269 results about "Embedded processing" patented technology

The invention relates to a final-stage cache design method based on near memory calculation, and belongs to the technical field of near memory calculation. And compared with a near memory processor near a traditional main memory, data access and calculation can be carried out more quickly, and the calculation throughput rate can be improved. According to the near storage calculation, a near storage processor and a last-stage cache storage array are directly designed in an integrated mode, faster memory access is achieved through an internal bus interconnection mode, and the problem of a storage wall between the calculation speed and the memory access speed and the problem of a power consumption wall of data handling are relieved to a certain degree. The last-stage cache controller cooperates with the embedded processor, and can dynamically coordinate between a standard cache function and a calculation mode. Through a synchronization mechanism in a last-stage cache controller, a memory access request from a host CPU and a calculation task executed by an embedded processor in a last-stage cache are transparently balanced, so that efficient near-memory calculation under cache consistency is realized.

The invention provides a code quality adaptive evaluation and optimization method, system and device and a medium, and belongs to the technical field of computers. The method comprises the following steps: performing redundant information cleaning and standardization processing on an original code; selecting an adaptive parser to convert the processed code into an abstract syntax tree, and optimizing the abstract syntax tree; performing embedding processing on each node of the abstract syntax tree by using a pre-trained deep learning model to generate a node semantic vector, and performing aggregation, feature extraction and vector normalization operation to obtain a semantic vector representing code semantic features; extracting a dependency relationship from the abstract syntax tree to construct a program dependency graph, extracting node features by using a neural network model, and generating a dependency relationship analysis result; based on the semantic vector and the dependency analysis result, identifying a code potential problem by using an optimization algorithm and generating an optimization suggestion; codes are automatically modified according to optimization suggestions, the optimization effect is re-evaluated, and related models and algorithms are iteratively optimized according to feedback information.

The invention relates to the technical field of electrical equipment state monitoring and fault early warning, and particularly discloses an electrical cabinet condensation risk online monitoring method and system, and the method comprises the steps: synchronously collecting the physical parameters of the same microscopic space point in an electrical cabinet; calculating the classical dew point temperature by adopting a Magnus formula; performing quantification processing on the air pressure micro-fluctuation data in the time window, and calculating information entropy of micro-air pressure fluctuation; the wall surface temperature gradient and the air flowing state are fused, and the equivalent dew point temperature is calculated; real-time safety margin is calculated, and the instantaneous change rate of the safety margin is solved; constructing a condensation risk trend index; comparing with a preset threshold value, generating an early warning signal and a risk diagnosis label, and completing condensation risk monitoring; the system comprises a multi-physical field sensing unit, an embedded processing unit, a self-adaptive control unit, a man-machine interaction unit and a communication unit which are integrated in the same shell.

The invention provides an energy coordination control system and scheduling method for multi-energy complementary new energy storage, and belongs to the technical field of new energy storage and energy technology management. The data acquisition and processing module is connected with a current and voltage sensor, a power transmitter, a meteorological station sensor and a battery management system sensor, and is used for acquiring voltage, current, power, state of charge and environmental data of photovoltaic power, wind power, energy storage and load through the sensors; and sliding window filtering and wavelet de-noising processing are carried out through the embedded processor. The method has the beneficial effects that the uncertainty of wind and light output is effectively processed by establishing a layered distributed optimization architecture and adopting a method of combining multi-objective optimization and robust optimization, and the consumption rate of renewable energy sources can be remarkably improved, the net load fluctuation can be reduced and the energy consumption can be reduced on the premise of meeting various system operation constraints. Therefore, dependence on conventional fossil energy is reduced, and the economic operation level and the environmental protection benefit of the whole energy system are improved.

The invention discloses a millimeter wave barrier gate radar multipath interference suppression method and device, and belongs to the field of millimeter wave radar signal processing. The method comprises the following steps: S1, generating an RD image by an echo signal, and executing 2D-CFAR detection to obtain a candidate target unit; s2, eliminating ground reflection (GB) through an RD domain multipath sensitive area mask (MVZ); s3, converting the candidate target unit into a Cartesian coordinate system point cloud; s4, through point cloud domain mirror symmetry verification, removing a pole side mirror image (PSR); s5, clustering the reserved point cloud to obtain a target cluster; s6, through vehicle contour modeling, constructing a rear suppression area to reject vehicle-brake box multiple reflections (VBR); and S7, outputting effective target information after tracking processing. The device comprises a radar radio frequency front end, a signal processing unit, a point cloud post-processing unit, a tracking and control unit and an embedded processing platform. Through a three-layer cooperative inhibition mechanism, three types of main multipath interferences of a barrier gate scene are comprehensively covered, the detection accuracy is improved, static targets are prevented from being mistakenly eliminated, the engineering practicability is high, and safe and reliable barrier gate control can be ensured.

The invention discloses a voice and music collaborative generation method and system based on a dynamic mixed attention and expert architecture, terminal equipment and a medium, and relates to the technical field of audio generation. The method comprises the following steps: acquiring multi-modal input containing at least one of audio, text and vision, performing embedding processing on the multi-modal input, and mapping the multi-modal input to a unified space to obtain a fusion sequence; setting a plurality of attention heads, dynamically selecting the attention heads to activate experts through attention head routing gating, and weighting and aggregating calculation results to obtain mixed attention features; generating expert probability distribution through expert set routing gating, dynamically selecting an expert set to activate experts, and calculating task adaptation features; and performing audio generation processing based on the audio language modeling head to obtain a voice or music result. According to the method, collaborative generation of voice and music, dynamic allocation of computing resources, balance of domain-specific and cross-domain general knowledge learning are realized, the problems of task conflict and data imbalance are solved, and the audio generation quality and efficiency are improved.

The invention relates to the technical field of chip delay error dynamic compensation scheme design, in particular to a chip delay error dynamic compensation method and system based on TDC. The method comprises the following steps: measuring the transmission delay of a critical path in real time by using a time-to-digital converter core integrated in a chip; dynamically predicting a future delay error through an embedded algorithm based on the delay measurement value; and according to the predicted delay error, dynamic compensation is carried out on the chip timing sequence, and the compensation comprises adjustment of a clock signal or modification of path delay. A corresponding system includes a TDC core for real-time measurement, an embedded processing unit for running a predictive algorithm, and a dynamic compensation unit for performing compensation. Closed-loop control is formed through real-time monitoring, prediction and compensation, the influence of process, voltage and temperature changes on the time sequence of the chip can be effectively resisted, the time sequence precision, working stability and reliability of the chip are remarkably improved, and the method is particularly suitable for high-frequency and precise measurement chips.

The invention discloses a new energy automobile electric compressor preheating structure and a control method, and relates to the technical field of new energy automobile heat management. The method comprises the three steps of heat source collaborative recovery, heat demand dynamic distribution and self-starting collaborative execution, specifically, motor stator waste heat and compression refrigerant waste heat are recovered through a heat conduction fin array and a spiral flow channel, and a shared heat conduction medium is guided in; based on a fuzzy PID controller, a heat flow guide valve is dynamically adjusted to distribute heat flow according to the lubricating oil way temperature, the suction superheat degree and the shell temperature; the generator is triggered to be rapidly heated in a locked-rotor mode at the ultralow temperature, the preheating requirement is pre-judged in combination with a heat attenuation model, the structure comprises a temperature sensor set, a three-way heat flow guide valve and an embedded MCU processor, efficient waste heat utilization is achieved, external energy consumption is not needed, preheating energy consumption is effectively reduced, and the low-temperature starting performance and operation stability of the compressor are remarkably improved.

The invention discloses a therapeutic footwear system comprising a footwear body (101) with insole (102), midsole (103), and outsole (105), integrated with a multi-modal sensor array (110) including a plantar pressure matrix (111), inertial measurement unit (112), photoplethysmography sensor (113) and temperature sensor (114). An embedded processor (120) houses an adaptive sampling module (121), an edge-AI inference engine (122), a federated learning client (123), and memory (124) for local analysis. A feedback interface (130) with zonal haptic actuators (131) under medial forefoot (132), lateral forefoot (133), medial heel, and lateral heel (135) delivers corrective prompts. A power subsystem (140) comprising a battery management circuit (141), energy harvesting elements (142), and battery (143) sustains operation. The integration reduces latency, conserves energy, preserves privacy, and enables closed-loop therapeutic intervention during ambulation.

The invention relates to the technical field of microwave semiconductor testing, and discloses a dynamic calibration system for the sensitivity of a microwave photoconductivity tester, which comprises a dual-port orthogonal coupling resonant cavity module provided with a primary port and a secondary port which are orthogonally isolated; the variable capacitance-variable resistance composite modulation perturbation module is connected to the secondary port and is used for generating analog photoelectric response complex impedance disturbance; the vector microwave transceiving and demodulation module is connected to the main port and outputs I / Q signals; the complex impedance track fitting control module is used for driving the perturbation module based on a physical model; the dynamic vector calibration and signal processing module is used for calculating a system transfer function and executing vector projection denoising, the composite modulation perturbation module is used for generating complex impedance disturbance simulating semiconductor photoelectric response at a resonant cavity secondary port, and the vector receiving and transmitting module is combined to obtain a system instantaneous vector transfer function; a system drift error is eliminated by performing de-embedding processing on a measurement signal, and orthogonal noise is eliminated by using a vector projection algorithm.

The invention discloses an optical system of an infrared thermal imager, and the system comprises a super-structure lens module which employs a silicon-based super-structure lens, and achieves 3-5 [mu] m and 8-12 [mu] m multi-band focusing through a single sheet; the free-form surface lens is matched with the super-structure lens to correct asymmetric aberration; the multi-view-field switching mechanism is used for driving a zoom group through a micro stepping motor to realize switching between a wide angle and a narrow angle; the active athermalization compensation unit comprises an MEMS driver and a temperature sensor, and the lens spacing is dynamically adjusted through closed-loop control; and the intelligent processing module is integrated with an embedded FPGA processor and operates a deconvolution algorithm to suppress diffraction stray light at the edge of the super-structure lens. According to the optical system of the infrared thermal imager, the problems that a traditional infrared optical system is large in size, high in weight and poor in environmental adaptability are solved through the innovative optical design and manufacturing process, and efficient and long-time patrol flight detection of the unmanned aerial vehicle-mounted infrared thermal imager is achieved.

The invention discloses a low-power-consumption electromagnetic water meter system with dynamic excitation and self-calibration functions, which comprises a sensor layer, a metering board, a main control board and a communication board, and is characterized in that the sensor layer comprises an excitation coil, an electrode pair, a temperature sensor and a pressure sensor; the metering board comprises an excitation driving circuit, a signal conditioning circuit, a high-precision ADC and a register block, the main control board comprises an embedded processor, an RTC clock, a data storage module and a display screen, and the embedded processor is used for receiving signals and processing and executing tasks according to a pre-written program; the RTC clock is used for providing an accurate and continuous time reference for the embedded processor; the communication board comprises an NB-IoT / 485 / PWM multimode communication interface, when water flow passes through the measuring pipe, the electrode pair generates an induced voltage signal, after the induced voltage signal is converted by the signal conditioning circuit and the high-precision ADC, the main control board optimizes excitation parameters in real time by combining environmental parameters and a cloud platform strategy, and meanwhile, triple calibration is periodically executed to ensure the precision of a signal chain.

The present application relates to the technical field of electrical equipment state monitoring and fault early warning, and specifically discloses a kind of electrical cabinet condensation risk on-line monitoring method and system, the method comprises: the physical parameters of the same microcosmic space point in electrical cabinet are synchronously collected;Adopt the formula of Magnus to calculate classical dew point temperature;Quantitative processing is carried out to the air pressure microfluctuation data in time window, and the information entropy of micro air pressure fluctuation is calculated;Fusion wall temperature gradient and air flow state, calculate equivalent dew point temperature;Calculate real-time safety margin, and solve the instantaneous change rate of safety margin;Condensation risk trend index is constructed;Compared with the preset threshold, generate early warning signal and risk diagnosis label, complete condensation risk monitoring;The system includes integrated in the same shell multiple physical field sensing unit, embedded processing unit, adaptive control unit, man-machine interaction unit and communication unit.

The invention discloses an automobile data recorder intelligent voice interaction system based on a cloud large model, and the system comprises a vehicle-mounted terminal device which integrates a multi-mode sensing module, a lightweight AI embedded processor and an Internet of Things communication module; the invention also discloses an automobile data recorder intelligent voice interaction method based on the cloud large model, and the method comprises the following steps: S1, synchronously collecting voice signals and video streams through the multi-mode sensing module, and then carrying out noise reduction processing on the collected voice signals through the lightweight AI embedded processor. According to the method, through cooperative work of vehicle-mounted end preprocessing and a cloud multi-modal fusion analysis engine, composite requests including space-time and target characteristics, such as a black car storing right side compaction line lane changing, can be accurately responded, the defect of semantic understanding mechanization in the prior art is overcome, and the success rate of instruction analysis is remarkably improved.

The invention discloses a mobile audio-visual cooperative unmanned aerial vehicle detection system, and belongs to the technical field of intelligent detection and automatic control. The system comprises a mobile ROS unmanned vehicle platform, an audio-visual perception module, an embedded processing unit and an execution mechanism. The mobile ROS unmanned vehicle platform comprises a motion control system, a serial port communication module and a wireless communication module; the audio-visual perception module comprises an infrared visual acquisition unit and a four-channel audio acquisition array and is used for acquiring infrared image information and acoustic signals of the unmanned aerial vehicle; the embedded processing unit comprises a multi-modal data fusion processor and a motion cooperation controller and is used for synchronously processing the infrared image information and the acoustic signal and generating a control instruction; the execution mechanism comprises a two-degree-of-freedom steering engine holder and a self-adaptive cruise system. According to the invention, through audio-visual cooperation and dynamic mode switching, a fixed detection scene is not needed, continuous tracking of the unmanned aerial vehicle and blind area complement detection can be realized, and detection continuity and accuracy are improved.

The invention discloses a mixed gas flow online calibration device and method for a cable combustion test device, and belongs to the technical field of cable and optical cable combustion performance detection. The device comprises a propane gas source, an air source, a pressure reducing valve group, a mass flow controller, a mixed gas online calibration module group, a Venturi mixer and a blowtorch, a mixed gas online calibration module is arranged on the downstream of a mass flow controller, and a temperature and pressure sensor, a differential pressure mass flow meter and an embedded processor are combined, so that real-time monitoring and online calibration of mixed gas flow are achieved; according to the method, a flow drift curve can be fitted, the output error of the controller is corrected, and the influence of dismounting calibration and pipeline leakage is avoided; compared with the prior art, the device has the advantages of being high in precision, simple in structure, capable of self-checking, high in stability and the like, the consistency of the blast burner fire source power and the flame shape in a GB / T 31248 test is effectively improved, and the accuracy and repeatability of a cable combustion performance test are guaranteed.

The present application relates to thin strip continuous casting production technical field, specifically to a kind of automatically controlled double-roller thin strip continuous casting production process in molten pool height method, comprising: laser ranging module and visual perception module are arranged;Embedded processing module processing;Using deep learning algorithm establishes molten pool height dynamic prediction model, outputs molten pool height prediction value;Using adaptive sliding mode control as main controller, the displacement of servo hydraulic stopper is output;Based on the displacement of servo hydraulic stopper, molten pool height control is carried out.The method utilizes laser to provide accurate distance reference, visual sensing provides molten pool liquid level information, improves detection accuracy and reduces environmental interference through machine vision algorithm;And based on process parameters and real-time information of sensor, molten pool height prediction model is constructed using neural network algorithm with physical constraint;Join reinforcement learning optimizer to adjust sliding mode control parameters online, so that it adapts to working condition change to carry out high-precision, fast-response molten pool height control.

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 invention relates to the field of photoelectric detection and remote sensing monitoring, and provides an intelligent forest fire identification method. The method comprises the steps that image data of a monitoring area are acquired in real time through a multispectral data acquisition module, and the image data comprise thermal imaging data and visible light data; an embedded processing unit is used for processing image data in real time, and automatic identification and positioning of fire points are achieved through a multispectral data fusion algorithm; the power output of a power supply system is dynamically adjusted by adopting an intelligent energy management strategy to support all-weather operation of equipment, and the power supply system comprises a renewable energy power generation unit and an energy storage unit; based on the topographic data, adjusting the spatial attitude of the intelligent forest fire identification equipment through an adjustable supporting structure; after a fire point is identified, alarm information is transmitted to a monitoring center through a communication module, and an emergency response mechanism is triggered. According to the application, high accuracy and all-weather stable operation of forest fire monitoring are realized through multispectral collaborative identification and intelligent energy management.

The invention discloses a 6U dual-core radio frequency signal processing method and system based on a heterogeneous computing architecture, and relates to the technical field of radio frequency signal processing. The method comprises the following steps: the embedded processing unit analyzes an external task instruction and distributes the external task instruction to the radio frequency direct sampling unit; the radio frequency direct sampling unit collects a radio frequency signal and generates a noisy baseband data stream. The auxiliary sampling ADC synchronously collects broadband switch noise, and generates an interference signal through a time sequence AI model; the hardware acceleration unit cancels the baseband data stream based on the interference signal to generate a clean data stream; and executing hardware acceleration processing on the clean data stream and writing the clean data stream into a shared memory unit. The technical problem that broadband synchronous switching noise generated by a digital circuit interferes with a radio frequency receiving link, and consequently the quality of a baseband data stream is reduced is solved, and the technical effect that the precision, reliability and real-time performance of radio frequency signal processing are improved through real-time noise cancellation and hardware acceleration processing is achieved.

Low size, weight, power and cost (SWaP-C), flexible programmability and rapid run-time reconfigurability are desired features for the design of adaptive, underwater wireless communication modems for micro autonomous underwater vehicle (μAUV) swarms. Disclosed herein are first-generation and second-generation software and hardware design of a new class of low SWaP-C underwater modems that can accommodate wide-band acoustic front-ends and achieve s-level reconfigurability during run-time. By dynamically mapping certain signal processing operations of the acoustic communication stack to the Field Programmable Gate Array (FPGA) and others to the embedded processing system of the modem's System-on-Chip, data rates of up to 2000 bps can be achieved in ranges up to 50 m with the implementation of custom programmable logic.

The invention discloses a system and a method for identifying electric equipment by using a null line electromagnetic waveform, and belongs to the technical field of non-intrusive load monitoring. The system comprises a high-frequency current sensor which is installed on a main zero line of a distribution box and is used for collecting an instantaneous current signal in a non-contact manner; the signal conditioning module comprises a 3kHz-150kHz band-pass filter and an amplifying circuit; the sampling rate of the analog-to-digital converter is not lower than 500kS / s, and the quantization bit is greater than or equal to 12 bits; the embedded processing unit is internally provided with a one-dimensional convolutional neural network model combined with a channel attention mechanism, and the model comprises at least three one-dimensional convolutional layers, two pooling layers, an attention layer and a full-connection output layer; and the storage module is used for storing equipment identity, waveform characteristics and type information. According to the method, live wire interference is effectively avoided by analyzing the high-frequency electromagnetic waveform of the null line, high-precision and high-robustness equipment identification can still be realized in complex scenes such as multi-equipment concurrency and power grid disturbance, and the method is suitable for smart home and energy management systems.

The present application discloses a kind of railway subgrade settlement detection system and detection method, railway subgrade settlement detection system includes being arranged on the wrist arm column of one side of track high-definition image acquisition device, track gauge measuring device, embedded processing board, data transmission device and battery module, and reference point being arranged on the wrist arm column of other side of track and target point being arranged on the sleeper of track two sides, high-definition image acquisition device is used to collect reference point and target point image and transmit to embedded processing board, track gauge measuring device is used to measure the distance between target point and high-definition image acquisition device and transmit to embedded processing board, embedded processing board is connected with data transmission device, the present application carries out settlement data calculation by embedded processing board, data is transmitted to data processing center by data transmission device, data processing center detects whether settlement occurs and timely early warning, can real-time monitoring and timely discover subgrade condition, improve the detection efficiency of subgrade settlement, and detection precision is high.

According to the near-zero-intervention intelligent regulation and control system and method for the power station unit, internal and external data of a power station are collected and gathered, sensing data of internal equipment of the power station and external power grid data are integrated for the first time, a data basis is provided for follow-up decision making, decision making deviation caused by data missing is avoided, and the accuracy of decision making is improved. Different types and different dimensions of data are converted into structured heterogeneous tensors through a multi-factor coupling dynamic self-learning decision model under a variable working condition, original characteristics and association relationships of the data are reserved, the heterogeneous tensors are mapped to a unified feature representation space through embedding processing, the potential value of the heterogeneous data is mined to the maximum extent, and the heterogeneous tensors are mapped to the unified feature representation space through the multi-factor coupling dynamic self-learning decision model. Decision precision loss caused by rough data processing is avoided, the reliability of intelligent decision making is improved, the self-learning decision-making agent autonomously generates action signals based on the fused feature data, manual judgment is replaced, human errors are reduced, and dependence on manual intervention is remarkably reduced.

The utility model is applicable to the technical field of interactive tablets, and provides an embedded interactive tablet system and an interactive tablet, the embedded interactive tablet system comprises a control module, a switching module, a signal conversion module, a port module, a switching module and an embedded processing module, the control module is electrically connected with the switching module and the switching module, and the signal conversion module is electrically connected with the embedded processing module. The signal conversion module is electrically connected with the switching module and the port module, and the switching module is electrically connected with the switching module and the embedded processing module. According to the embedded box, the switching module is additionally arranged, so that the first signal output by the signal conversion module can be transmitted to the control module in the mainboard and can also be transmitted to the embedded processing module in the embedded box through the switching module, and selection and switching of signal transmission paths are achieved. Therefore, only the port module needs to be arranged on the mainboard, so that the operation complexity of a user is reduced, and the user experience is improved.

The invention relates to the technical field of power grid operation, and provides a power grid operation safety distance accurate sensing device and method in a non-fixed scene, and the device comprises a hardware device and terminal software which is in communication connection with the hardware device. The hardware device is integrated and deployed on a movable support or a holder, and comprises a laser radar sensor which is used for rapidly scanning a power grid operation scene and obtaining three-dimensional point cloud data including three-dimensional coordinates, laser intensity and echo information; the high-definition visible light camera is approximately parallel to the optical axis of the laser radar sensor and is used for collecting real-time video images of the operation scene; an AI visual analysis module and a point cloud processing unit are arranged in the embedded processor; and the terminal software is deployed on the edge computing equipment or the background server and is used for receiving and processing the data from the hardware device. According to the invention, the safe distance can be accurately sensed.

The invention discloses a sea-ground-air-space integrated marine intelligent collaborative three-dimensional networking observation method, which belongs to the technical field of marine environment observation, is used for intelligent collaborative three-dimensional networking observation, and comprises the following steps of: generating an optimal navigation track covering a target area of dynamic characteristics by using a multi-target optimization algorithm; each observation platform executes a measurement task according to the received collaborative observation instruction, and each observation platform carries an embedded processor to analyze the collected temperature, salinity and flow field data characteristics in real time; and the data processing center completes data assimilation and ocean mode simulation by using supercomputing parallel calculation, and generates and outputs high-resolution three-dimensional temperature, salinity and flow field analysis results of dynamic characteristics. The coverage rate and the quality of observation information are obviously improved; the observation precision and efficiency are obviously improved, and optimal configuration and high-precision observation of resources are realized; a finer three-dimensional structure is captured with less navigation mileage, data redundancy and energy consumption are reduced, and observation efficiency is improved.

The approaches presented here enable the generation of alternative views from disparity data acquired for one or more objects in a scene. This generation can be performed using an embedded processor with DMA memory access or other hardware with limited capacity. An intermediate representation can be created, depicting a 2D histogram view of the disparity data. This intermediate representation can then be transformed, using the embedded processor, into an alternative view image, such as a bird's-eye view. Morphological or similar filtering can be applied to the one or more objects in the intermediate representation using a filter of equal size, regardless of their distance from the camera plane used to acquire the disparity data.

The invention relates to the field of underground detection, and discloses a super-large-scale water-containing underground goaf drilling and loading unmanned ship group detection system and method, and the system comprises a drill bit storage module, an unmanned ship group, an underwater communication module and a central data processing module. The drill bit storage module is used for realizing deployment and recovery of the unmanned ship group; underground goaf data collected by the unmanned ship group is sent to the central data processing module through the underwater communication module to be processed, and an underground goaf holographic three-dimensional model is obtained. The method has the beneficial effects that the detection efficiency is greatly improved, and efficient detection of the super-large-scale water-containing goaf can be completed with less manual intervention. The mode of cooperative operation of multiple unmanned ships significantly expands the coverage range of single detection, and the detection efficiency is improved by several times compared with that of a single robot. And meanwhile, the embedded processing module of the drilling and loading platform supports real-time data analysis and modeling, so that the post-processing time is further shortened.

This invention discloses an intelligent management method for digital creative marketing scenarios based on multimodal feature fusion, relating to the field of multimodal feature fusion technology. The method includes the following steps: processing raw data to obtain short-term and long-term interaction sequences; performing fixed-length processing on the short-term interaction sequences to obtain fixed-length short-term interaction sequences; performing fusion processing, first embedding processing, and second embedding processing on the multimodal feature set of creative content to obtain creative content embedding matrices, location embedding matrices, and time embedding matrices; processing the fixed-length short-term and long-term interaction sequences to obtain user short-term and user long-term interest vectors; adaptively fusing the user short-term and user long-term interest vectors to obtain a user dynamic interest representation; calculating a balanced interest score matrix based on the user dynamic interest representation; and performing propagation processing through an improved graph neural network model to generate a personalized creative content recommendation list.