104results about How to "Improve the extraction effect" patented technology

The invention relates to a load prediction method TSR-MAEN based on decomposition representation and an attention encoder network. Firstly, the algorithm provides a trend-seasonal decomposition representation method based on comparative learning so as to enhance the decoupling performance of a model on a trend component and a seasonal component of a load sequence and the representation capability of decomposition characteristics. Secondly, in order to deeply mine and decompose time sequence information contained in a feature hidden space and dependency between features, the algorithm proposes a mixed attention encoder network; through segmented embedding, a trend-season dynamic projection mixed self-attention mechanism and a multi-head mixed output mechanism, trend and season multi-dimensional features after load sequence decomposition are effectively utilized to express and enhance the long-range dependence information extraction capability of the model. Experiments show that the TSR-MAEN algorithm provided by the invention has excellent performance in an electrical load prediction task of oil and gas development, and the prediction error is reduced by 8.51% compared with that of an optimal baseline algorithm.

The invention belongs to the technical field of geological disaster monitoring and early warning, and discloses a slope geological disaster multi-source remote sensing image intelligent early warning method and system, and the method comprises the steps: a multi-source remote sensing data adaptive collection step: constructing a multi-source remote sensing data cube; a multi-scale spatio-temporal feature coupling extraction step: generating a spatio-temporal coupling feature vector based on an improved Swin Transform architecture and an adaptive spatio-temporal attention mechanism; a disaster risk dynamic assessment and feedback step: calculating a disaster risk level based on a multi-modal deep learning network, and dynamically adjusting a feature fusion weight according to a risk assessment result to form closed-loop feedback; and an intelligent early warning decision and response step: constructing a multi-stage early warning threshold system to trigger early warning and generate an early warning report, the slope geological disaster early warning method realizes multi-scale spatial-temporal feature adaptive fusion and a closed-loop cooperation mechanism, and significantly improves the accuracy and intelligent level of slope geological disaster early warning.

The invention discloses a resonant grounding power distribution network single-phase grounding fault line selection method and system. The method comprises the following steps: when a single-phase earth fault occurs in the power distribution network, acquiring real-time zero-sequence current data of each feeder line; and inputting the real-time zero-sequence current data into the fault line selection model to judge a fault line and output a result. The training process of the fault line selection model comprises the following steps: acquiring a noisy source domain data set from an electromagnetic transient simulation platform and a target domain data set from a real power distribution network; constructing a noise reduction automatic encoder as a feature extractor, and performing unsupervised pre-training by using the source domain data set to minimize a reconstruction error; pre-trained feature extractor parameters are frozen, and a trainable fault classifier is connected to form a transfer learning model; and carrying out supervised training on the model by using the source domain and target domain data sets and taking the fault line selection accuracy as a target to obtain a final model. By implementing the method provided by the invention, the accuracy and generalization ability of the line selection model in a real noise environment can be effectively improved through transfer learning and noise reduction feature extraction.

This utility model discloses a gas staged extraction pipeline structure, including an upper branch pipe, a lower branch pipe, and a connecting pipe. The upper branch pipe is located above the lower branch pipe and the two are connected by the connecting pipe. Multiple connecting pipes are provided, spaced at predetermined intervals. Each connecting pipe connects to a borehole extraction pipe, which in turn connects to a borehole. One end of the upper branch pipe is closed, and the other end is connected to the main gas pipe via a branch pipe. One end of the lower branch pipe is also closed, and the other end connects to a water drainer. By setting up the upper and lower branch pipes and connecting them with the connecting pipe, which is then connected to the boreholes, the gas and accumulated water from each borehole are discharged together into the short connecting pipe. Gas enters the upper branch pipe, and accumulated water enters the lower branch pipe. The upper branch pipe connects to the gas extraction pipeline, and the lower branch pipe connects to the water drainer, achieving separate extraction of gas and accumulated water, improving extraction efficiency, avoiding the need for a separate water drainer, and reducing equipment costs.

The present application provides a kind of inner slag discharging device and its slag discharging method, it includes high wear-resistant obturator, drill bit drill rod device, venturi water and drill slag dry-wet separator, the drill bit drill rod device is located in high wear-resistant obturator, drill slag dry-wet separator is connected with drill bit drill rod device by venturi water.The flow direction of drill slag can be changed by changing the running direction of slag discharging medium, so as to reduce the scouring, soaking and friction of slag discharging medium and drill slag on the hole wall, which is beneficial to the formation of drilling hole;It can achieve the purpose of point coring by achieving "falling and discharging simultaneously";Drilling cuttings can be well collected, which is beneficial to the civilized production of construction site and prevents dirtiness and disorder;If inert gas is used for slag discharging during construction, it can also achieve the purpose of inhibiting spontaneous combustion of coal seam.Through the above technical scheme analysis, the device is theoretically feasible and the operation is reliable, which provides good drilling quality for casing, greatly improves the success rate of drilling and casing and the extraction effect.

This invention relates to a vacuum extractor that improves extraction performance through a vacuum treatment unit, which improves extraction performance by recovering extracts contained in the air drawn in when a vacuum is generated. The vacuum extractor includes: an extraction chamber for extracting extracts from a raw material using a solvent; a chamber heating unit for heating the extraction chamber with a heat medium to improve extraction performance by heating the raw material; and a vacuum treatment unit for generating a vacuum in the extraction chamber to facilitate extraction of the extract from the raw material, while simultaneously recovering the extracts contained in the drawn-in air using a vacuum pump. The vacuum treatment unit includes: a first cooler for cooling the air and recovering the extracts when a vacuum is generated in the extraction chamber using a vacuum pump; a cyclone tank for generating a cyclone and recovering the extracts contained in the air passing through the first cooler; a second cooler for further cooling the extracts contained in the air passing through the cyclone tank and recovering the extracts; and a recovery tank for storing the extracts recovered from the second cooler.

This application relates to the field of semantic analysis technology, specifically to a method and system for extracting audit anomaly semantics from unstructured data. The method includes: acquiring a benchmark graph and a benchmark database and their entity mentions; obtaining the strings of entity mentions based on their positions; acquiring an audit candidate set of strings from the benchmark database; determining a matching score based on the consistency of audit roles, and then filtering the candidate entity set; encoding the candidate entities, their audit role sets, entity mentions, and strings to obtain the probability distribution of entity mentions and candidate entities, and determining the audit role consistency score based on the probability distribution; classifying the association between entity mentions and candidate entities based on the overall magnitude of the audit role consistency score, and applying different audit rules to each category; and extracting audit anomaly semantics using the audit rules. This application improves the accuracy and efficiency of semantic extraction.

The utility model discloses an exhaust device for an experiment operation table. The exhaust device comprises a fixed assembly, a sliding assembly, a rotating assembly and an air suction assembly, the fixing assembly is fixed to the table edge of the operation table and comprises a first fixing frame and a second fixing frame, and the first fixing frame and the second fixing frame are connected through a telescopic rod so as to adjust the length of the fixing assembly; the sliding assembly is arranged on the fixing assembly and moves in the length direction of the fixing assembly. The rotating assembly is arranged on the sliding assembly and used for connecting the sliding assembly and the air suction assembly. The elevation angle of the air suction assembly is changed through the rotating assembly, the air suction assembly comprises an air suction plate, the air suction plate comprises a plurality of telescopic plates, and a plurality of air suction ports are formed in the telescopic plates. The air suction assembly can reach most positions of the operation table, poisonous gas generated at any test point on the table top can be better sucked away, and the situation that due to the fact that the suction effect is poor, workers inhale the poisonous gas, and the bodies of the workers are damaged is prevented.

This application discloses a method, method, and apparatus for training an information extraction model. The main technical solution includes: acquiring a third dataset, which includes multiple unlabeled third text samples; using a pre-trained first information extraction model to extract target information of a preset type from the third text samples in the third dataset, as a second target sample; inputting the second target sample into a pre-trained text generation model to obtain text generated by the text generation model using the second target sample as a second text sample; forming a second sample pair by combining the second target sample and its corresponding second text sample; obtaining a second dataset using multiple second sample pairs; and training a second information extraction model using the second dataset. This application can improve the information extraction effect of the information extraction model while reducing manual costs and time consumption.

This invention discloses a maritime rescue method based on improved YOLOv4 target detection, belonging to the field of maritime rescue. It includes: an improved pixel-by-pixel filtering image preprocessing method (de-raining, de-fogging) for processing images captured by a high-definition camera mounted on a drone, converting blurry images into clear images to provide effective data for subsequent image detection; an improved YOLOv4 target detection algorithm for accurately identifying people in the water; and the use of a Feature Pyramid Transformer (FPT) to enhance the semantic extraction capability of images of small floating objects (people awaiting rescue), thereby improving the search and rescue success rate. Adding a Feature Pyramid Transformer (FPT) after the traditional feature pyramid structure enhances the feature map of semantic information for small targets, which is beneficial for detecting small objects at sea.

The invention belongs to the technical field of coal mine drilling tool equipment, and particularly provides a drilling, fracturing and hole protecting integrated drill bit which comprises a connecting rod section and a drill bit section, one end of the connecting rod section is connected with an external drill rod, the drill bit section is assembled at the other end of the connecting rod section, the drill bit section is divided into an inner layer and an outer layer, and a gap is reserved between the inner layer and the outer layer. An annular screen pipe moving power valve is arranged in the gap, a water outlet is formed in the end face of the drill bit section, and an inner sleeve is arranged in the connecting rod section. Drilling, fracturing and hole protection are integrated, after drilling construction is completed, in-hole real-time machining, hole protection screen pipe lowering and fixed-point multi-load fracturing are carried out from the bottom of a hole to the outside of the hole, the problem of hole collapse caused by drill withdrawal is solved, the protection pipe lowering rate is increased, the drilling fracturing precision is improved, the fracturing effect is enhanced, and the drilling gas extraction effect is greatly improved.

The invention discloses a sub-echo reflectivity correction method and system based on a B-spline model, and the method comprises the steps: firstly, fully quantifying the irregular features of a laser radar waveform signal through the B-spline model, improving the precision of multi-echo signal information interpretation in a complex environment, and providing reliable initial parameter support for the subsequent sub-echo reflectivity correction; then, by means of a distribution simulation algorithm and vegetation vertical structure distribution, ground and different vegetation target point clouds are classified into two types, and the two types are reflectivity correction algorithm priori knowledge; thirdly, deducing a reflectivity correction algorithm from a radar basic principle equation, and further deriving a multi-layer chain type correction equation; according to the method, the reflectivity of the multiple sub-echoes is further corrected from the standard ground reflectivity and the first target reflectivity correction, a single echo of a single vegetation target is not needed, the precision of a reflectivity correction algorithm is greatly improved, and the application range of the reflectivity correction algorithm is greatly widened.

The invention discloses a wake detection model construction method based on frequency domain multi-scale perception, which relates to the technical field of ocean observation, is used for capturing weak wake features to carry out sea surface internal wave wake detection, and comprises the steps of constructing a wake image data set, constructing a wake detection model by taking an RT-DETR model as a baseline model, and constructing a multi-scale texture information selection network. A backbone network of the baseline model is replaced, a frequency domain multi-scale attention perception module is constructed, an AIFI module of the baseline model is replaced, a golden cudgel convolution down-sampling module is adopted to replace a standard convolution down-sampling module of the baseline model, training and optimization are performed on the wake detection model based on the wake image data set, and a target detection result is obtained. According to the method, the wake detection model AUVRT-DETR is constructed, the detection precision is improved while the advantage of light weight is kept, the optimal balance is achieved between the detection precision and the calculation efficiency, and the model parameter quantity and the calculation overhead are remarkably reduced while the highest detection precision is achieved.

The invention discloses a mining-induced fissure zone gas extraction dynamic regulation and control system and method, and relates to the technical field of coal seam gas extraction, and the system comprises a data acquisition unit which is used for obtaining micro-seismic event source parameters, resistivity data and extraction data of a target measurement area; the data inversion unit is used for constructing a first objective function of the micro-seismic event source parameters, extracting micro-seismic parameters in the first objective function of the micro-seismic event source parameters through a moment tensor analysis mechanism, constructing a second objective function of the resistivity data, and performing inversion to obtain the gas concentration; the model construction and verification unit is used for the micro-seismic and gas concentration space mapping capability of the three-dimensional geologic model; and the optimal regulation and control execution unit is used for carrying out collaborative optimization on the extraction instruction by applying an intelligent extraction optimal algorithm model, so that the flow of each branch and the power of the pump station can be accurately scheduled, and the aims of saving energy and reducing consumption are fulfilled. Accurate targeted extraction of the high-concentration gas enrichment area is achieved, and energy consumption of the low-concentration area is reduced.

This invention discloses a passive RFID tilt detection method based on a transceiver separation architecture, comprising the following steps: acquiring a continuous wave radio frequency signal from an RF exciter; collecting a mixed signal containing a passive tag reflected wave and an RF exciter leakage wave, and simultaneously acquiring a self-interference reference signal; performing self-interference cancellation processing on the mixed signal based on the self-interference reference signal, and extracting the tag backscattered signal; when a dual-tag array is used, measuring the phase of the passive tag array separately and calculating the phase difference, calculating the initial tilt angle of the target object, and obtaining the final tilt angle after correction; when a single passive tag is used, extracting the phase data of the single passive tag and performing linear interpolation, and substituting the interpolated real-time phase into a linear model to calculate the tilt angle. This invention can effectively eliminate transceiver self-interference and automatically offset the influence of distance changes on angle measurement accuracy, significantly improving the robustness and accuracy of object tilt detection.

The utility model relates to camellia oil processing technical field, concretely relates to a kind of auxiliary structure for camellia oil extraction process, it includes: base, the one side of base top end is provided with extraction device, the other side of base top end is provided with separating device, extraction device includes extraction tank, heating pipe is around and connected in the middle part of extraction tank, the inner wall surface of extraction tank is fixedly connected with several temperature sensors, the top end of extraction tank is fixedly connected with cover plate, stirring device is provided in the inside of extraction tank, separating device includes separation tank, separation motor is fixedly connected at the bottom of separation tank, filter cartridge is rotatably connected in the inside of separation tank, the output end of separation motor is connected with the bottom of filter cartridge by spline, the setting of heating pipe and temperature sensor, heating pipe can evenly heat the material in extraction tank, temperature sensor real-time monitoring temperature in extraction tank, accurate control is carried out to extraction temperature, provide suitable temperature for camellia oil extraction, guarantee the quality of camellia oil.

The invention discloses a document information extraction method and device, equipment and a storage medium. The method comprises the following steps: firstly, obtaining a first to-be-processed document and a target document category and a target extraction field of the to-be-processed document, and converting image information in the to-be-processed document into an editable text format to obtain a second to-be-processed document comprising a plurality of service sub-documents; and in combination with a VLM model, determining the service sub-document, which conforms to the target document category, in the plurality of service sub-documents as a target document. And if the page number of the target document is greater than or equal to a page number threshold value or the Token number is greater than or equal to a Token threshold value, performing key information retrieval on the target document based on the target extraction field through a retrieval enhancement generation (RAG) technology to obtain a target text. And generating a cue word in combination with the target text, the target extraction field and a preset instruction, and inputting the cue word into a VLM model for information extraction to obtain an extraction result of the target document. According to the method and the device, document information extraction can be efficiently and accurately performed.

The utility model discloses a clear water extraction device applied to sewage treatment, which comprises a sedimentation tank, the sedimentation tank is connected with a sewage treatment tank through a pipeline, the treated sewage supernatant can flow into the sedimentation tank through the pipeline, the supernatant is separated for the first time, one side of the sedimentation tank far away from the pipeline is provided with a through hole, and the through hole is communicated with the sedimentation tank. An extraction pipe communicated with the through hole is arranged on the outer side of the sedimentation tank, a pump body is arranged on the extraction pipe, a supporting frame is fixed to the upper side of the sedimentation tank, a mounting hole communicated with the through hole is formed in the upper side of the sedimentation tank, a threaded rod is screwed on the supporting frame, and a filter plate is rotationally mounted at the lower end of the threaded rod and located in the through hole through the mounting hole; after clear water entering the sedimentation tank is settled in the sedimentation tank for a period of time, the pump body is started to pump out supernate in the sedimentation tank, the supernate is separated again, and impurities in the supernate can be filtered through the filter plate in the pumping-out process, so that the bottom suspended impurities are further prevented from being pumped into the clear water; and the clean water pumping effect is good, and recycling of clean water in the later period is facilitated.

This invention discloses a method, system, device, and medium for automatic extraction of black soil layer interfaces using ground-penetrating radar (GPR) based on an improved U-Net, belonging to the field of GPR data processing and intelligent interpretation technology. The method includes acquiring and preprocessing GPR B-scan image data, constructing a training dataset, building a layer extraction model based on an improved U-Net neural network model, constructing a hybrid loss function including at least two combinations of weighted binary cross-entropy loss, Dice loss, structural similarity loss, and smoothness loss, training the layer extraction model based on the training dataset and the hybrid loss function, and inputting the GPR B-scan image data to be detected into the trained layer extraction model to achieve automatic extraction of black soil layer interfaces using GPR. This invention effectively overcomes signal attenuation and noise interference in GPR images, achieving high-precision and continuous soil layer interface extraction.

The application discloses a kind of based on machine haptics's car gauge level semiconductor device surface microdefect detection method, belong to semiconductor device quality detection technical field.The application is by establishing the quantitative coupling relationship of microdefect geometric parameter and haptics signal characteristic parameter, design high-density flexible piezoelectric haptics sensor array, and using signal stable enhancement under high-speed scanning and multi-force level signal fusion technology, construct the car gauge level microdefect intelligent discrimination model based on improved C3D-Attention, realize to submicron level microcrack, nanometer scratch, glue surface wrinkle and other surface microdefects High-precision online detection.The application can overcome the bottleneck of traditional machine vision in high light, heterogeneous surface is susceptible to light interference, high miss rate, cannot obtain three-dimensional mechanical information, etc., false detection rate ≤100PPM, satisfy AEC-Q100 car gauge reliability requirement, can be adapted to automobile electronic production line high-speed scanning and batch detection, provide technical support for car gauge level semiconductor advanced manufacturing quality control and device reliability improvement.

The application discloses a visual positioning method and device, a storage medium and an electronic device, and relates to the technical field of information. The method comprises the following steps: acquiring a current frame image and a previous frame image corresponding to the current frame image; extracting feature points and feature lines corresponding to the previous frame image and feature points and feature lines corresponding to the current frame image, and performing feature point matching and feature line matching on the previous frame image and the current frame image to obtain a feature point matching pair and a feature line matching pair; constructing a point reprojection error and a line reprojection error between the previous frame image and the current frame image based on the feature point matching pair and the feature line matching pair; constructing a target optimization function according to the point reprojection error and the line reprojection error, and solving a relative pose between the current frame image and the previous frame image; and performing visual positioning on a target object according to the relative pose. The application can overcome the limitation of feature point extraction in a weak texture environment, thereby improving the accuracy and robustness of visual positioning.

The invention discloses a training method of a target detection model, a target detection method and related equipment. The training method comprises the following steps: training a teacher target detection model by using a source domain data set D1 in a cross-camera difference data set; training a student target detection model by using the source domain data set D1; the source domain data set D1 and the target domain data set D2 are used at the same time, a knowledge distillation method is adopted, feature knowledge and response knowledge obtained by the teacher target detection model are provided for student target detection models to learn, the student target detection models are adjusted, and the adjusted student target detection models are provided for target detection. Therefore, by adopting the cross-camera difference data set, the response-based knowledge distillation method and the feature-based knowledge distillation method, the feature extraction capability of the detection model can be improved, the prediction capability of the detection model can also be improved, and the detection capability of the target detection model is greatly improved under the combined action; and introduction of additional parameter quantity and calculation quantity is not needed.

The application relates to the technical field of automobile luggage rack connecting assemblies, in particular to a fixed connecting assembly capable of improving connecting strength and a use method thereof, which comprises a roof original luggage rack fixing rod and connecting bolts, the upper end of the roof original luggage rack fixing rod is connected with a connecting assembly, and the upper end of the connecting assembly is fixedly connected with a roof luggage rack through the connecting bolts; the connecting assembly comprises an assembly shell, the upper end of the assembly shell is fixedly connected with a connecting block, the inner sides of the assembly shell are fixedly connected with limiting columns at both ends, the outer sides of the limiting columns are provided with second springs, and the outer sides of the limiting columns are slidably connected with clamping plates; in the application, the sealing buffer mechanism can further improve the connecting tightness between the connecting assembly and the roof original luggage rack fixing rod, thereby improving the connecting strength of the connecting position; meanwhile, in the installation process, the adsorption between the sucker and the roof original luggage rack fixing rod can prevent the luggage rack from being displaced under the condition of frequent emergency braking of the automobile.

The application relates to the technical field of fire safety detection, and specifically discloses a smoke and fire detection method based on an enhanced YOLO algorithm, innovative data enhancement of constructing multi-scene sample data sets and fusing space-time features, proportionally dividing training, verification and test sets, designing an enhanced YOLO network containing feature extraction, enhanced fusion and a detection head, improving a backbone network and the detection head, designing a hybrid loss function fusing multiple types of losses, training and optimizing the model in stages, collecting images through a camera, realizing real-time detection and three-level early warning through model inference, and incrementally updating the model. The application solves the problems of feature distortion and poor small target detection of traditional methods, improves the adaptability to complex environments and detection precision, realizes timely early warning, is suitable for various scenes, and has high practicability.

This invention discloses a method for identifying pet breast cancer lesions based on abnormal sample DBT images in the field of pet medical imaging technology. The method acquires and preprocesses 3D images of abnormal DBT in pet breast tissue with lesion annotations, generating a 2D section sequence and a three-branch sample set. A pet breast lesion feature encoder is obtained through lesion prior comparison self-supervised training. Human breast DBT anatomical knowledge is aligned with pet data to achieve cross-species knowledge transfer. The 2D section sequence is input into the encoder and detection head to obtain 2D lesion detection results. A 3D lesion body is generated through lesion spatial trajectory growth and aggregation, and the final lesion identification result is output. This invention achieves high-precision detection relying solely on abnormal samples, eliminating the need for normal samples, diffusion models, and knowledge distillation. It adapts to the physiological structural characteristics of pet breast tissue, improves the detection effect of small lesions and the accuracy of 3D localization, and is suitable for rapid clinical screening of pet breast cancer.

The utility model relates to the technical field of American ginseng saponin extraction design, in particular to American ginseng saponin extraction equipment. Comprising an extraction box, and the extraction box internally comprises a feeding cavity used for conveying an American ginseng extracting solution, a working cavity connected with the feeding cavity and a condensation cavity installed on one side of the working cavity; wherein a filter cartridge, a first rotating shaft, a first stirring rod, a second stirring rod, a cutting knife, a driving assembly and a heating pipe are arranged in the working cavity; and a condensation pipe communicated with the upper part of the working cavity is arranged in the condensation cavity. According to the American ginseng saponin extraction equipment disclosed by the utility model, due to the arrangement of the feeding cavity, the working cavity and the condensing cavity, the extraction efficiency and the extraction effect of American ginseng saponin are greatly improved, and the American ginseng saponin extraction equipment is suitable for industrial application.

The application provides a non-invasive vibration diagnosis system and method for mechanical state of a transformer tap changer, the system comprising a mounting base, a pre-tightening coupling mechanism, a directional vibration transmission main probe, a background vibration transmission reference probe, a double-channel synchronous collector, an analysis host and a display communication terminal. The method comprises: positioning a single tap action event; decoupling an oil tank propagation mechanism and inverting compensation attenuation, and enhancing vibration under strong noise; extracting vibration characteristics, and performing action chain deconstruction, stage decomposition, multi-domain feature construction and fault deviation modeling; identifying mechanical defect modes and grading; evaluating reliability, and attributing and self-adapting repair parameters when the reliability is low, to form a closed loop. The application realizes non-invasive high-reliability diagnosis of the outer wall, improves information extraction, defect identification and grading accuracy in complex scenes, and supports abnormal early warning and online operation and maintenance.

The invention belongs to the field of artificial intelligence, and relates to a video key frame extraction method and device based on self-supervision, equipment and a medium, and the method comprises the steps: obtaining a frame sequence of a target video, extracting the feature vector of each frame in the frame sequence through a pre-training feature extractor, and obtaining a feature vector sequence; determining graph nodes according to the frame sequence, calculating edge weights according to the feature vector sequence, and constructing an initial directed graph; calculating a graph Laplacian matrix, screening positive and negative sample frame pairs from the frame sequence by using a target sampler, and calculating time sequence perception comparison loss; circularly using parameters of the loss updating feature extractor to finely adjust the feature vector, update the edge weight, construct the graph and the like until the loss converges; reconstructing a directed graph, and calculating a Laplacian matrix of the graph to obtain a Laplacian matrix of a target graph; and performing Laplacian matrix spectral clustering on the target image to obtain a target video key frame set. The method can be applied to the business fields of financial science and technology, insurance, medical treatment and the like, and can improve the extraction accuracy of the video key frame.

This invention discloses a method, device, electronic equipment, and product for monitoring the implementation of land spatial planning, relating to the field of image classification technology. The method includes acquiring aerial images of the target area of ​​the approved plan; extracting SIFT features from the aerial images and local color features of the neighborhood of each feature point in the SIFT features; fusing the SIFT features with the local color features of the neighborhood of each feature point in the SIFT features to obtain a first fused feature; extracting global color features from the aerial images; fusing the first fused feature with the global color features to obtain a second fused feature; and using the second fused feature as input to a pre-trained classification model to obtain the land use type of the target area, so as to monitor the implementation of land spatial planning in the target area based on the land use type. This invention can automatically and dynamically monitor the implementation of land spatial planning in real time, saving human resources.