45 results about "Adaptive compression" patented technology

This invention relates to the field of image processing and data transmission technology, specifically disclosing a method and system for adaptive compression and breakpoint resume scheduling of wildlife images. This invention acquires multi-dimensional data of wildlife images in real time, constructs a quantitative evaluation system, and dynamically optimizes data transmission and processing. First, raw data is collected from the wildlife image acquisition terminal, covering core information such as terminal identification, network status, and image content. Then, value assessment coefficients and breakpoint resume scheduling coefficients are calculated sequentially, and an image compression strategy is obtained accordingly. Adaptive compression and breakpoint resume transmission are then executed. Finally, verification, retransmission, and image enhancement optimization are performed on the server side, achieving closed-loop optimization across the entire link from terminal to server. This ensures high-definition transmission and reliable delivery of rare species images even in weak network environments, improving the efficiency and reliability of the monitoring system.

The application discloses a high-definition video low-delay compression method based on adaptive adjustment, and aims to solve the problems of insufficient adaptive adjustment capability, high coding and decoding delay and low matching degree of code rate and network bandwidth in the prior high-definition video compression technology. First, high-definition video stream and real-time network bandwidth parameters are acquired, and features of intra-frame texture complexity, inter-frame motion intensity and scene switching frequency are extracted, and a double-dimension adaptive compression strategy is constructed in combination with real-time bandwidth; low-delay coding is executed by adopting parallel frame slicing and simplified adaptive arithmetic coding, adaptive code rate control is realized by combining code rate overflow early warning and intra-frame key area priority bit allocation, and finally, fast inverse transformation, quantization recovery and motion compensation optimization are completed based on the strategy parameters synchronously transmitted by the encoding end at the decoding end to reconstruct high-definition video frames. The application improves the adaptive adjustment capability of video compression, realizes low coding and decoding delay, and is suitable for real-time high-definition video transmission scenes such as remote monitoring and video conference.

This application discloses a WAL log compression method, device, and medium under PostgreSQL streaming replication, belonging to the database field. The method includes: reading configuration parameters and establishing a streaming replication connection when the master database starts up, obtaining network I / O load values ​​in real time, and calculating the predicted WAL log generation rate based on the transaction processing rate to estimate future bandwidth requirements. When the network I / O or predicted bandwidth requirements reach a preset threshold, samples of the original WAL log data are extracted and compared using multiple algorithms for rapid compression. The optimal compression algorithm is dynamically selected for adaptive compression processing, and the corresponding compression flag field is set; otherwise, the data is transmitted directly in an uncompressed manner. The standby database completes the corresponding decompression or direct writing by parsing the compression flag field. This invention achieves a dynamic balance between network bandwidth and CPU resources through a dual mechanism of load prediction and content awareness, effectively reducing master-standby replication latency and improving the adaptability and overall operating efficiency of the streaming replication system in complex production environments.

The application relates to the technical field of geographic information and data storage, and discloses a high-performance map storage updating method, which solves the problems in the background art, and comprises the following steps: dividing map data into a static layer and a dynamic layer, adopting column storage for the static layer, and adopting an in-memory mapping file for the dynamic layer; adopting Z-Order curve coding for the static layer data in combination with Delta compression, and using Snappy compression for the dynamic layer data and merging continuous frames based on time sequence similarity; adopting a multi-version concurrent control mechanism, associating a version chain with each data block, generating a new version for an updating operation, and appending the new version to the tail of the chain; and constructing a space-time joint index, and storing index partitions in distributed nodes. Through hierarchical storage, adaptive compression, distributed version control and joint indexing, the application effectively reduces storage overhead, improves incremental updating efficiency and space-time query performance, and is suitable for large-scale map data management scenes such as automatic driving and smart city.

The application discloses a kind of high-frequency test data's adaptive compression and lossless fast transmission method and system.The method includes: real-time extraction of high-frequency test data multidimensional dynamic feature vector and mark feature point;Based on feature, network bandwidth and target compression ratio, through pre-training decision model dynamic selection compression algorithm combination and parameter;According to algorithm combination, data is framed and type is preprocessed;With multistage pipeline execution compression, and interstage insertion based on statistical characteristics correction link;Three-dimensional transmission state matrix is constructed, which fuses priority, link and quality score, dynamically allocates link and adaptively adds forward error correction code;After error correction decompression at receiving end, lossless verification is realized by multidimensional comparison of the feature point sequence of reconstructed data and original data.The application realizes the whole process adaptation of compression strategy and transmission scheduling, improves the compression and transmission efficiency while ensuring the lossless and reliable recovery of data.

The application discloses a mobile device-oriented image semantic adaptive compression transmission method and system, relates to the technical field of image transmission communication, and obtains original inertia motion data and interactive operation data of a device, analyzes to obtain an interference image index and an interactive operation fitting broken line; when the interference is significant, the interference is mapped into a global basic compression ratio, and a target area point is extracted from the broken line; a preset radius is calculated based on the target area point and the interference degree, and an interactive focus area is demarcated, a calibration coefficient is calculated according to the intersection relationship of each area; a high-quality first compression parameter used for the interactive focus area and a bandwidth-saving second compression parameter used for the remaining area are obtained according to the calibration coefficient, and differential encoding and transmission are performed according to the calibration coefficient; based on spatial density analysis of historical interactive data, the focus area range is dynamically corrected for subsequent periods. The application realizes adaptive clear transmission of a user attention area and efficient compression of a background area under the condition that the device shakes.

The application discloses a kind of based on federal learning's distribution of network unmanned vehicle distributed fault identification model collaborative training and safe updating method and system.The method includes: multiple edge nodes utilize local multi-modal inspection data, train local model by based on positive and negative sample cross-modal contrast distillation network and calculate local class feature prototype;Node is handled by layered processing and is combined with adaptive compression after uploading model information using homomorphic encryption and differential privacy noise;Center server decrypts correction and calculates global mean and variance prototype, combined with dynamic client selection and proximal term constraint, realize fault aggregation by hierarchical prototype alignment to generate global model;Node receives model, under the double restraint of global prototype, combined with time series attention mechanism to carry out individualized fine-tuning, obtain the identification model adapted to local environment and fault evolution trend;For fault identification and trend assessment.The application realizes efficient, robust, individualized cross-domain fault diagnosis under strict privacy protection.

The present application relates to a kind of based on virtual-real linkage perception large-scale digital twin model adaptive compression and streaming method, comprising: build edge perception networking, form perception data pool;Construction digital twin basic model;Design layered adaptive hybrid compression algorithm, adopt BP neural network dynamic decision each level model compression strategy, according to core layer, middle layer, edge layer executes different hybrid compression in turn and integrates model, obtains integrated layered compression model;Combined with perception data pool adjustment transmission priority, executes dynamic streaming.This application solves the problem that existing fixed compression algorithm is easy to lose key details or transmission delay is too high by layered adaptive hybrid compression algorithm;Reduce the deployment difficulty and operation cost of large-scale digital twin model;Effectively avoid loading lag, interaction lag and other problems, meet real-time monitoring, remote control and other practical application requirements, further promote the popularization of digital twin technology.

The application relates to a GPU-based adaptive compressed data direct query method, system, device and medium, which comprises the following steps: adopting an adaptive compression strategy to compress columnar data to obtain a block-level structure suitable for GPU storage and calculation; loading the compressed data into a GPU memory and performing Tile-level memory management by taking a Tile as a basic scheduling unit; according to a received query statement, performing direct query of the compressed data on the Tile level by the GPU, and outputting a query result. The application realizes high-performance query execution without decompression by designing a Tile-level direct query framework, a hardware-aware memory management and control flow coordination mechanism and an adaptive compression strategy, and fully releases the calculation potential of the GPU. The application can be widely applied to the technical field of big data processing.

This application relates to the field of computer science and discloses an artificial intelligence task scheduling system and method for edge computing. The system includes modules for semantic manifold kernel extraction, heterogeneous resource potential field construction, gradient tensor projection reconstruction, and streaming gradient optimization scheduling. The method utilizes a potential energy function to map heterogeneous network resources into a scalar potential field and calculates negative gradient vectors; it generates a semantic manifold kernel mask based on the semantic entropy density of feature channels, and combines it with an environmental pressure tensor mapped by the negative gradient magnitude to project and reconstruct the model weights to generate a dynamic inference model; it synthesizes a scheduling vector using the negative gradient vector and Lyapunov drift correction to determine the next-hop target node. This invention achieves adaptive model compression and network load balancing while preserving key semantic features through potential field gradient collaborative optimization of the model structure and transmission path, solving the problem of balancing transmission efficiency and inference accuracy in edge computing scenarios.

The deep learning-based video multi-level adaptive compression cross-layer transmission optimization method and system of the application relates to the technical field of video transmission. By designing a compression level selection model based on deep learning, the optimal compression level is output based on the original video data and network environment parameters, and the corresponding compression strategy of the optimal compression level is obtained. A transmission strategy selection model is designed to generate a transmission strategy based on video quality parameters and network environment parameters. A global utility function is defined to calculate the global utility value when the compression strategy and the transmission strategy are assumed to be executed. An optimization objective function is designed, which inputs the compression strategy and the transmission strategy and outputs the optimized compression strategy and the transmission strategy. A condition judgment logic is set to select the final strategy based on the compression strategy and the transmission strategy before and after optimization, thereby achieving global optimization of video compression and transmission.

In several aspects, a computing device analyzes data to determine its characteristics. The computing device selects at least one compression process based on the characteristics. A compression switch and a compression level are dynamically adjusted based on multiple factors including available system resources, desired storage savings and performance requirement. The compression level dynamically varies depending on the data being processed and a workload on a system. Performance of compression and decompression operations are continuously monitored for dynamically adjusting compression parameters to optimize performance.

This invention discloses a method for dynamically updating aeronautical data through air-ground collaboration, belonging to the field of aviation communication technology. The method constructs a dynamic geofence based on flight plans, performs spatiotemporal correlation filtering on multi-source data such as meteorological, aeronautical information, air traffic, and aircraft maintenance data, and transmits the data to the airborne terminal via satellite link after structured transformation, adaptive compression, and hybrid encryption. The airborne EFB equipment receives the data, decrypts it, transcodes it, and verifies its integrity, then presents it to the pilot in multimodal formats including text, voice, and images. This invention effectively solves the problems of poor timeliness and high satellite bandwidth consumption in traditional aeronautical data updates, achieving accurate delivery and dynamic updates of aeronautical data, significantly improving flight safety and operational efficiency.

This invention relates to the field of internet search services and discloses a big data-based investment lead analysis system, comprising: a query parsing unit that calls an event sequence template defining meta-event types and standard relative time window constraint intervals; a distributed retrieval and aggregation unit that retrieves and aggregates meta-event types to entity objects; a temporal logic verification unit that statistically analyzes the global document frequency of meta-events and generates a scaling factor through reverse mapping, dynamically corrects the time window, and determines only entity objects that pass the corrected dynamic verification interval as qualified; and an object sorting unit that generates a sorting score based on the template matching completeness. This invention constructs a reverse coupling mechanism between density and window to adaptively compress temporal logic boundaries in high background noise environments, maintaining the retrieval signal-to-noise ratio and logical confidence.

This application is directed to intelligent video data management and processing optimization. A method includes receiving video streams and analyzing individual frames to extract comprehensive object metadata including physical characteristics, contextual information, and spatial relationships. The method also includes calculating movement scores for detected objects based on frame-to-frame variations and classifying objects into active or passive categories using these scores. The system dynamically predicts object movement probabilities by analyzing historical patterns, object characteristics, and inter-object relationships. The system selectively processes video content based on object importance and predicted behaviors, implementing adaptive storage mechanisms where active objects receive higher processing priority. The method manages video data through intelligent compression and selective processing, improving storage efficiency while maintaining critical information quality. The system addresses challenges of large-scale video management by implementing predictive mechanisms that anticipate and adapt to object behavior patterns.

A multi-modal communication dynamic switching and adaptive compression system and a method thereof for low-altitude intelligent network are provided. The system includes an airborne terminal and a ground terminal; the airborne terminal comprises a NVIDIA Jetson Xavier NX embedded computing platform, an airborne Raspberry Pi CM4-5G expansion board, an airborne ad hoc network radio, a flight control computer and an optical pod camera; the ground terminal comprises a ground Raspberry Pi CM4-5G expansion board and a ground ad hoc network radio; the multi-modal communication dynamic switching and adaptive compression system and the method thereof for low-altitude intelligent network provided by the invention take advantage of the advantages of longer transmission distance, no dependence on ground infrastructure and flexible deployment of the ad hoc network radio, and use it as a supplementary communication means of 5G public network, so as to ensure the real-time online of UAV.

This invention proposes a local sensing adaptive time-series data storage and compression method for smart grids. It takes data feature-driven adaptive compression as the core, combines hierarchical decision tree and LSTM model to adaptively select encoding and compression strategies, dynamically manages fragmentation and builds a three-level index, and combines precision quantization to achieve efficient compression and storage optimization while ensuring query performance. This invention can adaptively select compression strategies, dynamically manage fragmentation, build multi-level indexes and eliminate storage redundancy based on data features.

The application discloses a transient electromagnetic three-dimensional inversion method based on vector field adaptive compression, comprising: data acquisition: constructing a three-dimensional inversion objective function; designing a forward equation according to three-dimensional forward theory; information recoverable vector field compression, compressing electromagnetic field vector data, and extracting key point data information on electric field and magnetic field attenuation curves; information recoverable vector field compression storage; comprising: storage preparation and intermediate data storage; hierarchical storage of key points; storage connection of compressed data; mixed format storage and conversion; final storage of global format. The application adopts hierarchical storage of initial key points and screened target key points, avoids excessive compression waste storage, uses the least square method for secondary compression in a smooth section, accurately retains core data, effectively reduces data redundancy and storage demand, and significantly improves the calculation ability and processing efficiency of large-scale complex model three-dimensional inversion.

The application relates to a large model privacy enhancement cloud-edge collaborative training method and system based on split federated learning, which comprises the following steps: constructing a cloud-edge hierarchical training architecture, implementing model cutting deployment according to the heterogeneity of edge devices, and only performing forward propagation of the frozen layer on the terminal side to generate activation values; a hierarchical adaptive compression strategy is designed for different device types, a feature-aware noise scale model is constructed based on the compression core associated parameters, and Laplace noise is added in the compression domain to realize differential privacy enhancement. After the privacy data is adaptively blocked and transmitted to the cloud, reverse decompression and layer alignment completion are performed. A double protection mechanism is constructed through aggregation shuffling and terminal side differential privacy, and finally, the LoRA low rank adapter is used to independently fine-tune the key layer of the Transformer in the cloud. While ensuring the privacy and security of the whole link, the training efficiency and convergence performance of the large model in the cloud-edge collaborative scene are greatly improved.

This invention proposes a space-based IoT data processing method driven by dual constraints of network and service quality. First, the CPE obtains network status information via a satellite terminal, including maximum network bandwidth, network latency, and packet loss rate, and calculates the maximum data transmission rate (Kbps) that the network can support. The CPE then acquires user-acceptable data error information from the space-based IoT application server; this error information is the data service quality parameter, used to define the upper limit of user requirements for data error. Next, the CPE performs adaptive data compression and transmits the processed data back to the space-based IoT application server. The satellite terminal converts the raw data into an abstract model using a V2M algorithm, and the satellite receives the data and restores the approximate value using an M2V algorithm. During transmission, OVM (Overbounded Data Message) is used to prioritize uploading overbounded data to ensure timely handling of anomalies, while AMM (Aggregated Model Message) is used for batch transmission of common technical solutions.

This invention discloses a multi-point vibration detection system and method for a large-scale passive RFID tag array, comprising a host computer, an RFID reader group, a reader antenna group, and a passive RFID tag array arranged on the surface of the structure to be detected. The signal terminal of the host computer is bidirectionally connected to one end of the RFID reader group, and the other end of the RFID reader group is bidirectionally connected to the signal terminal of the reader antenna group. The antenna terminal of the reader antenna group is bidirectionally wirelessly connected to the passive RFID tag array, and vibration sensing data of several passive RFID tags in the passive RFID tag array are simultaneously collected through the antenna terminal of the reader antenna group. The reader antenna group transmits several vibration sensing data to the RFID reader group, and the RFID reader group transmits several vibration sensing data to the host computer. The host computer sequentially performs adaptive compressed sensing reconstruction operation and adaptive random resonance operation on several vibration sensing data to obtain several enhanced vibration signals of the detected structure.

This invention provides a method for automatic synchronization of multiple databases under an AI platform, belonging to the field of automatic database synchronization technology. This invention initializes the multi-database synchronization topology and deploys a change data capture module to monitor data changes in real time. The captured change records are converted into unified format synchronization messages and input into an intelligent scheduling optimization model for dependency analysis and transaction reordering based on dynamic programming. Data conversion rules are dynamically generated based on mode version differences, and Merkle tree hierarchical structure is used to verify data consistency. After conflict detection and circular dependency handling are performed at the target node, multi-level adaptive compression transmission and data writing are completed. Finally, periodic full reconciliation tasks ensure eventual consistency among the multiple databases, solving the technical problem of low transaction scheduling efficiency leading to synchronization performance bottlenecks during automatic multi-database synchronization.

This invention discloses a token simplification and optimization method and system for multimodal large model calls. The invention extracts independent fingerprints for text, images, and audio, and maintains independent inverted indexes for each modality. It then performs whole-modal pruning based on fingerprint existence and data volume thresholds. Random subspace sampling is performed on the fused feature vectors. Edge features are identified, and three levels of adaptive compression are implemented using a lightweight fidelity predictor. Finally, a fallback mechanism is implemented through dual semantic hash verification. This invention features non-intrusive deployment, does not rely on a unified semantic space, avoids existing patent risks, achieves an overall token saving rate of 21%-26%, downstream task accuracy loss is less than 0.3%, and system resource consumption is extremely low, making it easy for large-scale commercial applications.

A method includes receiving from a client computing device (“CCD)” a request to perform a given post-processing type on compressed image data, receiving from the CCD a first amount of compressed image data, obtaining an image from the first amount of compressed image data, post-processing the image obtained from the first amount of compressed image data with a post-processing algorithm, in response to a result of the post-processing failing to satisfy a predetermined requirement, iteratively, and until the result of the post-processing satisfies the predetermined requirement: requesting a greater amount of compressed image data from the CCD, receiving the greater amount of compressed image data from the CCD, obtaining an image using the greater amount of compressed image data, and post-processing the image obtained using the greater amount of compressed image data with the post-processing algorithm, and returning to the CCD the result of the post-processing satisfying the predetermined requirement.

The present application relates to the technical field of wireless communication, in particular to a Beidou short message compression method and a mobile terminal. The present application collects power consumption parameters and thermal state parameters of the mobile terminal in real time through the drive layer interface of the mobile terminal; determines the current power consumption state level based on the power consumption parameters and the thermal state parameters; matches the corresponding compression algorithm from the preset adaptive compression algorithm library according to the power consumption state level; calls the matched compression algorithm to compress the Beidou short message data; adjusts the transmission parameters of the Beidou radio frequency module according to the power consumption state level, and cooperatively schedules the power consumption module and the thermal management module of the mobile terminal. The present application effectively solves the problem that the compression algorithm and the power consumption management are separated in the prior art, and the mobile terminal power consumption state cannot be dynamically adapted, avoids unnecessary power consumption and CPU temperature rise in low power or high heat scenarios, improves communication efficiency, and avoids the risk of heat superposition in the "compression-transmission" process.

This invention discloses an image processing method based on the Internet of Things (IoT), specifically relating to the field of IoT image technology, including the following steps: Step S1: Visible light image data and infrared thermal imaging data are simultaneously acquired by an IoT image acquisition terminal deployed in the target area, and the current ambient light intensity parameters are obtained; Step S2: In the edge computing module of the IoT image acquisition terminal, visible light and infrared images are acquired simultaneously, and a decision on whether to perform preliminary denoising is made based on real-time illumination detection, so that the preprocessing step forms a conditional response relationship with the environmental conditions; Subsequently, the fused image is adaptively compressed and uploaded to the cloud, where the cloud model analyzes the noise characteristics and generates targeted parameters to be sent back to the terminal, which then performs final denoising accordingly; In this process, the linkage between illumination judgment and denoising action avoids invalid calculations, and multimodal fusion provides complementary information for subsequent analysis.

This application provides an adaptive compression method and electronic device for image and text browsing scenarios, relating to the field of image processing technology. In image and text browsing scenarios, fixed bitrate compression can be applied to the image and text content. Based on user browsing operation information (such as browsing speed / dwell time / editing status), image attribute information (such as multiple images / thumbnail / large image), device power consumption mode (such as whether it is in power-saving mode), and device performance status (such as whether it is in low-performance state), the fixed bitrate compression ratio is dynamically adjusted to achieve adaptive compression control. This effectively controls system resource usage, reduces memory consumption and bandwidth consumption, lowers overall power consumption, improves performance, and provides visually lossless effects. Especially for electronic devices with lower performance (e.g., small memory, low CPU / GPU performance), it solves problems such as lag and insufficient battery life.

This invention discloses a method and system for classifying gas drainage coupled anomaly monitoring by integrating AI time-series analysis, belonging to the field of coal mine safety monitoring technology. The method includes: collecting time-series data of coupled parameters to generate a multi-time-resolution original time-series dataset; performing step-wise time-series processing on the original time-series dataset to generate regularized time-series data; constructing a multi-dimensional time-series hazard index sequence and mapping the regularized time-series data to this sequence; adaptively identifying coupled mutation trends using a good-point set-improved adaptive compression factor particle swarm optimization algorithm and outputting mutation identification results; weighted fusion of the multi-dimensional time-series hazard indicators to generate a dynamic hazard index, and dynamically adjusting the update frequency of the dynamic hazard index according to the mutation identification results; comparing the latest dynamic hazard index with a preset classification threshold to output the current risk level of the target monitoring area. This improves the real-time performance, accuracy, and stability of gas drainage anomaly monitoring.

The present application relates to a small model enhanced medical large model clinical decision reasoning method, belonging to the technical field of medical information processing, solving the problem that the input length of the large model is limited in the prior art, affecting the clinical reasoning performance. The method comprises: obtaining the medical information of a plurality of patients in each department to construct a sample set corresponding to the department; constructing a small model corresponding to each department for dynamic self-adaptive compression of the input data; training the small model of the department based on the sample set of the department to obtain a trained small model; fine-tuning the medical large model based on double-channel optimization; the medical large model is used for key field extraction and clinical decision reasoning; inputting the medical information of the patient to be predicted into the small model to obtain the compressed medical information of the patient to be predicted; inputting the compressed medical information of the patient to be predicted into the fine-tuned medical large model for clinical decision reasoning. The accuracy and reliability of medical text analysis and clinical reasoning are improved.

The application provides a first-view video key frame extraction method and system based on work semantics, and belongs to the technical field of computer vision; the method comprises the following steps: acquiring a first-view video stream and IMU sensor data collected by an AR device; performing motion de-jittering preprocessing based on the IMU data; extracting human key points through a lightweight human posture estimation model, and calculating work action saliency scores; detecting action start and end points based on a saliency curve, and extracting key frames; adaptively compressing the key frames to generate a simplified feature data packet; the application realizes key frame extraction from 'picture change' to 'work semantics' according to the characteristics of the first-view work operation video, guarantees work action retention rate, achieves a compression ratio of more than 10:1, can be applied to intelligent monitoring scenes of power distribution network construction sites, and effectively solves the video transmission problem under the condition that the network bandwidth of the distribution network site is limited.