277 results about "Cloud processing" patented technology

The invention discloses a laser scanning optimization system for high-precision ellipticity and diameter measurement, and relates to the technical field of cylindrical workpiece geometric dimension measurement. Comprising a laser contourgraph assembly, the laser contourgraph assembly comprises four laser contourgraphs, the four laser contourgraphs are arranged in the circumferential direction of the cross section of a to-be-measured workpiece at equal intervals, the included angle between every two adjacent laser contourgraphs is 90 degrees, and the laser contourgraphs are used for synchronously collecting contour point cloud data of the surface of the to-be-measured workpiece; and the data acquisition module is used for receiving and summarizing the contour point cloud data. The four laser contourgraphs are arranged at equal intervals in the circumferential direction of the cross section of the workpiece and collect synchronously, so that the problems that a traditional single laser module is incomplete in rotary scanning coverage and has timeliness deviation are solved; and the point cloud processing module improves the condition of insufficient point cloud processing refinement through splicing, direction correction, section extraction, adaptive region division and arc length interception processing.

The invention relates to a control method, system and device for a security robot and a medium, and the method solves the problem of perception failure in smog and low-light scenes through the adaptive adjustment of the weight of a multi-modal sensor through dynamic weighting space-time alignment processing. Constructing a cross-modal threat field function based on a lightweight Transform, and realizing quantitative modeling of threats in a complex environment; a federal reinforcement learning framework is utilized to fuse local decision and encrypted global strategy update, and cloud processing delay is avoided while data privacy is guaranteed; a mixed potential field path is generated in combination with the threat field gradient and strategy parameters, and the limitation of traditional path planning on dynamic threat avoidance is broken through; and finally, a multi-robot surrounding track is generated through a distributed cooperation protocol, so that the high-risk target surrounding efficiency and the system robustness are remarkably improved. According to the method, perception-decision-execution full-link optimization is realized, and a high-response, strong-privacy and collaborative intelligent solution is provided for a complex security scene.

The invention provides an Internet of Things smart farm monitoring method and system, and relates to the technical field of agricultural Internet of Things, and the method comprises the steps: executing the triggering type sensing operation of crop growth and environment abnormality at a field end side visual node of an Internet of Things smart farm based on a preset triggering condition, and outputting an end side initial sensing data set; performing non-redundant feature screening on an end-side initial sensing data set, extracting a key feature set, executing a staged slow aggregation operation through a low-power-consumption communication network, outputting batch feature data packets according to a link state, and realizing field light-flow transmission; the cloud processing platform receives the batch feature data packets, then calls the agronomy knowledge set for restoration and diagnosis, outputs a farm monitoring diagnosis result, generates a field equipment linkage instruction according to the farm monitoring diagnosis result, sends the field equipment linkage instruction to the execution equipment, and outputs an equipment linkage response result, so that the farm monitoring efficiency and accuracy are improved, and the farm monitoring accuracy is improved. Intelligent management is realized, and the field operation cost is reduced.

The invention relates to the technical field of computer vision, and particularly provides a tunnel video stream three-dimensional modeling system based on binocular stereo matching and SLAM (Simultaneous Localization and Mapping), which comprises a multi-modal image acquisition unit, a binocular infrared camera and an RGB (Red, Green and Blue) camera are configured to synchronously acquire an infrared image pair, an RGB image pair and inertial measurement unit data of a tunnel environment, the multi-source data is uploaded to the cloud processing platform in real time through the wireless transmission module; based on a dynamic switching mechanism of environmental perception, performing adaptive fusion pose estimation of a feature point method and a direct method on input multi-modal image data; the stereo matching and dense reconstruction unit is used for calculating a disparity map by adopting an improved self-adaptive window stereo matching algorithm, generating a three-dimensional point cloud in combination with the pose information output by the pose estimation unit, and constructing a global consistency point cloud model of the tunnel scene through a time sequence point cloud registration and splicing algorithm; and a multi-algorithm target detection and semantic fusion unit. The method can meet the requirements of precision and robustness of tunnel reconstruction.

The invention discloses a robot grabbing pose generation method and device, and the method comprises the steps: obtaining a natural language grabbing task instruction, and carrying out the analysis based on a large language model, and generating a structured task parameter containing target object identification information and grabbing constraint information; performing open vocabulary target detection on the image based on the identification information, and obtaining a target area mask in combination with pixel-level segmentation; generating a target point cloud based on the mask, the depth map and the camera internal reference, carrying out depth completion or denoising, outlier elimination and threshold confirmation filtering, inputting the processed target point cloud into a grabbing pose prediction network, outputting a plurality of candidate six-degree-of-freedom grabbing poses, carrying out constraint screening according to grabbing constraints, and carrying out collision detection; and selecting the candidate with the optimal confidence coefficient from the collision-free candidates as the optimal grabbing pose to be output. Task constraints penetrate through a target extraction process, a point cloud construction process and a pose screening process, and point cloud processing and collision verification are combined, so that an executable six-degree-of-freedom grabbing pose which meets the constraints is output.

The invention relates to the technical field of new energy automobile batteries, in particular to a new energy automobile battery state prediction method, optimization control method and system, and the method comprises the steps: collecting the real-time operation data of a new energy automobile battery; uploading the real-time operation data to a cloud processing platform in which historical operation data of the battery is stored, and dynamically predicting the state of the battery; fusing the charge state data driven by the real-time model and the charge state data driven by the historical data to carry out charge state estimation prediction, and outputting charge state estimation parameters; a circulation attenuation model is combined with reinforcement learning correction, historical circulation data is used for fitting a capacity attenuation curve, and a dynamic battery health state estimation value is output; a fusion thermal model and an internal resistance model are adopted, a safety voltage / temperature boundary is predicted through digital twin simulation, and a dynamic battery power state estimation value is output; and generating a battery control strategy based on the battery state parameters and issuing the battery control strategy to an onboard BMS module for execution.

The invention discloses a substation equipment fault diagnosis and defect early warning system supporting cloud-side cooperation. The substation equipment fault diagnosis and defect early warning system comprises a data acquisition module, an edge calculation unit, a cloud processing center, a state evaluation module, an early warning generation module and a feedback control unit. According to the application, primary data processing and real-time analysis are realized through the edge computing unit, comprehensive evaluation is completed in combination with a cloud deep learning engine, and the diagnosis efficiency and accuracy are remarkably improved; the self-adaptive adjustment module dynamically optimizes resource allocation and enhances the response capability of the system; graded alarm and multi-channel notification ensure timely early warning; and the feedback control unit realizes remote and local linkage to guarantee safe operation of equipment. According to the invention, large-scale complex equipment management requirements are effectively met, and reliable support is provided for intelligentization of a power system.

The invention discloses a sign language-voice conversion system, and belongs to the technical field of auxiliary communication and wearable computing. The system comprises a wearable myoelectricity acquisition module used for acquiring double-arm myoelectricity signals when a user executes sign language; the mobile terminal module is wirelessly connected with the acquisition module and is used for receiving and preprocessing the signal and uploading the signal; the cloud processing module is used for receiving the signal, converting the signal into text information through a sign language recognition model, and further calling a voice synthesis service to convert the text into voice data; and the wearable audio output module is used for receiving and playing the voice data. Through an innovative end-to-end hardware system architecture, natural, accurate and real-time translation and voice output of sign language gestures are realized, communication barriers between hearing-impaired people and healthy hearing people are effectively solved, and the system has the advantages of flexible deployment, user friendliness and privacy protection.

The invention discloses a friendly charging pile management system and method for a pure electric vehicle, and relates to the technical field of charging piles. Comprising a sensing layer, a cloud processing layer and a user interaction layer which realize data interaction through an encrypted communication protocol. The vehicle type identification module receives three frames of images of entering vehicles collected by a wide-angle camera, after Gamma gray correction, 3 * 3 Gaussian filtering and zooming to 224 * 224 and normalization preprocessing, vehicle types are classified, pure electric vehicles enter a priority queue of the double-queue scheduling module, and mixed extended-range electric vehicles enter a lag queue; the double-queue module builds queues according to Redis Zset and ranks the queues every 5 seconds according to a priority rule, and when the charging piles are idle, scheduling with the highest priority is selected; an identity verification algorithm of the parking lock control module is used for unlocking through license plate editing distance comparison and 120-second time sequence verification, queue jumping is avoided, and the problems that pure electric charging resources are occupied and queuing is disordered are solved.

The invention discloses a method for integrating an industrial field information system and a cloud platform based on edge computing. The method comprises the following steps: locally acquiring original monitoring data of a multi-source sensor of industrial equipment at an edge node; according to the method, the original data of the multi-source sensor is locally acquired and preprocessed through the edge node, and only the standardized time sequence data set and the feature vector are transmitted to the cloud, so that the original data transmission quantity is greatly reduced, and the network bandwidth pressure and the communication cost are relieved; feature engineering is carried out on the edge side in real time, a predictive maintenance model is deployed to carry out local reasoning, and an abnormal probability or an RUL predicted value is output, so that data processing link delay is remarkably shortened, and the real-time performance of health state assessment is improved; a maintenance decision is quickly triggered based on an evaluation result, so that instant response and accurate maintenance of equipment abnormity are realized, and fault expansion caused by cloud processing delay is avoided; meanwhile, an edge and cloud collaborative optimization model is combined with global data continuous iteration, and prediction accuracy and adaptability are improved.

The invention discloses a vehicle-mounted data transmission method, device and equipment and a computer readable storage medium. The method comprises the following steps: acquiring to-be-transmitted data, and determining a target cache according to a data type of the to-be-transmitted data; if similar data related to the to-be-transmitted data exists in the target cache, the similar data are removed, the to-be-transmitted data are stored in the target cache, the similar data and the to-be-transmitted data belong to the same parameter attribute, and the transmission content of the similar data can be replaced by the to-be-transmitted data; and if the sending condition is met, the data in the cache is sent to the cloud, so that the same instruction is prevented from being repeatedly transmitted to the cloud, and meanwhile, the vehicle-mounted network bandwidth and cloud processing resources are saved.

The invention relates to the technical field of steam turbine shafting vibration monitoring, and discloses a steam turbine shafting vibration intelligent diagnosis and edge early warning method based on multi-source data fusion, and the method comprises the following steps: deploying multiple types of sensors at key measurement points of a steam turbine shafting, and constructing a monitoring network with complementary spatial dimensions; according to the method, multiple types of sensors are deployed at key measuring points to collect multi-source data, time sequence and frequency domain features are extracted, feature layer fusion is carried out, the weight is dynamically adjusted in combination with real-time working conditions, the limitation of a single data source is broken through, and the shafting state is comprehensively reflected; local real-time diagnosis is realized by deploying a lightweight model through edge nodes, and meanwhile, the model is updated by incremental learning to improve the generalization ability of complex working conditions, so that the problem of high cloud processing delay is solved, and the real-time early warning requirement is met; the false alarm rate is reduced by combining a dynamic threshold adjustment mechanism with equipment health baseline and real-time working condition deviation correction, and performing graded early warning on the overrun amplitude and confidence of a reference vibration predicted value.

The invention relates to the technical field of water body pollution monitoring, in particular to a multispectral image fused water body pollutant accurate recognition system, which comprises a data acquisition module, a cloud processing module, a boundary processing module, a pollution recognition module, a diffusion prediction module and a visualization module, the system constructs sub-pixel representation of a water pollutant boundary by using a differential geometry manifold theory, and realizes high-precision pollutant boundary description through a multi-scale analysis and curvature flow optimization technology; enhancing pollutant characteristic expression by adopting multispectral image fusion and an optimal wave band selection technology; using a support vector machine model to accurately identify various pollutant types such as oil films, oil spots, algae blooms and the like; based on a boundary fine description result and a pollutant type identification result, the diffusion trend of pollutants is accurately predicted in combination with historical flow, wind direction and wind speed data, and the system improves the water pollutant boundary identification precision and enhances the identification capability of complex boundary forms and low-contrast regions.

The invention discloses a parking state intelligent sensing method and device based on target detection tracking, and relates to the technical field of image recognition. The method comprises the following steps: deploying a vehicle target detection model and a vehicle target tracking model in a cloud server to all edge computing gateways of a parking lot monitoring system; inputting a video stream collected by the camera into a vehicle target detection model of an edge computing gateway, and carrying out vehicle target detection to obtain a plurality of vehicle targets; inputting the plurality of vehicle targets into a vehicle target tracking model of the edge computing gateway, and performing vehicle target tracking to obtain active tracks of the plurality of vehicle targets; and performing dynamic optimization by using an improved optimization algorithm according to the active tracks of the plurality of vehicle targets to obtain a parking state list of the parking lot. The problems that in the prior art, cloud processing is relied on, detection and tracking are disjointed, state perception is simple, and the model generalization ability is weak are solved.

The invention relates to the technical field of gem detection and three-dimensional modeling, in particular to a gem high-precision three-dimensional modeling and flaw detection method based on reverse engineering, which comprises the following steps: S1, fixing a gem to be modeled, and acquiring image data of different wavelengths by using a multispectral imaging device; s2, performing distortion correction, noise suppression and light intensity normalization processing on the multispectral image data; s3, based on the preprocessed multispectral image data, refractive index inversion is carried out in combination with a gem optical model and light propagation characteristics, and point cloud data containing space coordinates and refractive index information is generated. According to the method, the gemstone is subjected to multispectral imaging acquisition, and image preprocessing, refractive index inversion, point cloud processing and self-adaptive grid reconstruction are combined, so that high-precision three-dimensional modeling and defect marking are realized, and the problems that traditional gemstone detection mostly depends on manual observation, and due to the lack of multispectral information, the gemstone detection accuracy is low are solved. Therefore, the problem that internal defects cannot be accurately identified is solved.

The invention provides a reservoir operation risk real-time early warning method based on digital twinning, and relates to the technical field of reservoir operation engineering safety monitoring, and the method comprises the steps: firstly constructing a reservoir digital twinning body, collecting hydrological, structural and environmental data in real time through a multi-source sensor, and uploading the hydrological, structural and environmental data to a cloud processing center; then dynamically calibrating twin model parameters by using a machine learning algorithm based on historical and real-time data to realize virtual-real synchronous mapping; then calculating a hydrological situation deviation degree representing a hydrological abnormal state and a structure response coupling coefficient representing a structure response characteristic; generating a comprehensive risk index through a risk fusion algorithm, and triggering graded early warning according to a dynamic threshold value; and finally, the early warning information is displayed through a visual platform, and disposal suggestions are provided. According to the invention, real-time, accurate and panoramic perception and intelligent early warning processing of the reservoir operation risk are realized, and the initiative and reliability of reservoir safety management are significantly improved.

The invention relates to the technical field of building automation and machine vision, and particularly discloses a building decoration assembly system and method based on image point cloud processing. According to the system, semantic annotation and feature enhancement are performed on a three-dimensional point cloud through an image semantic segmentation map, and a semantic enhanced point cloud is generated; iterative registration based on semantic consistency is carried out on the decoration component digital model and the semantic enhancement point cloud, and an accurate installation pose is output; planning a collision-free assembly track according to the pose and the semantic segmentation map; and the robot is driven to execute assembly along the track, and pose deviation detection and dynamic compensation are carried out based on the real-time image and the point cloud. According to the method, deep penetration and unified driving of image semantic information in the whole process of environment perception, pose calculation, path planning and closed-loop execution are realized, and the technical problems of insufficient precision, poor adaptability and dependence on manual intervention caused by disjunction of perception, planning and execution links of an existing assembly system are effectively solved.

The invention provides a laser SLAM (Simultaneous Localization and Mapping) navigation method for point cloud processing and pose optimization bidirectional closed-loop regulation, which comprises the following steps: screening a current frame point cloud under a spherical coordinate system by using a previous frame pose uncertainty scalar and a previous frame loop importance scalar and adopting a point cloud screening method to obtain a current frame reserved point cloud; judging whether the current frame is a key frame or not based on the reserved point cloud of the current frame, storing the current frame to a key frame set only when the current frame is the key frame, and obtaining a current optimal pose and a pose uncertainty scalar of the current frame by adopting a pose optimization method based on the key frame set; the current optimal pose is used as a current positioning result of the navigation system.

The invention discloses an edge cloud collaborative multi-modal data real-time processing method and system. The method mainly comprises the steps that 1) a data sensing system directionally collects multi-modal data; 2) the edge device preprocesses the multi-modal data and uploads the multi-modal data to the cloud computing center; 3) the cloud computing center constructs a multi-modal data model to extract high-order feature vectors of the multi-modal data; 4) the cloud computing center performs data fusion processing on the extracted feature data to obtain fusion feature data with unified dimensions, and transmits the fusion feature data to the edge device; and 5) the edge device performs post-processing and application output on the fusion data. According to the invention, multi-modal data is processed through edge cloud, multi-source information is fused and real-time response is carried out to improve processing efficiency and accuracy, bandwidth pressure and cost are reduced through local preprocessing and edge resource scheduling, and system reliability is enhanced through a distributed architecture and data redundancy.

The invention relates to auxiliary equipment for preventing and resolving contradictory disputes and a control method thereof, and belongs to the technical field of corollary equipment for preventing and resolving contradictory disputes, the auxiliary equipment comprises a cloud processing device, a plurality of computer terminals and mobile terminals are arranged outside the cloud processing device, and signal lines are arranged between the computer terminals and the cloud processing device and between the mobile terminals and the cloud processing device. The cloud processing device comprises a case, a memory is arranged in the case, a signal switch is arranged at the rear end of the memory, a CPU processor is arranged on the side edge of the signal switch, a power supply is arranged at the lower end of the memory, and a power socket communicated with a power supply phase circuit is arranged at the rear end of the power supply. The cloud processing device serves as a core and is connected with computer terminals and mobile terminals used by a plurality of workers through a wired or wireless network to form a cooperative work network. According to the invention, the conversion from passive response to active prevention and from artificial experience to intelligent assistance is realized, and the efficiency, standardization and safety of contradictory dispute handling are improved.

This invention discloses a method for 3D mesh reconstruction of buildings from multispectral LiDAR point clouds, belonging to the field of multispectral LiDAR point cloud processing technology. The method includes: extracting building instances from the input point cloud using a multispectral LiDAR point cloud classification method; extracting geometric features from the height map obtained from the building point cloud, and then extracting spectral features from the spectral map using a spectral adaptive convolution module; inputting the data into a spectral-geometric fusion module to fuse the features of the spectral map and the geometric map to achieve accurate prediction of 2D corner points; projecting the 2D corner points onto the height index of the height map to obtain 3D corner points, and using an attention mechanism jointly guided by multispectral and geometric features to predict the boundary structure of the 3D corner points, generating a wireframe model; finally, achieving seamless and watertight 3D mesh reconstruction of the building under the guidance of the topological relationships contained in the wireframe model. This invention can effectively reduce artifacts and planar misalignment problems during the reconstruction process.

An integrated cyber-physical system and method for high-fidelity distress monitoring and automated alert error mitigation. The system comprises a wearable multi-parametric sensing suite (100) for primary anomaly detection and an autonomous mobile robotic agent (200) for secondary contextual verification. Upon detection of a potential distress event (S110), the robotic agent (200) executes a priority undocking sequence (S120) from a charging station (500) to perform multi¬ modal acquisition (S130), including 3D skeletal pose estimation and acoustic sentiment analysis. A Cloud Processing Engine (300) synthesizes these data streams — including optional spatial occupancy data from an environmental sensor network (400) — to calculate a Validation Confidence Score (VCS) (S140) using a dynamic weighted fusion algorithm. This score enables the system to autonomously perform alert de-escalation (S150) for verified safe-states or execute prioritized escalation protocols (S160) to a caregiver interface (600) and an emergency services gateway (700). The invention effectively reduces alarm fatigue and improves emergency response reliability by providing contextual validation of biometric triggers.

The application discloses a kind of based on multi-mode wireless sniffing and cloud computing mobile device real-time positioning monitoring method and system, the method includes: distributed sniffing terminal is in listening mode, the sniffing data of surrounding mobile device is collected, and the sniffing data is encrypted and uploaded to cloud processing platform;Cloud processing platform carries out unified access and time synchronization to the sniffing data of multiple terminals, through multi-mode heterogeneous data fusion and association mechanism, WiFi sniffing data and Bluetooth sniffing data from the same mobile device are matched, and position estimation is completed in combination with multi-point observation and multi-band signal characteristics, and the positioning result of mobile device is output;The cloud processing platform is based on the positioning result of multiple mobile devices, generates dynamic heat map in user interface display, to support historical trajectory query and regional people flow density statistical analysis.The application realizes high concurrency, real-time, wide coverage mobile device sniffing, positioning and heat visualization.

The invention provides an image information extraction method and device, electronic equipment and a storage medium, and relates to the technical field of image processing, and the method comprises the steps: carrying out the model compression processing of an optical character recognition model, so as to obtain a lightweight optical character recognition model; performing text recognition on the input image by using the lightweight model at the edge equipment end to obtain text content and position information, then judging whether to trigger cloud processing according to a text recognition result, and sending the input image and the text recognition result to the cloud during triggering; the cloud server performs multi-modal information extraction through a large language model to obtain structured information, so that the problems that in the prior art, an optical character recognition model is difficult to adapt to the computing power of edge equipment, and end-cloud cooperation lacks a reasonable triggering mechanism, so that response delay or resource waste is caused, and the efficiency is high can be solved. And edge equipment cannot directly deploy a large model to realize high-precision structured information extraction.

The present application provides a point cloud processing method and apparatus, a device, and a storage medium. The method comprises: acquiring a current frame point cloud collected by a point cloud collection apparatus and a global model, wherein the global model is obtained by fusing historical frame point clouds collected by the point cloud collection apparatus, any three-dimensional point in the global model carries a weight, and the weight of any three-dimensional point represents the possibility that the three-dimensional point is a cluttered point or a target object; after the current frame point cloud is projected into the global model, projecting a plurality of light rays between the projected current frame point cloud and the point cloud collection apparatus to determine a target three-dimensional point through which the light rays pass in the global model; reducing the weight of the target three-dimensional point; and deleting the target three-dimensional point if the reduced weight of the target three-dimensional point satisfies a cluttered point deletion condition. Therefore, the cluttered point is accurately deleted.

The invention relates to the technical field of data processing, and discloses a method for controlling NAS (Network Attached Storage) equipment based on a large model and network attached storage device.The method comprises the following steps: extracting instruction intention information of user instruction information according to a tool list, mapping an instruction intention into an instruction task in combination with context information of the user instruction information, determining a logic dependency relationship of the instruction task according to the instruction object and the constraint condition, determining a task step according to the logic dependency relationship, generating a tool calling request according to the target tool screened in the tool list, and executing the instruction task in the NAS device according to the tool calling request. It can be seen that by implementing the method, the use threshold of NAS advanced functions can be reduced, the user function control and interaction efficiency can be improved, a large model locally deployed by the NAS does not need to depend on cloud processing, the risk of user operation instruction or task leakage is reduced, the data safety and reliability are improved, end-to-end automation from user instruction to task execution can be achieved, and the user experience is improved. And the NAS control efficiency is improved.

A point cloud processing apparatus includes circuitry to generate a selection screen displaying a plurality of options each of which indicates a corresponding one of a plurality of qualities for point cloud processing that is performed on point cloud data representing a three-dimensional point cloud. The circuitry stores, in a memory, a processing result of the point cloud processing performed with one of the plurality of qualities on the point cloud data. The one of the plurality of qualities is a selected one of the plurality of options.

A point cloud processing device and a point cloud processing method are provided. The point cloud processing method includes: receiving a vertically scanned point cloud; performing a random sample consensus algorithm on the vertically scanned point cloud to obtain a fitting plane; filtering a first ground point of the vertically scanned point cloud according to the fitting plane to update the vertically scanned point cloud; and generating a processed point cloud according to the updated vertically scanned point cloud and outputting the processed point cloud.

The invention discloses a road surface point cloud multi-frame registration method and system based on an improved ICP algorithm, and relates to the technical field of road inspection, and the method comprises the steps: obtaining the multi-frame road surface point cloud data and initial attitude estimation of a frozen soil region, and calculating a space overlapping ratio and a dynamic proportion through a three-dimensional grid; calculating a density increasing demand score and a density reducing pressure score based on a space overlapping rate and a dynamic proportion, adaptively determining a thinning proportion, adopting non-uniform voxel sampling, and recalculating a normal vector; the method comprises the steps of establishing an initial corresponding relation between a source point cloud and a reference point cloud, calculating normal reliability, voxel occupancy uniformity and a preheating interior point rate, adaptively selecting a registration strategy through net information revenue, including a point-to-point ICP algorithm, a point-to-surface ICP algorithm or a two-stage ICP algorithm, and solving optimal rigid body transformation; applying the transformation to the original high-density point cloud, processing an overlapping region by adopting a weighted fusion strategy, and generating a registered high-density point cloud and log; the method effectively improves the registration precision and efficiency, and is suitable for frozen soil region pavement point cloud processing.

The invention discloses a mountain building extraction method based on deep learning, and relates to the technical field of remote sensing image processing and deep learning, and the method comprises the steps: obtaining multi-source remote sensing data, and carrying out the cloud removal processing and annual median synthesis in sequence, and obtaining a cloud-free annual synthesis image; constructing a high-resolution reference data set and a low-resolution reference data set, and performing division and data enhancement according to a target proportion to obtain an enhanced training set and an enhanced verification set; training the enhanced super-resolution semantic segmentation network model to obtain an initial enhanced super-resolution semantic segmentation network model; jointly optimizing model parameters by using high-resolution loss, time comparison loss and spatial generalization loss to obtain a target enhanced super-resolution semantic segmentation network model; and detecting and extracting the to-be-detected multi-source remote sensing data to obtain a high-precision resolution mountain building range extraction result. According to the method, the detection precision and generalization ability of the mountain building are improved by using the multi-layer residual error and the pixel rearrangement structure.