105 results about "Complex field" patented technology

The invention provides an electromagnetic field prediction method and device and electronic equipment, and relates to the technical field of electromagnetic field solving. The method comprises the following steps: acquiring historical electromagnetic original data in a field-line coupling scene, preprocessing the historical electromagnetic original data, inputting the preprocessed historical electromagnetic original data into an LSTM-PINN model, and outputting a physical field quantity mapping result; wherein the physical field quantity comprises an electric field component and a magnetic field component; setting a weighted loss function, and performing optimization training on the LSTM-PINN model based on a physical field quantity mapping result and an error of a real physical field quantity corresponding to historical electromagnetic original data to obtain a trained electromagnetic field prediction model; wherein the weighted loss function comprises a data loss function, a physical residual loss function, an initial condition loss function and a boundary condition loss function; and inputting real-time electromagnetic original data into the trained electromagnetic field prediction model to obtain a spatio-temporal distribution electromagnetic field prediction result. The method can effectively extract the spatial distribution features and the time sequence features at the same time, and is suitable for a complex field-line coupling problem.

The invention discloses a multi-platform real-time field rice ear detection and counting method based on a visual converter, which is suitable for various acquisition platforms from a portable mobile terminal to an unmanned aerial vehicle terminal and the like, and meets the requirements of small-range accurate identification and large-range high-throughput monitoring. Aiming at the characteristics of multiple scales, multiple varieties, phenotype difference, time sequence change, complex field environment and the like in a data set, a rice spike detection counting network comprising a feature enhancement module, an efficient cross-space feature extraction network, a lightweight bidirectional feature fusion module and a number estimation and detection head module is constructed. The average precision mean value (mAP50) of the model reaches 0.8906 and is improved by 11.5% compared with that of an original model, the parameter quantity is reduced by 29%, and the calculated quantity (GFLOPs) is reduced by 48%. The method can effectively process rice ear images of different scales and growth periods, adapts to morphological change, overcomes interference such as illumination, phenotypic difference and shielding, and provides technical support for rice yield evaluation and field rice ear quality analysis.

The invention discloses a multi-agent cooperative task method and device based on a hierarchical task mechanism, and relates to the technical field of artificial intelligence, and the method comprises the steps: carrying out the hierarchical abstraction based on a task space of a target domain, obtaining a plurality of hierarchies, configuring a corresponding agent and a tool for each hierarchies, constructing a state management mechanism to manage intermediate data among a plurality of hierarchies; in response to target information input by a user, decomposing and planning a target task by using a plurality of mutually associated hierarchies to obtain a directed acyclic graph of the target task; calling and executing the subtasks corresponding to the nodes from bottom to top based on the directed acyclic graph so as to complete the target task; and integrating a completion result of the target task and then rendering the result to a user interface. The problem that an existing intelligent agent system is difficult to meet the requirements of complex fields with highly structured and hierarchical features is solved. A more efficient, accurate and robust scheme can be provided for complex tasks with highly structured and hierarchical features.

The invention discloses a YOLOv11n-based modeling method for an insulator defect detection model of an unmanned aerial vehicle polling high-voltage transmission line, and the method comprises the following steps: 1, collecting an insulator defect image through an unmanned aerial vehicle, and carrying out the preprocessing of the insulator defect image, and forming a data set; step 2, constructing an insulator defect detection model based on the improved YOLOv11n; step 3, training the insulator defect detection model based on the improved YOLOv11n constructed in the step 2 to obtain an optimal model; and 4, performing precision verification on the optimal model constructed in the step 3 by adopting a verification set to obtain an insulator defect detection model meeting a precision requirement. According to the method, by deeply optimizing a model structure and introducing an innovative module, high-precision identification of insulator defects is realized on the premise of ensuring low computing resource consumption. The method can effectively adapt to efficient inspection tasks in various complex field environments, and the detection performance and the practical value at the edge equipment end are remarkably improved.

The invention belongs to the technical field of agricultural machinery and intelligent control, and particularly relates to an intelligent precise weeding robot based on multispectral recognition. The robot aims at solving the problems that in the prior art, due to visible light morphological recognition, the forms of crops and weeds are similar, and the recognition precision is low and the robustness is poor in the complex field environment. The robot comprises an autonomous mobile platform and a sensing operation assembly, the sensing operation assembly integrates a multi-mode sensing unit, a data fusion and decision core and a high-energy thermal ablation module, and the multi-mode sensing unit provides eight narrow-band pulse light with specific wavelengths through an active light source illumination module. Realizing synchronous acquisition of spectrum and space information by combining multispectral imaging and three-dimensional profile measurement; the data fusion and decision core performs registration, segmentation and classification on multi-source data to generate weed target coordinates; and the high-energy thermal ablation module implements 1470-nanometer laser accurate ablation based on the coordinates, and verifies the operation effect in real time through coaxial infrared thermal imaging. By means of the technical scheme, high-confidence-coefficient distinguishing and millimeter-level lossless precise weeding of crops and weeds in the complex farmland environment can be achieved.

The invention discloses a method and a system for improving the operation precision of a rotary tillage seeder. The method comprises the following steps: acquiring working environment and machine state information in real time through a soil hardness sensor, a humidity sensor, a gyroscope and a GNSS (Global Navigation Satellite System) module; the rotary tillage depth and the sowing depth are detected, and operation deviation is calculated based on a multi-source data fusion algorithm; according to the deviation result, the rotating speed of a rotary tillage cutter shaft, the downward pressure of the seeder and the posture of the machine are dynamically adjusted through a self-adaptive control strategy, and the seeding depth and the rotary tillage depth are consistent; the operation track is optimized in combination with GNSS navigation, and the row spacing and plant spacing precision is ensured; and after the operation is finished, evaluating the uniformity and consistency of the operation. The system comprises a data acquisition module, a data fusion module, a control decision module, an execution mechanism and an operation evaluation module. According to the method and the system, the operation precision and the stability of the rotary tillage seeder can be improved in a complex field environment.

The invention discloses a tilling depth detection method and system based on three-dimensional data, and relates to the technical field of precision agriculture, and the method comprises the steps: building nonlinear mapping based on a corrected attitude angle and a lifting arm angle, updating the nonlinear mapping through extended Kalman filtering, obtaining a high-precision real-time attitude angle, and obtaining a tilling depth detection result; and constructing an initial tilling depth model based on the high-precision real-time attitude angle, and fusing the target tilling depth and the LiDAR depth to obtain a target tilling depth value. By fusing data of multiple sensors such as LiDAR, RGB-D and IMU, adopting complementary filtering and an RANSAC algorithm to correct an attitude angle and improve attitude estimation precision, a point cloud density and signal-to-noise ratio adaptive weight mechanism is introduced, fusion of target tilling depth and visual depth is optimized, robustness and accuracy of tilling depth detection under a complex field condition are improved, and the method is suitable for large-scale popularization and application. The problem that a traditional method is insufficient in adaptability to mechanical dynamic changes and terrain slopes is effectively solved.

The invention discloses an acquisition device based on rock mass structural surface roughness, and relates to the technical field of rock mass structural surface roughness measurement, the acquisition device comprises a frame type base, a probe assembly, a laser scanning module and a real-time data processing unit, the lower surface of the frame type base is symmetrically provided with liftable supports, and the probe assembly is arranged on the lower surface of the frame type base; the frame-type base is of a square frame-type structure, the bottom of the frame-type base is open, the frame-type base is provided with a height-adjustable supporting column, the center of the top face of the frame-type base is provided with a linear sliding rail, the sliding rail is provided with a slidable measuring table top, and the measuring table top is driven by a motor to move. Vibration interference caused by irregularity of a device or a rock surface is suppressed, and laser scanning and a multi-point synchronous acquisition technology are combined, so that the information space continuity of the structural surface is improved; real-time dynamic measurement in a complex field environment is achieved, efficiency and data authenticity are improved, and the method has popularization value.

The invention discloses a pseudo-ginseng flower maturity detection method based on an improved RT-DETR model, and belongs to the field of agricultural intelligent detection and deep learning application, the pseudo-ginseng flower maturity detection method is an improvement of a deep learning algorithm taking RT-DETR as a baseline model, and mainly performs optimization and enhancement on three aspects of backbone network replacement, neck network optimization and AIFI module partial replacement. On the premise of ensuring the detection speed, the lightweight detection model with lower complexity and higher detection precision is successfully realized. The model provides a core detection method for intelligent picking of panax notoginseng flowers, fills the blank of small-target and easily-confused growth stage detection technologies in a complex field environment, has important engineering application value, and promotes automatic upgrading of the panax notoginseng industry.

The invention discloses a crop insect pest detection method and system based on improved ConvNeXt, and relates to the technical field of crop insect pest detection, and the method comprises the steps: constructing an EDA-ConvNeXt model, training the EDA-ConvNeXt model to obtain a crop insect pest detection model, according to the method, the detection precision is improved through the EDA-ConvNeXt model, the multi-scale features are captured by the backbone network, the boundary information is enhanced by the edge enhancement branch, the dynamic feature extraction branch adapts to the morphological change, and the key features are highlighted by the multi-branch fusion. The classification layer ensures the recognition reliability, and finally higher accuracy is achieved in a complex field scene.

The invention discloses a multi-mode joint information source channel coding system for satellite-ground semantic communication. The multi-mode joint information source channel coding system comprises a joint information source channel coder and a decoder, the encoder is used for sequentially performing feature extraction, feature fusion, data cutting and phase modulation operation on input multi-modal data to generate a transmitting signal, and transmitting the transmitting signal to a ground base station through a channel; and the decoder is used for sequentially performing real and virtual part separation, data expansion, feature extraction and modal data recovery operation on the received signal so as to reconstruct multi-modal data. According to the system, public semantic information among multi-modal data can be mined, the time-varying low-signal-to-noise-ratio channel characteristics of a satellite downlink are combined, the information bottleneck theory highly related to data reconstruction quality is used as guidance to optimize a coding space, complex field transmission signals with the high compression ratio are generated in a self-adaptive mode, and the multi-modal multi-domain transmission signals are obtained. The reconstruction quality of the multi-modal data can be ensured, and the bandwidth and the energy consumption overhead of satellite communication can be remarkably reduced.

The invention discloses a self-leveling anti-seismic field emergency intelligent operating bed, and relates to the technical field of operating beds. Comprising a supporting table and a supporting box fixedly installed at the top of the supporting table, an operating bed body is installed on the top wall of the supporting table, and according to the self-leveling anti-seismic field emergency intelligent operating bed, through arrangement of a leveling assembly, when the operating bed needs to be arranged in the field, firstly, switching of a stable supporting state of a moving state can be rapidly completed; under the self-weight action of the whole device, the pressing seat and the cone are pushed to be inserted into the ground to form primary support, and intelligent self-leveling is conducted on the supporting table through the horizontal control mechanism according to the horizontal state of the whole supporting table, so that the supporting table is rapidly kept in the horizontal supporting state, and the inclination or shaking condition is prevented; and during intelligent leveling, adjustment and adaptation can be carried out according to each fulcrum, especially in the complex field environment, stable leveling can be completed on the uneven ground, and better environmental adaptability is achieved.

The invention discloses a rice ear grain identification method and system based on a deep learning neural network, and the method comprises the steps: firstly carrying out the precise mask marking and systematic preprocessing of a rice ear image, and enhancing the quality of input data; secondly, a self-attention mechanism is introduced into the MaskR-CNN model, the long-distance dependency relationship between any positions in a feature map is effectively captured, the understanding ability of the model for the overall structure of the rice panicles is enhanced, and particularly the recognition effect on the rice panicles which are shielded or incomplete in shape is improved; furthermore, a multi-point regression strategy is fused, regression weights are dynamically distributed according to different IoU thresholds, and regression results of a plurality of high-quality candidate boxes are comprehensively utilized for fine correction, so that the accuracy and robustness of rice spike positioning are remarkably improved. Therefore, the method can effectively cope with interference caused by a complex field environment, reduces false detection and missing detection while ensuring a high recall rate, realizes more accurate and stable automatic statistics of rice ears, and provides reliable data support for subsequent agricultural decision making.

The invention discloses a multi-focus composite Fresnel phase diagram rapid generation and power equalization method and system for one-to-many free space laser communication. According to the method, aiming at a scene in which a plurality of user targets are located at different directional angles and different propagation distances, a focusing + directional single-beam phase is constructed for each target on the basis of analyzing a Fresnel zone lens phase and a linear tilt phase, and the phase is taken after complex field superposition is performed according to a preset power weight, so that the target target is obtained. A composite phase diagram for a phase-type spatial light modulator (SLM) is directly obtained. The power weight is preset according to a link budget or a task demand so as to realize on-demand distribution and equalization of multi-beam output power; and meanwhile, fixed phase bias can be introduced to each beam to suppress inter-beam interference fringes. According to the method, frame-by-frame iterative solution is not needed, the calculation complexity is low, and high-frame-rate and low-delay online updating is easy to realize.

The invention discloses a strawberry growth period identification and detection method based on improved RT-DETR. The method comprises the following steps: constructing a data set based on collected image samples of different growth periods of strawberries; annotating the image samples in the data set to generate annotation files corresponding to the image samples; dividing the image samples in the data set and the annotation files corresponding to the image samples into a training set, a test set and a verification set; carrying out growth period identification detection on the strawberry photo through an improved RT-DETR algorithm module; according to the method, the common problems of difficulty in small target detection and fuzzy stage characteristics in strawberry growth period identification tasks are effectively solved; the method not only enhances the combined modeling capability of shallow details and deep semantic features, but also improves the expression and interaction of cross-scale and remote dependent features, and effectively improves the recognition accuracy of the model for densely shielded and small-sized targets and similar categories in a complex field environment. And a reliable technical support is provided for crop growth monitoring in intelligent agriculture.

The present specification relates to a technique by which a hologram generation device generates a hologram with reduced speckle noise by optimizing a carrier according to a specific distance. A method by which a hologram generation device generates a hologram, according to an embodiment of the present specification, comprises the steps of: setting a carrier by initializing an arbitrary complex field having a random phase and a random amplitude; removing a specific band component from the frequency band of the set carrier so as to transform the set carrier; propagating the transformed carrier for each specific distance, which is a depth plane in which a target object is present; setting a loss function for each specific distance on the basis of a target uniform amplitude at the specific distance and the amplitude of the carrier propagated to the specific distance; and updating the propagated carrier for each specific distance on the basis of the loss functions set for each specific distance.

The invention provides an acoustic-magnetic time difference calculation method based on signal quality self-adaption, and belongs to the technical field of cable fault detection. The method comprises the following steps: acquiring an acoustic signal and an electromagnetic signal of high-voltage cable partial discharge, and detecting the electromagnetic signal to determine a triggering moment; intercepting a sound signal segment and a background noise signal segment by taking the moment as a reference, and calculating a signal-to-noise ratio and a maximum peak-to-peak value after preprocessing; the trigger point is compared with a preset threshold value, and a trigger point positioning scheme is dynamically selected; determining an effective trigger point according to the scheme, and calculating an original acoustic-magnetic time difference; and calculating the current sound signal energy, and correcting the original time difference in combination with the reference energy to obtain a final time difference. According to the method, through an adaptive matching scheme, the accuracy and reliability of time difference calculation are improved, different signal quality scenes are adapted, and the application adaptability of a complex field environment is enhanced.

The present application relates to the technical field of image recognition, and proposes a rice ear grain detection and counting method and system in a complex field environment. The method comprises: acquiring a to-be-tested image of rice ears in an un-picked state in a complex field environment; inputting the to-be-tested image into a trained rice ear grain detection and counting model; the rice ear grain detection and counting model comprises a feature extraction and fusion module, a shielding perception module, a detection frame proposal module and a classification regression module; the feature extraction and fusion module extracts and fuses the ear grain features of the input image, and outputs an ear grain feature map and a fused feature map; the shielding perception module generates a shielding mask map according to the ear grain feature map; the detection frame proposal module generates a target detection proposal frame according to the fused feature map; the classification regression module performs target classification and detection frame regression of the ear grains according to the fused feature map, the shielding mask map and the target detection proposal frame, and outputs a detection frame and a counting result of the rice ear grains.

The invention discloses an electric tower mechanical defect detection method based on RGB and depth image fusion. The method comprises the steps that 1, an image to be detected is acquired, and electric tower component detection is carried out; 2, extracting and projecting RGB image features of each electric tower component; step 3, estimating and projecting depth features of each electric tower component; 4, fusing the multi-modal features and carrying out anomaly detection; and 5, generating a visual thermodynamic diagram of the mechanical defect. According to the invention, through a depth-guided unsupervised anomaly learning strategy, model training can be completed only by depending on normal samples, and large-scale defect labeling data is not needed. According to the method, the adaptability and robustness of the model in a complex field scene are effectively improved, the data acquisition and labeling cost is reduced, and the generalization ability of identifying novel or unknown defects is achieved. In addition, the lightweight efficient architecture can be deployed on edge equipment such as an unmanned aerial vehicle in real time, and an end-to-end efficient and reliable solution is provided for intelligent inspection and maintenance of a power system.

The present application relates to the field of agricultural forestry informatization technology, in particular to a tree sap flow and growth synchronous monitoring method, system and storage medium, through deploying multi-modal sensor nodes to synchronously collect multi-modal data; after preprocessing the original data, constructing a tree ecological and physiological data set; training a neural network model based on the data set, realizing end-to-end mapping from original multi-modal data to sap flow density and net growth, and embedding the model in a digital twin; the digital twin collects the predicted values at each time to construct a historical sequence, and calculates the sap flow density and net growth at future time. The present application organically integrates ecological and physiological mechanisms with deep learning, improves the accuracy, anti-interference ability and automation level of tree physiological monitoring in complex field environment, and improves the defects of environmental noise and physiological signal aliasing in original monitoring data.

The invention discloses a small target detection method based on a complex number dynamic attention mechanism, and the method comprises the steps: taking a complex number matrix obtained through the pulse pressure and moving target detection processing of a target reflection echo received by a radar as the input, and constructing an I / Q dual-channel standardized tensor as a training sample; establishing a background, clutter and target three-classification labeling graph as a label and setting an effective mask; in a training stage, a small target detection and segmentation network based on U-Net is constructed, and a complex field dynamic attention module is embedded behind a complex residual block of an encoder; the complex field dynamic attention module comprises a complex field channel attention module, a multi-scale space attention module and a dynamic fusion gating module which are respectively used for performing channel enhancement and dynamic fusion after space enhancement aiming at complex features; a mixed loss function is adopted during training; in the reasoning stage, a prediction category is output, and a detection result is obtained through a target category mask.

The application relates to the technical field of radio frequency testing, in particular to a radio frequency performance testing method and device and electronic equipment. The method comprises the following steps: obtaining structure information of a radio frequency front end to be tested; constructing a composite excitation signal, which is used for exciting the radio frequency front end; obtaining spatial complex field distribution information of at least one nonlinear response signal generated by the radio frequency front end excited by the composite excitation signal, and combining the structure information of the radio frequency front end to determine a propagation path of the nonlinear response signal from a generation source point in the radio frequency front end to a radiation point; positioning at least one target circuit node in the radio frequency front end according to the propagation path, and simultaneously determining a nonlinear distortion type caused by the target circuit node; and determining and outputting a test result according to the target circuit node and the nonlinear distortion type. The application can position the target circuit node causing the nonlinear distortion in the radio frequency front end, and solves the problem that the prior art cannot determine a specific fault position.

The application discloses a kind of complex field SAR image super-resolution reconstruction method and system.Therein, method includes the following steps: S1, complex field data pre-processing;S2, data set construction is carried out and data augmentation is carried out;S3, depth learning model is constructed, and training area and verification area are divided, and model training is carried out;S4, using high-resolution original image data carries out precision evaluation;S5, according to the depth learning model of optimal scale training, obtains high-resolution SAR product.The scheme is under the improvement hardware processing capacity, obtains the result that only relies on amplitude diagram super-resolution reconstruction is superior, can restore more ground object details, improves the application value of existing image data natural disaster field and the identification ability of military target.

The invention relates to a rice disease detection method based on a YOLO algorithm, in particular to high-precision detection achieved by improving a YOLOv11 model. The method combines a plurality of innovative modules, and aims to improve the detection accuracy of rice diseases (such as bacterial leaf spot disease, brown spot disease and leaf tumor disease). By introducing a CPA-Enhancer image enhancement module, a bidirectional feature pyramid network (BiFPN), a CARAFE content perception feature recombination module and a CBAM double attention mechanism, the small target identification problem of rice leaf scab can be effectively solved, and the challenges of background interference and multi-scale feature fusion in a complex field environment are overcome. According to the method, through multi-stage optimization, the recognition precision and the detection speed of the model on rice diseases are remarkably improved, and compared with a traditional method, the method has obvious advantages in the aspects of small target detection and scab boundary detail recovery. Experimental results show that the model of the invention is superior to the existing YOLO series and other common deep learning models in multiple performance indexes, and is suitable for large-scale field real-time monitoring and intelligent agriculture application.

The application discloses a ginseng flower and seed lightweight field detection method based on RT-DETR improvement, and belongs to the field of agricultural intelligent detection and deep learning application.The application takes RT-DETR as a baseline model, and makes targeted improvements from three aspects of backbone network, backbone-neck feature interaction and neck network downsampling in view of problems such as complex environmental interference, small target feature sparsity and difficulty in edge deployment of the model in field detection of ginseng flower, green seed and red seed.The model lightweight design is realized, the anti-interference ability and detection precision of the model to the complex field environment are improved, and the real-time detection demand of edge equipment in the field is met.The training of the improved model is completed, the model is deployed on a JetsonOrinNX edge computing platform, end-to-end real-time detection of ginseng flower, green seed and red seed in the complex field environment is realized, core visual technology support is provided for intelligent picking and field fine management of ginseng, and the application has important engineering application value for promoting the automatic and intelligent upgrading of the ginseng industry.

This invention proposes a neural network model enhancement and electronic device for small target defect detection in oilfield inspection images under complex field scenarios. It addresses the problems of severe feature degradation, scarcity of real samples, poor spatial localization ability, low anchor box matching degree, and limited computing power at the edge in small target defect detection in oilfield inspection images under complex field scenarios. This invention constructs synthetic and real datasets through dual-source data processing; performs four-fold super-resolution amplification of the input data through residual dense block concatenation and sub-pixel convolution, and introduces a generative adversarial network detail enhancement branch to generate super-resolution reconstructed images; extracts multi-scale basic features through a lightweight backbone; generates coordinate attention weights through global pooling of height and width dimensions to enhance the spatial features of small targets; inputs the enhanced feature map into a defect detection layer embedded with anchor boxes specific to small targets for classification and regression; and optimizes model training using a hard sample mining loss function.

The invention discloses a golden snub monkey lightweight detection method based on an edge perception double fusion network, and relates to the technical field of image detection. The invention provides a lightweight detection method for solving the problems of complex background, serious shielding, difficulty in small target identification, limited edge equipment computing power and the like in field monitoring of golden snub monkeys. The method comprises the steps of firstly collecting images and constructing a data set; secondly, establishing an EADFNet network improved based on YOLOv12n, and introducing an edge sensing Stem module (EAS) at the front end of the network to enhance the boundary and suppress the background; in the feature extraction stage, a partial dynamic weighted residual block (PDW) is used to stabilize feature learning; in the feature fusion stage, a high and low frequency correlation guidance double fusion module (HCF) is used for realizing splicing-free efficient fusion; and a small target detail head (SOD) is introduced into a detection head. According to the method, the parameter quantity and the calculation cost are greatly reduced, and meanwhile, the detection precision and the recall rate of the golden snub monkeys in the complex field environment are remarkably improved.

The invention discloses a simulation test method for an asymmetric support effect of a weakly-cemented and strongly-expansive soft rock roadway, and belongs to the technical field of coal mining. The method aims at solving the problems that in the prior art, it is difficult to truly reproduce roadway asymmetric large deformation under the weak cementing-strong expansion-rich water multi-field coupling effect, and the effectiveness of an asymmetric supporting scheme cannot be rapidly and quantitatively evaluated. According to the technical scheme, real rock mass parameters are obtained through on-site rock core sampling; similar materials are prepared based on the similarity theory, a physical model is constructed, and the surrounding rock deformation environment is reproduced in a high-fidelity mode; designing a plurality of asymmetric support schemes for simulation comparison; a non-contact full-field strain measurement system is adopted to monitor surrounding rock deformation, and an optimal supporting scheme is optimized and applied to field verification. According to the method, complex field problems are solved to a controllable laboratory environment, rapid and accurate optimization of a supporting scheme is achieved, and a scientific and reliable basis is provided for soft rock roadway supporting design.

The invention discloses a remote sensing image parcel extraction method and device based on a diffusion model, and belongs to the technical field of deep learning, remote sensing satellites, image processing and the like, the method can accurately capture boundary information of parcels of fields through the strong feature learning and generation capability of the diffusion model, effectively avoids adhesion between adjacent parcels of fields, and improves the efficiency of remote sensing image parcel extraction. According to the method, independent and accurate extraction of each field is realized, clearness and accuracy of field boundaries can be ensured even in a field scene with similar or complex field features, and the requirements of large-scale and refined cultivated land resource management are met; according to the method, by introducing a cross attention mechanism and other technologies, original image information is directly fused in the denoising process, so that the generated field boundary is purer and finer, a high-quality extraction result can be obtained without depending on complex post-processing operation, and the whole field extraction process is greatly simplified; the method has higher robustness for different illumination conditions, seasonal changes, image noise and the like.

The utility model provides an explosion-proof electrostatic discharge device for the top of an oil storage tank, the device comprises a sampling tool, a magnetic sucker, a magnetic base, an electrostatic discharge rod and an explosion-proof touch ball, one end of the sampling tool goes deep into the oil storage tank through a sampling port at the top of the oil storage tank, and the other end of the sampling tool is connected with the magnetic sucker; the magnetic base is fixedly installed on the side face of the electrostatic discharge rod, the magnetic suction cup and the magnetic base are connected in a magnetic suction mode, the anti-explosion touch ball is arranged at the top of the electrostatic discharge rod, and the electrostatic discharge rod is connected with the ground. According to the utility model, the sampling tool and the electrostatic discharge rod are in magnetic attraction connection through the magnetic assembly, so that the sampling tool can be in grounding connection through the electrostatic discharge rod, the problems of complex field grounding operation and unreliable temporary grounding are solved, and the reliability of grounding connection is improved through integration with human body electrostatic discharge; the device is simple in structure, easy and convenient to operate and convenient to use, and the possibility of electrostatic safety accidents is reduced.