molecules, has the effects of bacteriostasis and oxidation resistance, is completely deodorized, has strong product stability, is safe and non-irritating, has no secondary pollution, and can be widely applied to foul odor treatment in multiple scenes such as livestock breeding, garbage treatment, sewage treatment, public environment and the like.

The application relates to the technical field of reservoir siltation monitoring, and discloses a reservoir underwater siltation dynamic monitoring system and method. The method comprises a layered intelligent sensor array of a pressure sensor, an acoustic Doppler current profile sensor and a fiber bragg grating strain sensor, and a point-line-surface three-dimensional monitoring network is formed; a plurality of source monitoring data such as pressure data, flow velocity data and strain data output by the array are collected in real time through an edge computing node; the plurality of source monitoring data are subjected to real-time fusion processing to generate fusion data; based on a water flow dynamics model and a sediment settlement model, dynamic inversion calculation is carried out on the fusion data to obtain siltation evolution trend data; an adaptive calibration algorithm is used to calibrate the siltation evolution trend data, sensor drift and water body interference are eliminated, and finally, calibrated siltation state monitoring results are generated. The method can realize comprehensive and real-time monitoring of reservoir underwater siltation, and provide accurate state information for reservoir siltation management.

The invention relates to the technical field of intelligent assistance, and discloses a method for assisting a blind person in daily life through voice of intelligent glasses, and the method comprises the steps: obtaining an environment image data stream, and obtaining spatial feature description through multi-level visual feature analysis; reconstructing a space-time motion mode to obtain a pixel displacement field, and performing cross-modal feature integration on the space feature description and the pixel displacement field to obtain fused visual features; constructing a semantic segmentation map by taking the feature vectors of the pixels in the fused visual features as nodes and taking the spatial adjacency relation of the pixels as edges; performing pixel clustering analysis on the semantic segmentation map to obtain candidate instances, and performing structure perception aggregation on the candidate instances based on spatio-temporal topological relevance of the candidate instances to obtain context enhancement features; calibrating and screening to obtain a key target description set, generating a natural language description sequence, and converting the natural language description sequence into voice output; the efficiency of the intelligent glasses for assisting the blind in daily life through voice can be improved.

The application discloses a kind of based on partition electrode's switchable dual-function acoustic metasurface device and method.Partition electrode acoustic piezoelectric transducer and flat plate polymer metasurface are included, partition electrode acoustic piezoelectric transducer includes the lower surface partition electrode, lead zirconate titanate piezoelectric ceramic sheet, upper surface non-partition electrode and matching layer sequentially stacked from bottom to top, and flat plate polymer metasurface is arranged on the upper surface of matching layer;Lower surface partition electrode is divided into two partial electrodes of mutually insulated center internal area electrode and peripheral external area electrode, and upper surface non-partition electrode is whole electrode.The structure of the present application only switches the input voltage of the area electrode of partition electrode to realize the dual-function switching of focusing sound field and bottle-shaped sound field, so that it simultaneously has the dual-function demand of acoustic energy focusing and "sound tweezers", and flat plate polymer metasurface is directly imprinted on piezoelectric ceramic transducer, improves sound field coupling efficiency, and the thickness is thinner and the price is good.

The invention discloses a dust treatment device for semiconductor material preparation, and belongs to the technical field of semiconductor processing, the dust treatment device comprises a box body, one end of the box body is fixedly connected with a first connecting plate, one end, far away from the box body, of the first connecting plate is fixedly connected with a first driving mechanism, the first driving mechanism is fixedly connected with a first dust collection head, and the first dust collection head is detachably connected with a flexible hose; a flexible hose is arranged in the box body and communicates with the interior of the box body, an air inducing mechanism is arranged in the box body and communicates with the flexible hose, the air inducing mechanism communicates with a dust treatment mechanism, the dust treatment mechanism comprises a first partition plate fixedly connected into the box body, and a first filtering part is fixedly connected between the first partition plate and the inner wall of the box body; a second filtering part is arranged on the side, away from the first partition plate, in the box. The flexible hose is matched with the driving mechanism, long-distance and multi-angle dust capturing is achieved, the application range is wide, by arranging the first filtering part and the second filtering part, the cleanliness of exhausted gas is ensured, and the high requirement of semiconductor production for the environment is met.

The application provides a knowledge representation learning method of an R-GCN and a Transformer fusion network in the field of information retrieval, relates to the technical field of knowledge graph representation learning, and comprises the following steps: constructing input features of entities and relations based on an information retrieval knowledge graph; constructing a Transformer branch representation vector of the entities, an R-GCN branch representation vector of the entities and a Transformer representation vector of the relations based on a Transformer model and an R-GCN model, and obtaining a final embedding representation of the relations; obtaining a fusion feature of the entities in the information retrieval knowledge graph as the final embedding representation of the entities based on the R-GCN branch representation vector of the entities and the Transformer branch representation vector of the entities; and optimizing the effect of knowledge representation learning through information retrieval knowledge graph entity classification and information retrieval knowledge graph path query answering tasks. The application captures local dependency relationships in a graph structure through a GCN, dynamically captures dependency information at different positions in sequence data through a Transformer model, captures global structure information, fully utilizes global structure information and context information in the information retrieval knowledge graph, and realizes more accurate knowledge representation learning.

This application relates to the field of seismic measurement technology, and provides an apparatus and method for measuring cross-fault motion in experimental earthquakes. The apparatus includes a metal rod, an elastic compression spring installed at the first end of the metal rod, and a rod axial force gauge installed on the metal rod. The method involves installing the apparatus on a fault model and calculating fault motion. Based on the tensile stiffness K1 of the apparatus for measuring cross-fault motion in experimental earthquakes and the data recorded by the rod axial force gauge, the axial displacement of the metal rod, the shear displacement in the fault shear direction, and the real-time velocity and acceleration of the fault motion at time t are calculated. The apparatus for measuring cross-fault motion in experimental earthquakes of this application has a simple structure, is easy to manufacture, and can capture vibration information during fault instability without the need for complex field equipment deployment. It is not easily affected by the environment, and the test results are accurate. The method can accurately describe the instability process of experimental earthquakes.

This invention proposes a large-scale array acoustic tweezers control device and method, relating to the field of ultrasonic technology. It includes: a host computer and an ultrasonic control system; the host computer is communicatively connected to the ultrasonic control system for sending control data to the ultrasonic control system; the ultrasonic control system includes: an interactive circuit board, a control circuit board, and ultrasonic transducers; the ultrasonic transducers are arranged in multiple ultrasonic transducer arrays, forming a control space; the interactive circuit board distributes control data to the ultrasonic control board, which controls the ultrasonic transducers to emit ultrasonic waves, synthesizing a preset sound field within the control space to control the position of the target object, achieving precise control of the ultrasonic transducers, capturing and manipulating the target object, with high synchronization accuracy between signal channels, a large number of controlled transducer array elements, and support for adjusting the frequency, phase, and amplitude of the driving waveform of a single array element. The array can be arranged into a planar array or other shapes to meet different experimental needs.

The invention relates to a Mars hyperspectral image classification method based on CNN and GCN, and belongs to the technical field of remote sensing image classification. The method comprises the following steps: constructing a double-branch hybrid neural network comprising a CNN branch and a GCN branch; wherein the CNN branch is of a lightweight network structure and comprises a denoising and dimensionality reduction module, a depth separable convolution module and a multi-scale convolution module; the GCN branch is composed of an HGNNConv network; inputting a to-be-classified Mars hyperspectral image into the double-branch hybrid neural network, and respectively extracting pixel-level features and super-pixel-level features of the Mars hyperspectral image through a CNN branch and a GCN branch; and introducing a channel attention mechanism and a space attention mechanism, carrying out weight distribution and fusion on the pixel-level features and the super-pixel-level features to obtain comprehensive features, and completing classification of the Mars hyperspectral images based on the comprehensive features. The objective of the invention is to solve the technical problems that pixel-level and super-pixel-level features cannot be fused efficiently and distortion is easy to occur in a feature propagation process in the prior art.