150results about How to "Low hardware requirements" patented technology

The invention provides an agricultural field large language model training method and device and a medium, and the method comprises the steps: collecting agricultural field data book information, and carrying out the preprocessing of the information, and obtaining a first Chinese data set and a first minority language data set, collecting the first Chinese data set and the first minority language data set, generating a first mixed language data set, processing the first mixed language data set by adopting a word segmentation library, performing secondary processing by adopting a preset special character processing method, constructing a word segmentation list, and constructing a WordPiece word piece model to process and divide the word segmentation list; converting each word in the word segmentation list into a word vector by adopting a word embedding algorithm; the method comprises the following steps of: constructing a coding component and a decoder component which are based on a Transform architecture and adopt a self-attention mechanism, generating an agricultural field large language model, and training the agricultural field large language model by taking a word vector as an input to obtain a trained agricultural field large language model. The problem of supporting a large language model of Chinese and minority languages is solved.

The invention relates to the technical field of artificial intelligence digital persons, and particularly discloses an interactive digital person generation method and system based on artificial intelligence, and the method comprises the steps: collecting behavior data, voice data and image data of a real person; constructing a model for generating a digital human according to the collected data; a deep learning algorithm is adopted to analyze the model, initial digital human behavior characteristics, voice characteristics and appearance characteristics are generated, the initial digital human behavior characteristics, voice characteristics and appearance characteristics are input into a generative adversarial network (GAN), the behavior logic, voice logic and appearance logic of the digital human are optimized through the iterative training and optimization process of a generator and a discriminator, and the digital human behavior logic, voice logic and appearance logic are obtained. And finally, the interactive digital human with high authenticity is generated, the high authenticity of the digital human in behavior, appearance and sound is ensured, a personalized digital human image conforming to the expectation of the user is created, the user experience is further improved, the digital human image with high authenticity is generated, and the naturalness and interactivity of communication are enhanced.

The invention discloses a live broadcast method and device, equipment and a storage medium, and belongs to the field of human-computer interaction. The method comprises the following steps: receiving an uploading operation of a material video, wherein the material video is obtained by pre-recording; receiving a configuration operation of the live broadcast commodity, wherein the configuration operation is used for configuring commodity information of the live broadcast commodity; and playing a live broadcast video of the first live broadcast room, the live broadcast video comprising a video picture of the material video and an explanation audio, the explanation audio being a human voice audio synthesized based on an explanation text, and the explanation text being generated according to the commodity information and / or the material video. The method can improve the digital human live broadcast efficiency.

The invention provides a method, device and equipment for sensing the environment around a vehicle and a storage medium. The method for sensing the environment around the vehicle comprises the steps that image information around the vehicle is acquired; target feature recognition is carried out on the image information, and point cloud features of all target objects are generated; determining three-dimensional position information of each target feature in a vehicle coordinate system according to the point cloud feature of each target object; and according to the three-dimensional position information of each target feature under the vehicle coordinate system, converting to a top view plane of the vehicle body, and generating environment information around the vehicle at a bird's-eye view angle. The method for sensing the environment around the vehicle is easy to implement, the requirement for the sensor and the hardware requirement for algorithm implementation are low, high-cost schemes such as a high-precision map and cloud computing are not needed, the price is low, and the coverage degree is high.

The invention discloses an unmanned aerial vehicle self-adaptive inspection method based on multi-layer characterization learning, which comprises two stages of automatic coarse inspection and self-adaptive fine inspection, and learns the space structure, component characteristics and disease characteristics of a bridge layer by layer. In the automatic rough detection stage, in combination with a spatial clustering post-processing method and a point cloud semantic segmentation network, instance attributes of bridge components are quickly learned. Then, the point cloud space is compressed and converted by using a simulation view field model and a dimension reduction technology, and the unmanned aerial vehicle is guided to carry out efficient and comprehensive space patrol in plane geometry. In the adaptive fine detection stage, a space-channel two-dimensional optimized hybrid convolution structure Light-PVIT is utilized to learn and extract disease features in the automatic coarse detection stage as prior information. The information guides the unmanned aerial vehicle to carry out fine detection on a disease area in a small field of view. According to the unmanned aerial vehicle self-adaptive inspection method, the bridge inspection efficiency and precision are remarkably improved, and reliable technical support is provided for bridge maintenance work.

The invention discloses a method for identifying embryo ends of brown rice grains, a method for separating the brown rice grains from whole brown rice grains and a separation device, and relates to the field of brown rice grain whole embryo extraction. The invention aims to solve the problem that the germ concentration of brown rice grains is low due to the fact that tip sections of two rice grains cannot be separated in an existing germ-remaining rice processing mode. According to the recognition method, on the basis of the image of the brown rice grains transmitted by the light source, the end with the embryo is recognized through differences of the gray levels of the two ends of the image, the brightness and / or geometric morphology of the HSV space and the like, and then separation of the end is achieved. According to the device, when brown rice grains pass through a light-transmitting piece in the sliding process of a rice falling plate, an image recognition system takes a picture and recognizes that brown rice grain germs are located at the left end or the right end, the positions of the germs are determined according to the image recognition system, and a pneumatic unit is controlled to blow / suck the brown rice grains with germ tips into inner grooves in the two ends of a rice falling groove in the length direction; the cutter assembly located on the rice falling plate cuts off the brown rice, and the germ tip section and the non-germ section respectively fall off and are collected in the separation opening of the rice supporting plate.

The invention discloses a stratified fuel cell temperature control method based on adaptive model-free predictive control. The method comprises the steps that firstly, an upper-layer decision maker determines the optimal stack temperature under different working conditions based on a polarization curve of a stack, and the optimal stack temperature serves as the optimal reference temperature to be output to a lower-layer controller; in the lower layer controller, based on real-time input and output data of battery operation, nonlinear prediction of the stack temperature is realized by introducing an auxiliary slack variable. And then an optimal control variable is obtained by solving an optimization problem with constraints in each execution period, and a water pump and a radiator are driven to complete optimal control of the temperature of the galvanic pile. Compared with a traditional control method, the method has the advantages that the dependency on the model is reduced; compared with a traditional model prediction control method, the method has the advantages of being low in calculation complexity, low in requirement for controller hardware and the like, and the problems that the proton exchange membrane fuel cell is high in nonlinearity and coupling performance and large in control difficulty are solved to a certain extent.

The present invention discloses a PROFINET distributed IO synchronization method, comprising configuring a scan lock timer T for an interface module and each IO module respectively. tim , all modules T tim Have the same target reload cycle time and target synchronization lock offset time; each time the synchronization pulse edge arrives, according to each module T tim The deviation between the actual synchronization lock offset time and the target synchronization lock offset time dynamically adjusts the T of each module tim The reload cycle count value is used to realize each module T tim The synchronization between modules is tim The synchronization period is calculated and determined in the overload event and the synchronization pulse edge event, and the synchronization of input and output is realized. The present invention is easy to implement with FPGA or ARM, MCU peripherals, and can effectively improve the synchronization accuracy between the PROFINET communication interface module and the IO modules in each slot, and realize the high-precision synchronization of the input and output of the IO modules in each slot.

According to the map display method, the mobile device, the electric two-wheeled vehicle and the storage medium, under the condition that an image processor in a mobile terminal stops drawing, a map image is drawn on the mobile terminal through a central processing unit in the mobile terminal according to the position information of the mobile terminal, and the problem that a target application is switched to the background to run is solved. Or under the condition that the mobile terminal is in a screen-off or screen-locked state, the image processor cannot draw the map image. And then, the map video stream is generated according to the map image drawn by the central processing unit through the target application, and the map video stream is sent to the electronic equipment, so that the electronic equipment can display the map in real time, and the electronic equipment only needs to display the corresponding map video according to the received map video stream, so that the hardware requirement on the electronic equipment is low, and the user experience is improved. And the electronic equipment can display the map in a low-hardware environment.

The invention discloses an automatic detection device for an ultrasonic phased array of a spinning cylinder. The automatic detection device comprises a workpiece rotating table, a ground rail truss system, a scanning mechanical arm, a probe module, a coupling system, an air drying module, a defect calibration system, a data acquisition system and the like. According to the device, full-coverage scanning detection of the barrel is achieved through the workpiece rotating table and the ground rail truss system, automatic scanning and thickness measurement of the barrel are achieved through the mechanical arm and the probe module, and coupling of the probe and the barrel and drying of the barrel during detection are achieved through the coupling system and the air drying module. A defect marking device is designed to calibrate detected defects in time, a data acquisition system processes defect signals and marks the defects, and full-automatic detection of the spinning cylinder is achieved.

A dynamic cable dynamic response monitoring method based on deep learning comprises the following steps: collecting six-degree-of-freedom motion time history data of a floating platform and three-degree-of-freedom motion time history data of a buoy section through a monitoring system, and forming a training data set for different fatigue hot spots in combination with dynamic response data of each monitoring node along the length of a cable; then, a multi-task integration model comprising a plurality of parallel regression modules is constructed, and each module focuses on predicting the tension or bending moment of a specific monitoring node; and training each regression module by using training data sets aiming at different fatigue hotspots, and integrating the regression modules to form a complete multi-task integration model. And finally, inputting motion time history data collected in real time into the model so as to predict tension and bending moment of each monitoring node and realize dynamic response monitoring of the dynamic cable. While the health monitoring of the dynamic response in the dynamic cable operation and maintenance stage is realized, the monitoring cost is reduced, and the economic benefit of the floating wind power plant is improved.

The invention provides a deep-sea shadow area target positioning estimation method suitable for underwater maneuvering platform observation, and belongs to the technical field of deep-sea positioning measurement, and the method comprises the steps: firstly, receiving a narrow-band signal of a near-sea surface sound source of a deep-sea shadow area through an M-element horizontal array, and obtaining an initial target direction of arrival estimation value through conventional beam forming; and then clockwise rotating the horizontal array for T times, recording a course angle difference value of each rotation, receiving a signal, and performing beam forming on the rotated received signal to obtain a target direction of arrival estimated value. And constructing a cost function and searching an optimal solution based on the estimated value of the direction of arrival and the difference value of the course angle before and after rotation to obtain the estimated values of the horizontal azimuth angle and the vertical pitch angle. And finally, substituting the vertical pitch angle estimation value into a sound source distance estimation formula to calculate and obtain a sound source distance estimation value, thereby realizing three-dimensional positioning. The problem that in the prior art, sound source orientation and distance estimation of an underwater mobile platform in a deep sea environment needs a large-aperture horizontal array or large-broadband information, and practical application is difficult is solved.

The invention discloses an analog signal measurement method and device based on relative frequency control, and belongs to the technical field of measurement. The method comprises the following steps: setting a frequency F0 of a signal to be measured and a period T0, determining a frequency difference delta f, and generating a sampling frequency F1; generating a sampling clock for measurement according to the sampling frequency F1 to obtain a voltage group V1 (k); by adjusting the delta f, different new sampling frequencies F1 are generated and measured, and voltage groups {V2 (k)} to {Vn (k)} are obtained respectively; and according to the data set {V1 (k), V2 (k),..., Vn (k)}, recovering a measurement waveform and calculating a related measurement value. According to the invention, equipment occupation, test difficulty and test cost can be reduced, and a stable and accurate analog signal parameter measurement result can be rapidly obtained.

The invention relates to the technical field of photovoltaics, and discloses a global maximum power point searching method and device, electronic equipment and a storage medium, which are used for reducing hardware requirements for searching a global maximum power point. The method comprises the following steps: acquiring operation data of an inverter, wherein the operation data comprises the voltage of each path of MPPT and the current of each photovoltaic string in each path of MPPT; determining a target MPPT based on the voltage of each path of MPPT and the current of each photovoltaic string in each path of MPPT, wherein the target MPPT is an MPPT falling into a local maximum power point; judging whether the inverter has a global scanning function or not; if the inverter does not have the global scanning function, performing multiple times of short circuit and recovery on part of photovoltaic modules of the target MPPT through a plurality of preset bypass switches, and determining the MPPT power after each recovery to obtain a plurality of candidate MPPT powers; and according to the maximum value of the MPPT power in the plurality of candidate MPPT power, carrying out short circuit on a part of photovoltaic modules in the target MPPT, so that the inverter finds a global maximum power point within a preset range of the maximum value of the MPPT power.

The invention discloses a forklift collision early warning system adopting three-dimensional point cloud information to correct two-dimensional image coordinates, and belongs to the field of industrial safety monitoring. Comprising a video data acquisition module, a three-dimensional point cloud data acquisition module, an image target detection module, a three-dimensional point cloud target detection module, a correction model training module, a coordinate correction module, a target tracking module, a target distance calculation module and an alarm module. The invention further discloses a forklift collision early warning method adopting the three-dimensional point cloud information to correct the two-dimensional image coordinates. According to the method, the image target coordinates are corrected by collecting and detecting the image target and the three-dimensional point cloud target and mapping the model between the image target coordinates and the three-dimensional point cloud target coordinates, so that the influence of perspective distortion of the two-dimensional image is effectively reduced, and the accuracy of the forklift collision early warning system is improved while the real-time performance is ensured.

The invention discloses a pavement flatness detection method based on a vibration sensor. The method comprises the following steps: acquiring collected vibration sensor data; the vibration sensor data comprises acceleration perpendicular to the ground direction; determining relative variation and direction according to the acceleration and the reference acceleration; judging whether the data of the vibration sensor is a peak value or not according to a preset upper threshold value, a preset lower threshold value, a preset relative variable quantity and a preset direction, and combining adjacent peak values for reporting so as to obtain a complete event; and determining the pavement flatness according to the complete event, and obtaining a flatness detection result. According to the method, the peak values in the vibration signals are detected in a mode of combining multiple groups of threshold values and statistical values, the adjacent peak values are combined into one event, and finally, the pavement flatness is judged by analyzing the event, so that repeated reporting of a single disease can be avoided, the accuracy of a detection result is improved, the requirement on hardware is low, the cost is low, and the method is suitable for popularization and application. And the method has a certain self-adaptive capability to road condition changes.

The invention belongs to the field of computer vision, and discloses a multi-target detection method, a multi-target detection model training method, an edge computing device and a medium, and the multi-target detection model training method comprises the steps: extracting a first-class feature map; respectively inputting the first type of feature map into a first target detection module, a multi-layer feature fusion module and a second target detection module; calculating a target loss function value based on the obtained target detection result and the target labeling result; and adjusting network parameters of the to-be-trained model until a preset model training convergence condition is met, and obtaining a multi-target detection model. According to the method, the generalization ability of the model is improved, and the driving safety is enhanced.

The invention discloses a nested non-uniform dial plate reading method based on machine vision, which comprises the following steps of: 1, training an instrument detection model, 2, inputting a picture to be detected into the instrument detection model to obtain a set formed by a plurality of single dial plate images, 3, preprocessing the dial plate images and establishing a rectangular coordinate system, 4, obtaining the circle center coordinate and radius of the single dial image based on an improved Hough transform algorithm, 5, obtaining a pointer fitting linear equation of the single dial based on a regression method, 6, obtaining ordered scale information of the single dial image based on a real scale extraction algorithm, and 7, obtaining a single dial reading by using a distance method based on the scale and pointer information. And 8, performing linear operation on each dial plate according to known information to obtain the final reading of the picture to be tested. According to the invention, while the meter reading precision is ensured, the reading efficiency is improved, and the required hardware condition is reduced, so that the method can be widely applied to the working scene of real-time reading by relying on light-weight equipment.

The application discloses a visual perception method for assisted driving, comprising the following steps: S100, collecting assisted driving image data, and constructing input X after data set preprocessing; S200, constructing a multi-task neural network model F and substituting the input X into the model to obtain a plurality of branch task-specific outputs f i ; S300, calculating the final loss function L all of the multi-task neural network model F; S400, optimizing the minimum as the target of the loss function L all , and finally obtaining the optimal multi-task neural network model F * . A system using the above visual perception method for assisted driving is also disclosed, specifically comprising a data input module, a shared layer feature extraction module, a task layer structure search module and a model output module. The application has the beneficial effect that a multi-task model is constructed to effectively reduce the network parameter quantity and the calculation quantity, simultaneously reduce the online reasoning time, and is very beneficial to practical application.

The present application relates to a digital twin three-dimensional model rendering method and system, and relates to the field of digital twins, wherein the method includes: receiving a rendering request from a front-end device, the rendering request including hardware parameters of the front-end device; judging whether the rendering performance of the front-end device meets preset requirements based on the hardware parameters; in response to the rendering performance failing to meet the preset requirements, detecting the resolution of the front-end device, and obtaining a target image matching the resolution in a pre-rendered image; and sending the target image to the front-end device.

The invention discloses a method for detecting surface defects of a steel wire rope on a production line based on machine vision. The method comprises the following steps: constructing a defect detection model comprising a frequency domain screening network, an enhancement network, a feature extraction network and a classification network; acquiring a steel wire rope image and inputting the image into a frequency domain screening network to obtain screening frequency domain features; inputting the screened frequency domain features and the steel wire rope image into an enhanced network to obtain enhanced features; inputting the enhanced features into a feature extraction network to obtain extracted features; inputting the extracted features into a classification network to obtain a defect prediction result; training the defect detection model based on the defect prediction result and the defect real label joint loss function to obtain a trained defect detection model; and inputting a to-be-detected steel wire rope image into the trained defect detection model to obtain a corresponding defect prediction result. The target features of the steel wire rope image can be accurately extracted, so that the defect detection accuracy is effectively improved.

The application provides a display device, which comprises a liquid crystal prism panel, the liquid crystal prism panel comprises a plurality of electrode groups, the electrode groups comprise a plurality of electrodes arranged in sequence and at intervals; the absolute values of voltages on at least two electrodes in the electrode groups comprise a first voltage segment, and the proportion of the time length of the absolute values of voltages on the at least two electrodes in the first voltage segment in the time length of the first time segment is different. The display device provided by the embodiment of the application reduces the complexity of the driving circuit of the liquid crystal prism panel and reduces the hardware requirements for the driving circuit of the liquid crystal prism panel.

This invention relates to the technical field of seismic data interpretation, and discloses a method and computer equipment for identifying the locations of beaded reflections in carbonate rocks. The identification method includes: S1, obtaining a two-dimensional seismic migration profile; S2, extracting typical beaded reflections, determining their dimensions, statistically analyzing their structural morphology, and constructing adaptive hit-and-miss structural element pairs; S3, using a hit-and-miss transformation to identify locations that match the adaptive hit-and-miss structural element pairs. This invention's identification method extracts multiple typical beaded reflections, constructs adaptive hit-and-miss structural element pairs, and only identifies locations that match these pairs. Based on a hit-and-miss transformation algorithm, this algorithm has low hardware requirements, high computational efficiency, and high accuracy.

The invention discloses a spine segment space key point positioning method, device and equipment and a medium, and relates to the technical field of image data processing, and the method comprises the steps: inputting a spine original CT image into a spine segment space region positioning network, so as to obtain the position information of each spine segment space region, cutting the original CT image of the spine based on the position information of the spatial region of each spine segment to obtain a local CT image of each spine segment; and inputting the local CT image corresponding to the inpatient area spine segment into the spine segment anatomical key point positioning network to obtain position information of the inpatient area spine segment anatomical key points. The method is high in precision, strong in robustness and low in calculation complexity, and accurate positioning of the spine segment anatomical key points in the CT image of any view field can be realized.

The present invention provides a near-field speckle scanning phase contrast imaging method and system based on DIC, including an image acquisition step: using a speckle scanning phase contrast imaging experimental system to acquire a speckle image and a speckle sample image; the speckle scanning phase contrast imaging experimental system includes a ray source, a speckle generator, a sample, a turntable and a detector; an image processing step: using a unified modulation mode analysis to process the speckle image and the speckle sample image, and outputting the processed image. Under the premise of ensuring that the sample structure information can be effectively extracted, the present invention uses a post-processing step combined with DIC to replace the step of the sample constantly entering and exiting the optical path during the experiment, thereby reducing the requirements of the experimental system on hardware facilities, shortening the experimental time, and enhancing the feasibility of the experiment.

The invention discloses a laser scanning system and a laser scanning method, a control unit receives a processing file and a galvanometer scanning mode to calculate a scanning action used for driving a galvanometer driving unit, and the scanning action of a galvanometer is implemented to completely cover a first scanning path of the processing file; the control unit adjusts and calculates a second scanning path according to the processing file and the first scanning path, and the second scanning path divides the processing file into at least one scanning path in an equal interval range; and finally, the control unit outputs a galvanometer scanning command to the galvanometer driving unit according to the second scanning path, and outputs a switching command to the laser source for executing laser processing.

The present application relates to a method, device, computer equipment and storage medium for evaluating a driver's concentration, wherein the method comprises: detecting each frame of the driver's head and shoulder image, obtaining the driver's head posture angle data, and generating a head posture angle feature map corresponding to each head posture angle data; obtaining the first posture angle data of the current vehicle corresponding to each frame of the head and shoulder image, and generating a vehicle posture angle feature map corresponding to each first posture angle data; extracting the driver's first binocular image from each frame of the head and shoulder image, and generating a gaze point feature map corresponding to each first binocular image; performing multimodal feature fusion on the head posture angle feature map, the vehicle posture angle feature map and the gaze point feature map, and evaluating the driver's concentration based on the fusion result to obtain the driver's concentration level. Through the present application, the problem of being unable to accurately and comprehensively evaluate the driver's concentration is solved, and an accurate and comprehensive evaluation of the driver's concentration is achieved.

The embodiment of the application relates to the radar technical field, and discloses a signal processing method, an electronic device and a storage medium. The signal processing method comprises the following steps: transmitting a detection signal in a Doppler multiplexing (DDM) mode through a plurality of transmission channels; processing a return signal to obtain distance-Doppler two-dimensional data; calculating a likelihood probability of a Doppler spectrum under different aliasing combination conditions of transmission channel orders of at least two targets according to the distance-Doppler two-dimensional data; and determining Doppler spectra of the targets according to an aliasing combination corresponding to a maximum value of the likelihood probability. At least, it is beneficial to correctly restore signals before Doppler spectrum aliasing in the presence of at least two targets.

The application discloses a three-level coordinated production line body intelligent control method and system, and belongs to the technical field of intelligent manufacturing. The method comprises the following steps: establishing a global task planning layer, performing semantic analysis and logical disintegration at a first preset frequency, and generating an atomic action sequence; establishing a local space perception layer, receiving the atomic action sequence at a second preset frequency, constructing a dynamic local map by using sensor data, determining a workpiece pose by point cloud registration, and mapping the atomic action sequence into a continuous motion trajectory flow; establishing a real-time action execution layer, receiving an instantaneous target pose at a third preset frequency, calculating a theoretical driving torque by combining a Lagrange dynamics equation, switching to an impedance control mode at a contact moment, and driving an execution mechanism to complete work; and establishing an asynchronous coordination layer, realizing three-layer asynchronous communication through an industrial network, and controlling the three-layer frequencies to be distributed in a ladder type in an increasing manner. The application has the effects of improving the flexibility, precision and safety of a production line.

The present invention discloses a method and related apparatus for interaction between vehicle-mounted driver and passenger assistance devices. The method for interaction between vehicle-mounted driver and passenger assistance devices comprises: the vehicle-mounted master control device transmits the personalized interface corresponding to the at least one vehicle-mounted slave control device to each vehicle-mounted slave control device respectively through a wireless local area network connection pre-established with the vehicle-mounted slave control device; in response to an operation request for the personalized interface forwarded by the vehicle-mounted slave control device through the wireless local area network, the vehicle-mounted master control device processes the operation request and generates a processing result interface, and returns the processing result interface to the corresponding vehicle-mounted slave control device through the wireless local area network. The present invention reduces the cost of vehicle-mounted driver and passenger assistance devices in a car. When the driver and passengers are riding in a car, they can share the application of the vehicle-mounted master control device through the vehicle-mounted slave control device, and can interact with the vehicle-mounted master control device to enjoy the application service, and the user experience is good.