96 results about "Adaptive resolution" patented technology

The invention discloses a low-altitude path planning method and system based on a dynamic adaptive graph neural network, and belongs to the technical field of artificial intelligence and robot autonomous navigation. Aiming at the problems of large calculation amount of a fixed grid map, slow dynamic response of a static octree, failure of a traditional heuristic function on a non-uniform graph and the like, the invention provides the following scheme: constructing a resolution self-adaptive non-uniform graph by using the octree; a Morton code is adopted to realize rapid positioning of a dynamic obstacle and millisecond-level updating of a map; an attention mechanism is introduced into a graph attention network (GAT), weights are adaptively allocated to different neighbor nodes, features are aggregated, and a high-precision heuristic cost estimation value is output; and finally, integrating the estimated value into an algorithm to perform efficient path search. According to the method, the memory occupation and the calculation complexity are remarkably reduced, and the planning efficiency and the path safety in a dynamic environment are improved.

The invention provides an SLAM (Simultaneous Localization and Mapping) optimization method and equipment based on point cloud registration and adaptive resolution, and a medium, and relates to the technical field of AGV (Automatic Guided Vehicle) positioning and mapping, and the method comprises the following steps: (1) inputting and preprocessing sensor data; (2) front-end scanning and matching: based on the preprocessed data, carrying out point cloud registration through a GICP algorithm, updating the pose of the current frame and constructing a local subgraph; (3) global SLAM: receiving sensor data, poses and subgraph data from the front-end scanning and matching step, and carrying out loopback detection and global optimization; wherein a self-adaptive resolution switching strategy based on environment complexity is adopted in the loopback detection process; according to the GICP algorithm, the surface of the point cloud is modeled as Gaussian distribution, local features are described through a covariance matrix, and a target error function constructed based on the mahalanobis distance is optimized to estimate pose transformation. According to the method, the local matching precision and robustness are improved by replacing ICP with GICP, and the global calculation efficiency is adjusted and optimized in combination with the dynamic resolution.

The invention relates to a high-fidelity generative video stream transmission system based on a visual base model, which belongs to the field of image communication, and is characterized in that a visual enhancement-oriented generative codec is designed, and high-fidelity video reconstruction under a high compression ratio is realized through an asymmetric space-time compression strategy and time sequence consistency enhancement; a resolution scaling module is provided, the calculation complexity is remarkably reduced through a video super-resolution recovery module of adaptive resolution control and joint optimization, and real-time high-definition video processing is achieved; and constructing a network adaptive video stream transmission controller, an intelligent token discarding mechanism based on semantic importance and a mixed packet loss processing strategy to realize code rate scalable control and robust transmission under network fluctuation.

The invention belongs to pattern recognition and artificial intelligence, and particularly provides a multi-modal large language model optimization method and system based on adaptive resolution, and the method comprises the steps: dividing multi-modal question and answer data into a target task training set and a target task test set; constructing a resolution selection training set by using the recollected image data in the target scene and the paired text instruction; constructing a multi-modal large language model with a two-stage image coding mechanism, and training the multi-modal large language model by using the target task training set and the resolution selection training set to obtain a target model; the prediction accuracy of the target model is evaluated on the target task test set, and optimization of the multi-modal large language model is completed. According to the technical scheme, the optimal resolution is adaptively selected for reasoning through a two-stage image coding mechanism, so that the visual perception capability of the model is remarkably improved.

The invention discloses a robot fine operation control system based on a diffusion model and 3D vision, particularly relates to the technical field of industrial robot fine operation control, and aims to solve the problem that when an existing industrial robot executes fine operations such as bottle cap screwing and connector plugging in and out in a complex three-dimensional operation scene, the operation efficiency is high. The problems that three-dimensional perception is unstable, the action resolution is fixed, space and contact constraints are difficult to meet, and the task success rate and safety are insufficient are solved. The method comprises the following steps: constructing a voxel feature field updated along with time under a self-body coordinate system, generating interaction complexity and a physical constraint field based on task coding, and linking adaptive resolution and action granularity scheduling, diffusion trajectory generation and closed-loop execution correction to uniformly map environmental geometry, contact conditions and task stages into tail end trajectory control, so as to realize real-time tracking control. Therefore, fine operations such as bottle cap screwing, connector inserting and pulling and flexible object folding can still be stably completed under complex shielding and contact constraint, and the effects of safety and task success rate are both considered.

A railway roadbed disease diagnosis method based on a convolutional neural network comprises the following steps: acquiring vibration acceleration signals of a train and a railway roadbed, and performing data cleaning on the acquired vibration signals; signal preprocessing, time-frequency domain transformation and feature vector extraction, convolutional neural network-based model training and feature extraction, feature classification and disease recognition, and real-time disease detection and location positioning are realized. According to the invention, a modern signal processing technology and a deep learning algorithm are combined to realize efficient and automatic detection of railway roadbed diseases; signal segmentation, time-frequency domain analysis, image segmentation and convolutional neural network technologies are adopted, a self-adaptive resolution adjustment mechanism is achieved, accuracy and high efficiency of signal feature capture and disease diagnosis are achieved, and the nonlinear distortion problem of multi-band signals can be effectively solved through signal segmentation processing and summarizing fusion.

An X-ray detector comprises a detector panel (100) comprising a surface with a plurality of detector pixels, wherein the detector pixels are configured to convert received X-rays into electric signals representative of values indicative of the received X-rays. A data interface (308) transmits images having a fixed number of output pixels. A control unit controls to set (401) a region of interest comprising at least a subset of the plurality of detector pixels, wherein a number of detector pixels in the region of interest is independent of the number of output pixels. The control unit maps (403) the region of interest onto an output image having the fixed number of output pixels. The control unit resamples (404) the detector pixels within the region of interest to obtain values for the output pixels of the output image, and transmits (405) the output image via the data interface.

A method includes receiving a reference frame, determining a scaling constant for a current block, determining a scaled reference block using the reference frame and the scaling constant, determining a scaled prediction block using the scaled reference block, and reconstructing pixel data of the current block using the scaled prediction block. Related apparatus, systems, techniques, and articles are also described.

This disclosure signaling of motion vector or motion vector difference dependent on whether magnitude-dependent adaptive motion vector difference pixel resolution is employed or not in video encoding and decoding. An example method for processing a video block of a video stream is disclosed. The method may include receiving the video stream; determining that the video block is inter-coded based on a prediction block and a motion vector (MV), wherein the MV is to be derived from a reference motion vector (RMV) and a motion vector difference (MVD) for the video block; extracting or deriving, from the video stream, a data item associated with at least one of the RMV or the MVD.

A method, device, and non-transitory computer-readable medium for decoding an encoded video bitstream using at least one processor, including, obtaining from the encoded video bitstream a first flag indicating that reference picture resampling is enabled for coded video sequence (CVS); determining a value of a second flag indicating whether a picture resolution is changed in the CVS; based on the value of the second flag indicating that the picture resolution is changed in the CVS, decoding the CVS using the reference picture resampling for spatial scalability and adaptive resolution change (ARC); and based on the value of the second flag indicating that the picture resolution is not changed in the CVS, decoding the CVS using the reference picture resampling for the spatial scalability, and without using the reference picture resampling for the ARC.

A video decoding apparatus and a method for adaptively setting a resolution are disclosed. According to one embodiment of the present invention, a method for adaptively setting a resolution on a per-picture basis comprises the steps of: decoding maximum resolution information from a bitstream; decoding, from the bitstream, resolution information about a current picture; and setting the resolution of the current picture on the basis of the maximum resolution information or the resolution information, wherein the resolution information has a size less than or equal to that of the maximum resolution information.

The invention particularly relates to a hierarchical search three-dimensional airspace path planning method and system based on an adaptive resolution grid. The method comprises the following steps: constructing a corresponding multi-resolution grid model based on a three-dimensional grid map; global path planning is carried out on the coarse resolution grids, coarse paths of all coarse grid nodes are determined, and a coarse path sequence is constructed; performing non-overlapping segmentation on the coarse path sequence; planning local windows matched with the fine-resolution grids, and determining window entrance and exit points corresponding to the local windows; performing obstacle recognition on each local window based on the fine resolution grid, calling a corresponding path generation strategy according to an obstacle recognition result of the local window, and determining a local refined sub-path corresponding to each local window; and splicing the path information of each local window in sequence to obtain a path planning result. According to the method, efficient and reliable path planning of a large-scale and complex three-dimensional airspace environment can be realized.

The application discloses a segmented adaptive Fourier transform analysis method of power quality disturbance. The method firstly collects power quality disturbance signals in a power grid and performs Fourier transform, acquires peak envelope points in a frequency spectrum through a maximum envelope method to dynamically select characteristic frequency points and calculate frequency spectrum segmentation points; subsequently, a composite objective function is used to optimize Gaussian window parameters for each segment; finally, the signal is analyzed by using the optimized window function to obtain time-frequency results with adaptive resolution. The application allocates higher time resolution to strong time-varying components and higher frequency resolution to near steady-state components through segmented adaptive resolution and a composite objective function optimization strategy, effectively avoids the spectrum overlap and energy diffusion problems that are prone to occur in traditional methods, improves the accuracy and robustness of disturbance feature extraction, and can be widely applied to power grid fault diagnosis, disturbance identification and early warning.

The invention provides a distributed deployment method and system for an unmanned aerial vehicle three-dimensional wireless communication network, and relates to the field of unmanned aerial vehicle communication networks, and the method comprises the steps: obtaining the geographic information of an inspection region, constructing a three-dimensional terrain matrix, and carrying out the self-adaptive resolution voxelization; based on a voxelization model and unmanned aerial vehicle communication parameters, considering path loss and obstacle penetration loss, and constructing a three-dimensional communication quality potential field; designing an anti-collision repulsion potential field and a communication blind area gravitational potential field, and obtaining a total potential field gradient of unmanned aerial vehicle inspection; based on the real-time information, updating the three-dimensional position of each unmanned aerial vehicle in a distributed manner by adopting a gradient descent method; determining the responsibility coverage area of each unmanned aerial vehicle according to the updated position, and cooperatively determining the beam direction and the beam width; and defining a network coverage efficiency and communication quality balance index, and dynamically adjusting a learning rate to complete real-time online optimization. According to the invention, autonomous deployment and optimization of the unmanned aerial vehicle communication network can be realized, and communication coverage efficiency and quality balance are improved.

The invention relates to the technical field of robot environment perception, in particular to a real-time ground segmentation method and system. The method comprises the following steps: receiving and preprocessing laser radar point cloud data, and dividing the data into a plurality of concentric annular areas; carrying out adaptive ground fitting on each area and processing the multilayer structure through vertical plane fitting to obtain an initial ground point set; correcting an under-segmentation region of the current frame by using a time sequence backspacing mechanism and updating parameters; constructing an elevation map based on the initial point set, and obtaining a fine ground point set through adaptive resolution CSF simulation; and carrying out confidence evaluation on the two types of point sets, and generating a final result through multi-criterion fusion. Accurate segmentation of the complex terrain is realized, and robustness and real-time performance of a dynamic environment are improved.

The application discloses a serial port screen GUI adaptive resolution rendering method and system, and belongs to the technical field of human-computer interaction. The method comprises the following steps: segmenting string data to obtain at least one data segment, and generating a unique sequence identifier; embedding the sequence identifier, the total length of the string data and the effective data length of the current segment in the frame structure of each data segment based on the data frame format limitation of the bus protocol to obtain a data frame carrying segment metadata; sequentially sending the data frame through the bus; dynamically allocating a buffer according to the sequence identifier and the total length information at the receiving end, and recombining each data segment according to the sending order according to the sequence identifier, obtaining display instructions after recombination and performing a rendering operation. Through the segment transmission and metadata identification mechanism, the application breaks through the limitation of the bus protocol on the length of single-frame data, realizes reliable transmission and accurate recombination of large-capacity string data, and improves the stability and adaptability of serial port screen rendering.

The invention relates to a self-adaptive display method for post-processing resolution, a three-dimensional scanning system and a storage medium, and the method comprises the steps: transmitting scanning data obtained through the scanning of a three-dimensional scanning device to a processor; responding to the operation of the user on the interactive interface, and generating an adjustment instruction of the self-adaptive resolution; transmitting the adjustment instruction to a processor; in the interactive interface, displaying target point cloud data returned from the processor based on the adjustment instruction; the target point cloud data is obtained by performing fusion processing on the scanning data at least according to the first target resolution and the second target resolution; the first target resolution and the second target resolution are obtained by performing adaptive adjustment on a preset second resolution under the adjustment instruction; the problem that the use experience of a user is affected in the related technology is solved. And the processor adaptively adjusts the first target resolution under the adjustment instruction, so that the multi-resolution fused high-quality target point cloud data is displayed in the interactive interface, and the user experience is optimized.

The invention discloses a flatness error hybrid iteration identification method based on an on-machine measurement point cloud, and relates to the field of precision manufacturing and digital measurement, and the method comprises the steps: collecting three-dimensional point cloud data of the surface of a workpiece through line laser scanning, carrying out the normalization modeling of the three-dimensional point cloud data, and carrying out the mapping through employing a self-adaptive resolution; performing three-dimensional convex polyhedron iterative optimization on the processed point cloud data, and quickly screening and eliminating redundant points; carrying out dimension reduction processing on the screened data from three-dimensional dimension reduction to two-dimensional dimension reduction, and further compressing the data scale by adopting two-dimensional convex polygon iterative optimization; and introducing a key point control strategy based on a minimum region criterion, rapidly solving a flatness error, and outputting a flatness error identification result. According to the method, a mixed iteration solving framework combining convex optimization and key point control is constructed, the point cloud data scale and calculation complexity are effectively reduced, rapid processing of large-scale point cloud data is achieved, and the strict requirements of online measurement for real-time performance and rhythm are met.

The invention discloses a self-adaptive green screen image matting management system and method based on artificial intelligence, and relates to the technical field of neural networks, and the method comprises the steps: collecting multi-source data of a green screen site through different sensors, and constructing an end-to-end self-adaptive image matting neural network; processing the multi-source data by using an encoder to output features of different dimensions, and fusing the three paths of features by using a self-adaptive fusion weight to output a comprehensive feature; constructing a decoder by using a U-Net structure, outputting three predicted values, and correcting the foreground by using the three predicted values to output a final foreground; constructing a self-adaptive resolution adjustment mechanism, and dynamically adjusting the down-sampling proportion of the image; setting a physical constraint to separate a foreground from a background, and constructing an optical flow guided time domain smoothing mechanism to carry out refined filtering on an image edge region; and inputting the green screen multi-source data into the reasoning assembly line to output an image matting result, and performing quality evaluation and feedback updating on the image matting result.

The invention belongs to the technical field of deep learning rotary machine fault diagnosis, and particularly relates to a rotary machine multi-modal fault diagnosis method based on a multi-level cross attention fusion network, which comprises the following steps of: adopting a time-frequency transformation method of kurtosis screening based on variational mode decomposition and constant Q non-stationary Gabor; original signals from different sensors are respectively converted into time-frequency diagrams with adaptive resolution; constructing a fusion adaptive multi-branch attention module, and comprehensively and finely extracting fault features through a parallel multi-scale structure and a double attention mechanism; and constructing a multi-level fusion strategy and a final fusion module based on cross attention, carrying out dynamic and adaptive feature fusion and refining on the extracted fault features from a shallow layer to a deep layer, and outputting final feature representation with high discrimination. According to the method, complementary information from different sensors is effectively fused, so that the accuracy and reliability of fault diagnosis are remarkably improved.

A method includes receiving a bitstream, using the bitstream and determining, for a current frame, whether the current frame is usable as a reference frame, responsive to determining that the current frame is usable as a reference frame, setting a variable characterizing that an adaptive resolution management mode is not allowed, and reconstructing pixel data of the current frame, wherein the adaptive resolution management mode is not allowed. Related apparatuses, systems, techniques, and articles of manufacture are also described.

A mechanism is described for facilitating adaptive resolution and viewpoint-prediction for immersive media in computing environments. An apparatus of embodiments, as described herein, includes one or more processors to receive viewing positions associated with a user with respect to a display, and analyze relevance of media contents based on the viewing positions, where the media content includes immersive videos of scenes captured by one or more cameras. The one or more processors are further to predict portions of the media contents as relevant portions based on the viewing positions and transmit the relevant portions to be rendered and displayed.

The application discloses a single-level C-ADC circuit structure based on double-voltage zoom signal and single reference voltage comparison and a double-zoom voltage sampling method thereof. The converter comprises an adaptive resolution module, a double-voltage zoom module, a three-input comparator module and a control logic module. The sampling method comprises the following steps: generating two signals which are synchronous with the change of an analog signal to be converted; comparing the two generated synchronous signals with a single reference voltage signal respectively; triggering sampling when any synchronous signal crosses the reference voltage from a high bit or a low bit; generating an output pulse signal to indicate the direction of crossing the reference voltage when the sampling is triggered, and resetting the two signals used for synchronizing the input change, and repeating the above process. The event-driven analog / digital converter and the sampling method thereof reduce the accuracy requirement for the reference voltage, improve the accuracy of the analog / digital converter and save the power consumption of the analog / digital converter.

This application discloses a method, apparatus, and device for generating images of multimodal railway foreign objects, aiming to improve the accuracy of identified railway foreign objects. The application acquires a railway track surface image containing the target section and textual description information; inputs the textual description information into a text encoder to obtain a semantic condition vector; inputs the railway track surface image into a track surface detection model to obtain a rail bounding box; obtains a first target region and a second target region based on the rail bounding box; fuses the first and second target regions to obtain a target image; and inputs the target image and semantic condition vector into a multimodal diffusion model to obtain railway foreign object information. By fusing textual description information and the railway track surface image as generation conditions, and employing an adaptive resolution differentiation processing strategy for the first and second target regions, not only is high-precision spatial control of the foreign object generation process achieved, but the physical rationality of the generated foreign object is also significantly improved.

The invention relates to the technical field of cloud AR real-time rendering, and discloses a self-adaptive resolution cloud AR real-time rendering and special effect processing method. The method comprises the steps of collecting network bandwidth and GPU utilization rate data, generating a resolution adjustment parameter through a multi-factor weighted decision model, and calculating a rendering resolution target value in real time by using a PID controller. Performing three-dimensional space division on the AR scene based on the target value, and dividing the AR scene into a static area and a dynamic area; the static region is pre-rendered and cached, and the dynamic region is rendered in real time. And synthesizing the pre-rendering cache and the real-time rendering result, performing low-delay coding, marking a transmission priority, generating a coded data stream, and sending the coded data stream to a terminal. And monitoring the network state and the equipment performance in real time, dynamically adjusting the weight of the decision model, and updating the rendering resolution and the coding parameters. According to the invention, adaptive balance between the rendering quality and the resource consumption is realized, and the efficiency of cloud AR rendering and the user experience are improved.

The application relates to the technical field of cloud AR real-time rendering, and discloses a cloud AR real-time rendering and special effect processing method with adaptive resolution. The method comprises the following steps: collecting network bandwidth and GPU utilization data, generating a resolution adjustment parameter through a multi-factor weighted decision model, and using a PID controller to calculate a rendering resolution target value in real time. Based on the target value, an AR scene is divided in three-dimensional space into static regions and dynamic regions; the static regions are pre-rendered and cached, and the dynamic regions are real-time rendered. The pre-rendering cache and real-time rendering results are synthesized, encoded with low delay and marked with transmission priority, and then encoded data streams are generated and sent to terminals. Network status and device performance are monitored in real time, the weights of the decision model are dynamically adjusted, and the rendering resolution and encoding parameters are updated. The application realizes adaptive balance between rendering quality and resource consumption, and improves the efficiency and user experience of cloud AR rendering.

The application provides a kind of aircraft view fusion method and system and anti-collision early warning method and system, belong to aircraft technical field, the aircraft view fusion method includes generating synchronous radar data and synchronous inertial navigation data in the process of aircraft flight;Synchronous radar data is filtered by background noise power dynamic threshold false alarm point, and effective radar data is formed;The synchronous inertial navigation data corresponding to the time stamp of effective radar data is obtained as effective inertial navigation data, and the initial terrain grid is constructed using effective radar data and effective inertial navigation data;The initial terrain grid is processed by double-line interpolation and adaptive resolution, and the optimized terrain grid is generated, the obstacles in the optimized terrain grid are updated and rendered, and the view fusion of the aircraft is completed.The application removes noise interference in the original radar data by false alarm point filtering processing, improves the accuracy and reliability of radar data.

The invention provides an automatic parking path planning method, device and equipment. The method comprises the following steps: acquiring obstacle information of a parking environment and calculating a distance field; dynamically allocating spatial resolutions to different areas based on the distance field so as to construct a self-adaptive resolution occupancy map; and collision detection information is provided for automatic parking path planning according to the map. According to the method, the distance field drives the map resolution to continuously and adaptively change, so that the overall data volume and calculation load of the map are remarkably reduced while the detection precision of the area near the obstacle is guaranteed, the problem of misalignment of collision risk assessment caused by sudden change of the resolution in a multi-resolution map is avoided, and the accuracy of collision risk assessment is improved. Therefore, the cooperative improvement of the path planning precision and the calculation efficiency is realized.

This disclosure relates generally to video coding and particularly to methods and systems for providing signaling schemes for jointly coding of motion vector difference with adaptive resolution in compound-reference inter-prediction. An example method for processing a current video block of a video stream is disclosed. The method includes receiving the video stream; determining from the video stream whether joint motion vector difference (MVD) coding is applied to the current video block; determining from the video stream whether adaptive MVD pixel resolution is applied to the current video block; and decoding the current video block based on whether joint MVD coding and whether adaptive MVD pixel resolution are applied to the current video block.