90 results about "Image perception" patented technology

The invention provides an active train sensing system and method based on thunder and vision fusion. Comprising an acquisition module used for acquiring multi-path image data containing different focal length images and three-dimensional point cloud data corresponding to the front of a train; the time synchronization module is used for generating synchronous image data and synchronous three-dimensional point cloud data; the coordinate transformation module is used for establishing a spatial transformation relation between the three-dimensional point cloud coordinate system and the image coordinate system; the image sensing module is used for inputting the synchronous image data and the three-dimensional point cloud data into the target segmentation network to obtain track coordinates; the point cloud sensing module is used for obtaining three-dimensional point cloud data in front of the train; and the thunder-vision fusion module is used for distributing an accurate center representative coordinate for each obstacle, generating structured obstacle information and outputting the structured obstacle information to a vehicle-mounted train control system. The multi-source sensing information utilization rate can be improved, the track obstacle recognition and three-dimensional positioning precision is improved, and the system sensing stability in a complex scene is enhanced.

The invention discloses an extensible reconstruction method for a super-large scene based on 3D (three-dimensional) Gaussian Splitting. The balance of the number of sub-block images is ensured through image perception self-adaptive blocking; a low-contribution view is removed through SSIM pruning; a priori depth is obtained based on virtual stereo rendering, a confidence mask is generated through multi-view verification, boundary consistency loss is constructed, and Gaussian densification is driven; and finally, carrying out TSDF fusion to obtain a seamless grid. Experiments show that reconstruction is completed in a 2.7 km scene within 6 h, MatrixCity-Aerial F1 reaches 0.784 and is improved by 41% compared with City GaussianV2, precision, efficiency and boundary consistency are considered, and the method is suitable for being applied to oversized scenes such as digital twin cities and automatic driving.

The invention provides a multi-source heterogeneous traffic light data fusion optimization method based on vehicle and road cloud, and the method comprises the following steps: S01, traffic light data collection: carrying out the collection of the traffic light data in three modes, the method comprises the following steps: S02, carrying out image perception and identification on traffic lights through a camera, obtaining traffic light data of an intersection through V2X short-distance communication, and obtaining traffic light data through cloud traffic light data service, S02, carrying out data cleaning, S03, carrying out multi-source data fusion, and S04, carrying out HMI data display. More than 99.9% of service availability is realized; the average delay is controlled within 100ms; and city-level scale deployment is supported. Under the extreme condition, if the communication delay fluctuation is large, the phenomenon of transient display delay or second skipping may occur, but the system can still ensure the basic service continuity through a data compensation mechanism.

The invention provides an endoscope image sensing method and system based on self-supervised learning, and the method comprises the steps: constructing and training an endoscope image sensing model, obtaining a real-time endoscope image sequence pair, inputting the real-time endoscope image sequence pair into the endoscope image sensing model for sensing, outputting a predicted retracement endoscope image sequence, and outputting a predicted retracement endoscope image sequence. An image sequence matching mechanism based on deep learning is adopted, the predicted retracement endoscope image sequence and the real-time retracement endoscope image sequence are matched, difference evaluation and screening are carried out on the predicted retracement endoscope image sequence and the real-time retracement endoscope image sequence after matching, and a corresponding uncertainty evaluation result is generated; and performing interpretability prompting based on the uncertainty evaluation result, and optimizing a corresponding module of the endoscope image sensing system according to a feedback result of the interpretability prompting, thereby realizing efficient and accurate identification of the abnormal region in the endoscope image.

A system and a method for medical image processing are provided. The method includes: obtaining a first medical image; obtaining a trained image perception restoration model; the trained image perception restoration model includes an image quality perception sub-model and an image restoration sub-model; and inputting the first medical image into the trained image perception restoration model to obtain a second medical image. The image quality perception sub-model is configured to determine either or both of a first quality evaluation value of the first medical image and a second quality evaluation value of the second medical image, the image restoration sub-model is configured to determine the second medical image, and the quality of the second medical image is higher than that of the first medical image.

The embodiment of the invention provides a multi-modal problem solving method and a multi-modal problem solving system.The multi-modal problem solving method comprises the steps that an input image is obtained, and the input image comprises a problem text and a problem graph; analyzing the question text to obtain a question analysis result and a target graphic element; performing vector graphic marking on the target graphic element in the topic graphic to obtain a vector graphic marking result; and based on a vector graphic marking result, reasoning a question analysis result to obtain an answer text. Attention on image details is improved in a vector graphic marking mode, image perception precision is improved, a vector graphic marking result can serve as an intermediate representation in a processing process, interpretability of a problem solving process and understanding of a verification method on the image details are enhanced, and a problem analysis result is reasoned based on the vector graphic marking result, so that the accuracy of image perception is improved. Direct perception of possible errors can be avoided, and the problem solving accuracy is improved.

The application provides a panoramic image splicing system and method for weak light enhancement and color temperature correction. The system mainly comprises a multi-thread control module for overall controlling the execution sequence of each module; a video reading and image correction module responsible for reading four camera images and performing image correction operations; a parameter initialization module for processing images, performing key point detection and matching, and calculating the rotation matrix between images and estimating camera parameters; an image projection module for projecting each image onto a cylindrical surface according to the rotation matrix and camera parameters; an optimal seam line finding module for calculating the optimal seam line of each overlapping area; an image fusion module for fusing images according to the optimal seam line and a multi-band fusion method to obtain a panoramic image; and a panoramic image display module responsible for displaying the panoramic image. The system can realize efficient and accurate panoramic image splicing and effectively handle weak light and color temperature problems to improve image perception.

The invention discloses an autonomous representation processing method, system and device and a storage medium, and relates to the technical field of artificial intelligence, and the method comprises the steps: receiving a representation demand input by a user; in combination with a large language model and a preset thinking chain reasoning mechanism, analyzing and disassembling the representation demand to obtain a task process; executing imaging and positioning operation according to the task process, and obtaining imaging data; performing target object instance extraction and attribute calculation on the imaging data to generate an instance-level structured object; characterization conclusions and / or decision information or physical execution instructions are formed based on the instance-level structured object. According to the method, task autonomous planning is realized by using a large model, and automatic positioning, screening, measurement and statistics are supported by parallel visual analysis, so that the labor cost is remarkably reduced; the individual recognition precision in a complex scene is improved through dynamic process disassembly; an executable instruction is generated based on pixel-physical mapping, an image is opened to perceive a physical operation closed loop, and the problem of cross-device collaboration is effectively solved.

The application provides an intelligent image processing system and method for a car based on a perception algorithm model, and the method comprises the following steps: S1. spatio-temporal reference double anchoring and dynamic intrinsic extrinsic parameter calibration of a vehicle-mounted image acquisition node; S2. multi-node image heterogeneous domain normalization and adaptive preprocessing based on a self-adaptive kernel regression non-local mean denoising algorithm; S3. hierarchical image feature extraction and semantic anchoring based on a graph neural network dynamic feature interaction network; S4. cross-node and cross-frame feature mutual checking and pseudo-feature elimination; S5. full-scene semantic completion and dynamic target trajectory prediction based on a variational autoencoder trajectory prediction model; S6. dynamic lightweight adaptation and algorithm power adaptive scheduling of the perception algorithm model; and S7. risk scene grading identification and image targeted enhancement output based on semantics and trajectories. The application provides stable, accurate and efficient vehicle-mounted image perception support for intelligent driving of a car, and improves the safety and adaptability of environmental perception of intelligent driving.

The utility model relates to railway vehicle detection technical field, concretely relates to a kind of EMU high-voltage cable terminal acoustic image detection device, including acoustic image acquisition unit and acoustic image analysis unit, both constitute data connection signal transmission, wherein the acoustic image acquisition unit includes microphone array, camera, the microphone array is used for the acoustic perception of high-voltage cable, the camera is used for the image perception of high-voltage cable;The microphone array includes the MEMS type sensor combination of several array elements, and the arrangement of several array elements MEMS type sensor uses the irregular sparse array based on multi-arm spiral array.The utility model has the advantages that: realize EMU high-voltage cable terminal partial discharge abnormal sound non-contact detection, accurately master EMU high-voltage cable use condition and life cycle, scientific guidance roof high-voltage cable overhaul and replacement, can effectively guarantee EMU operation safety and stability, substantially reduce EMU operation and maintenance cost expenditure.

The application relates to an insulator image segmentation method based on an unmanned aerial vehicle infrared enhanced image, and relates to the technical field of image processing, and comprises the following steps: acquiring an infrared image sequence through an infrared enhanced camera to obtain a standard infrared image sequence, performing state recognition by using an image perception prior engine, and outputting a predicted insulator attention heat map and a predicted boundary confidence map; optimizing and adjusting an initial threshold segmentation algorithm and an initial region growing algorithm to obtain an adaptive threshold segmentation algorithm and an adaptive region growing algorithm; performing image segmentation on the standard infrared image sequence to output an insulator image sequence, and performing early warning judgment on the insulator image sequence based on an adaptive early warning mechanism. The application solves the problem that the traditional insulator image segmentation method cannot effectively deal with the problems of low temperature contrast, large noise interference and complex background of the infrared image, so that the insulator segmentation is prone to false segmentation and missed segmentation, and the high-precision requirement of fault detection cannot be met.

This invention discloses a method for scene classification of remote sensing images with a small sample size, belonging to the field of image perception. The method first divides the remote sensing image scene classification dataset into training, validation, and test sets; then, it constructs scenario sets based on these datasets; next, it constructs a deep nearest neighbor neural network model based on an attention mechanism; then, it trains and validates the attention-based deep nearest neighbor neural network model based on the scenario sets; finally, it tests the trained attention-based deep nearest neighbor neural network model based on the scenario sets. This invention addresses the problem of complex scene backgrounds in remote sensing image scene classification by proposing a deep neural network model based on an attention mechanism. It introduces a scenario training method, uses deep local descriptors for feature representation, and introduces an attention mechanism to obtain scene category-related attention maps. The deep local descriptors are divided into scene-related and scene-independent parts. A metric module calculates the similarity between the local descriptors of the query image and the category, and a weighted sum is performed using the attention map, thus solving the problem of scene classification in small samples of remote sensing images.

The invention provides an image compression method and device, electronic equipment and a readable medium, and relates to the technical field of image processing, and the image compression method comprises the steps: obtaining an input image and a target code rate; inputting the input image into the content feature analysis model to input a feature vector of the image, the feature vector being capable of reflecting a perception feature of the input image; and inputting the target code rate and the feature vector into a joint parameter decision model to generate a spatial scaling parameter and a quantization parameter adaptive to the input image and the target code rate. According to the embodiment of the invention, under the limitation of the target code rate, the details of the visual salient region are reserved preferentially according to the adaptive control of the image prediction optimal compression parameter, and the method is widely applied to application scenes such as extremely-low bandwidth image transmission, terminal front-end compression and Internet of Things image perception.

The application discloses a kind of corn earing process parameter on-line monitoring and intelligent control test bench, belong to the field of agricultural engineering technology, including frame, stem conveying mechanism is arranged in the middle of frame, earing mechanism is hinged on frame, image perception unit is fixedly arranged in the rear of frame, angle adjusting mechanism is hinged between frame and earing mechanism, angle adjusting mechanism is symmetrically arranged, one side of frame is provided with intelligent control platform, system can realize multi-parameter linkage regulation and control. Through central controller and host computer software communication, operator can set or adjust stem roller speed, earing plate gap adjusting mechanism gap, header frame angle, stem feeding speed and other parameters on man-machine interface, realize multi-factor combination control based on test demand, system can automatically record the working state data under different parameters, for subsequent data analysis and algorithm optimization.

The application discloses a quality detection method of a light field image; the method is as follows: 1, a measured light field image is decomposed to obtain a sub-aperture image sequence; 2, the sub-aperture image sequence is paired two by two to obtain multiple left-right view pairs; each left-right view pair generates a light field three-dimensional view; 3, a neural network for quality evaluation is constructed; and the neural network is used to generate a quality score S of the measured light field image. The application comprehensively considers the image perception experience and the image recognizability of the light field image, and can accurately analyze the quality of the light field image.

The invention discloses a control method and device for automatically positioning an intelligent logistics conveying line, and a medium, and relates to the technical field of intelligent logistics conveying, and the method comprises the steps: planning a target position and a current timestamp in a data package according to a task path; respectively extracting respective real-time data streams from a wireless ranging positioning mode, a radio frequency identification mode and an image perception mode, calculating the positioning confidence of each mode, performing multi-mode state fusion through an unscented Kalman filter, and generating a positioning fusion control data packet; the real-time data stream comprises real-time distance data, real-time RFID positioning data and real-time image positioning data; the transmission line is divided into a plurality of independent control areas by using the positioning fusion control data packet, event triggering conditions are set, and when any event triggering condition is met, the corresponding independent control area immediately executes a local MPC algorithm to generate a control parameter set; and the logistics conveying efficiency, reliability and automation level are improved.

The embodiment of the invention provides an unstructured data processing system based on artificial intelligence, and the unstructured data processing system based on artificial intelligence comprises multi-modal original data obtained through a distributed probe array which comprises a text scanning unit, an image perception unit and an audio and video synchronization unit; the characteristic smelting center is connected with the probe array and is provided with an anti-decoupling module and a cross-modal attention grafting module; the evolutionary engine is dynamically connected with the smelting center and comprises a periodic self-checking neural architecture searcher and a knowledge incubator based on analogy reasoning; and the system outputs a 128-dimensional composite feature code with a timestamp. Through the cross-modal fusion technology of the feature melting center, the semantic understanding depth which cannot be achieved by a traditional single-modal system is achieved, a dynamic evolution engine ensures that the processing precision is continuously optimized along with time, and the common industry problem of model performance degradation is overcome.

This invention provides a method and system for predicting brain diseases by fusing amplitude and phase information with graph-aware hybrid expert graph neural networks. It relates to the field of neuroimaging analysis technology. The method includes: acquiring raw resting-state functional magnetic resonance imaging (fMRI) data of a subject and corresponding brain disease category labels; preprocessing the raw resting-state fMRI data; extracting the mean oxygenation level dependent signal time series of multiple brain regions of the subject and using the mean oxygenation level dependent signal time series as the signal time series; constructing a functional connectivity matrix and a phase adjacency matrix based on the signal time series; constructing a training dataset; constructing a brain disease prediction model based on a two-branch hybrid expert graph neural network; training the brain disease prediction model using the training dataset; acquiring resting-state fMRI data; inputting the resting-state fMRI data into the trained brain disease prediction model for prediction, and outputting the brain disease prediction result.

The invention discloses an anomaly detection method and device for fusion of a visible light image and an infrared image, and relates to the technical field of artificial intelligence and big data or other related technical fields, and the method comprises the steps: obtaining a visible light image and an infrared image which are synchronous in time, and carrying out the preprocessing of the visible light image and the infrared image; inputting the preprocessed visible light image and infrared image into a registration model, and outputting the registered visible light image and infrared image; inputting the registered visible light image and infrared image into a multi-level feature extraction network, and outputting fusion features; and inputting the fusion feature into a target detection model, outputting a target detection result, and performing anomaly recognition based on the target detection result to obtain an anomaly detection result. According to the invention, the technical problem of low detection result accuracy of an anomaly detection mode based on single-mode image perception and recognition in the prior art is solved.

The invention relates to privacy protection, in particular to an AMI network-oriented portrait perception adaptive privacy protection method, which comprises the following steps: quantifying the importance of users through a fuzzy comprehensive evaluation algorithm, and providing different protection levels for different users according to the importance, the higher the importance is, the higher the protection level is; distributing a privacy budget for each user according to the protection level of the user; based on the distribution privacy budget, adopting a square wave mechanism to disturb data of a numerical value type in the original data of the user; on the basis of distribution of privacy budget, calculating a theoretical mean square error of generalized random response and optimal unary coding, and selecting a mode with a relatively small error to disturb discrete type data in the original data of the user; and after the disturbed data and the privacy budget thereof are uploaded to the server, the server recovers the user data with the same privacy budget by adopting a hierarchical EMS algorithm. According to the method, the non-uniform transfer matrix and the virtual grid projection are constructed, so that seamless fusion of different privacy budgets and different discrete granularity data is realized.

The invention relates to the technical field of stereoscopic video signal processing and interactive television terminal energy efficiency management, and discloses an energy-saving interactive screen projection control system based on image perception, which is characterized in that a multi-channel load demand mapping unit analyzes an input signal flow into power demand space distribution of each node of a load network; the gradient threshold power isolation logic unit defines a power supply cut-off region according to the power gradient mutation characteristics; the distributed current injection regulator carries out truncation zero level operation on the load nodes in the cut-off region, and carries out time domain damping modulation on the non-cut-off region according to migration vector data. By constructing a differential power supply protocol of a physical level, the dynamic voltage stabilization precision of the core load nodes is ensured, and meanwhile, time domain damping modulation is carried out on the non-cut-off region according to the migration vector data. Redundant electric energy input of a low-demand background area is cut off, and the energy utilization efficiency of the interactive stereoscopic display system and the stability of video picture depth reconstruction are improved.

The present disclosure provides a system and a method for perceiving an object in a scene. The method comprises collecting features of a first radar image of the scene captured from a first sensor and a second radar image of the scene captured from a second sensor, each of the first radar image and the second radar image includes depth data. The method further comprises processing selected features of the collected features with a transformer neural network having a transformer architecture with self-attention over the selected features and cross-attention between object queries and the selected features to produce 2D+ embeddings of the object. The method further comprises processing the 2D+ embeddings with a detection neural network to perceive the object and produce an image of the scene with markings of the perceived object, and outputting the image of the scene with the markings of the perceived object.

The application discloses a multi-field coupling unmanned ship pollution tracing path method, comprising a small unmanned ship platform, wherein an environment perception component, an automatic power supply component and a control communication component are arranged on the small unmanned ship platform, image perception, hydrological and water quality perception environment perception components are carried, three-dimensional multi-element perception is realized, a central control module is interacted with a client, task assignment, instruction distribution, module cooperation and other whole-process automation are achieved, the multi-field coupling unmanned ship pollution tracing path method is independently modeled and coupled through a water body dynamic field, a pollution field and other fields, a scene is accurately described, dynamic changes of the pollution field are reflected, the tracing technology is scientifically realized, a multi-objective algorithm is established, a target function and a constraint condition are defined, and intelligence is realized through multi-objective optimization, and on the basis of a traditional algorithm, a penalty function and a local search strategy are introduced, applicability is enhanced, and high efficiency is realized.

The invention relates to the technical field of sensor integration, and discloses a system and method for integrating pulse neural network training in a sensor, the sensor is integrated with a CMOS sensing module, a pulse conversion circuit, a weight storage module, a neuron module and a weight updating module, and a clear and mappable circuit basis is provided for a sensing and calculation integrated architecture; the spiking neural network training realizes physical random number generation and random STDP learning process, and drives the spiking neural network to complete adaptive training and classification in the sensor; the CMOS sensing module deeply integrates two functions of image sensing and random number generation and realizes a combined technical characteristic of dynamic switching, so that sensor hardware can be flexibly reconstructed between a data source and an entropy source, and the utilization rate of hardware resources and the integration level of a sensor chip are greatly improved.

The invention relates to the technical field of intelligent control, in particular to a make-up water resin regeneration automatic control system and method, and the system comprises an interface recognition processing module, a liquid level stability judgment module, a state linkage triggering module, a path action reconstruction module and a regeneration process recovery module. According to the method, synchronous recognition of the resin boundary state and positioning analysis of an abnormal section are completed through linkage comparison of the image gray gradient and the liquid level change, reversible action node sequences with a correlation sequence are extracted by combining response offset conditions in a control path, valve control states are sequentially cancelled, feedback data are continuously tracked, and therefore the control accuracy of the valve is improved. A path backtracking and state repairing mechanism driven by image information is established, a process control mode integrating image perception, liquid level linkage, path comparison and control correction is constructed, and the execution continuity, the fault handling capacity, the process regulation and control stability and the coordination level of system operation of the regeneration process in an unexpected state are enhanced.

The invention discloses a high dynamic range display effect optimization method based on human visual characteristics, and the method comprises the steps: carrying out the standardized nonlinear coding and decoding and HDR post-processing of an SDR image, and obtaining the linear brightness distribution of an HDR image after the SDR image is converted; linear brightness distribution of the HDR image is used as input, EOTF and brightness normalization are carried out, a glare spread function filter convolution kernel is calculated, retinal HDR image perception simulation is carried out, a retinal contrast distribution result is obtained, an evaluation index of a contrast peak signal-to-noise ratio is established, and high dynamic range display effect optimization is carried out. Theoretical support is provided for the HDR adjustment strategy conforming to the human visual characteristics, and practical support is provided for improving the display quality of the display device.

The embodiment of the invention provides an image-based human body measurement method, device and equipment, and relates to the technical field of image processing. The method comprises the following steps: acquiring image perception information of a to-be-detected image; determining a human body measurement task strategy corresponding to the to-be-detected image based on the image perception information; and executing human body measurement operation based on the to-be-detected image according to the human body measurement task strategy to obtain a human body measurement result. According to the embodiment of the invention, the image perception information is acquired, and the human body measurement task strategy is generated according to the image perception information, so that human body measurement is performed on different types of images in a targeted manner, and the universality, flexibility and robustness of human body measurement are effectively improved.

The invention discloses a training method of an end-to-end visual motion model, a trajectory planning method and an application thereof to meet the requirements of high-level automatic driving application, and the training method comprises the steps: inputting a sample image at a T moment into a first model, and obtaining a perception task query and an image Token corresponding to an image perception result at the T moment; based on the perception task Query at the T moment and a second model, obtaining a predicted trajectory of the target object and a planned trajectory of the target vehicle at the T moment; based on the predicted trajectory of the target object at the T moment, the planned trajectory of the target vehicle and a third model, obtaining a predicted trajectory of the target vehicle at the T + 1 moment; based on the predicted trajectory of the target vehicle at the T + 1 moment, the planned trajectory of the target vehicle at the T moment, the image Token and a fourth model, obtaining a predicted image at the T + 1 moment; and executing back propagation of the model gradient based on the predicted image at the T + 1 moment so as to update parameters of the model and construct an end-to-end visual action model.

The invention relates to a control method of a mobile home environment adjusting robot based on user behavior perception. The mobile home environment adjusting robot comprises a mobile chassis module, an environment adjusting execution module, a multi-mode non-imaging perception module and a local intelligent control module. The control method comprises the following steps: S1, multi-modal non-image sensing; s2, behavior feature extraction and intention recognition; s3, service stationary point mapping and strategy generation; s4, path planning and user avoidance are carried out; and S5, executing autonomous movement and environment adjustment. According to the method, the user behavior and environment information is obtained in a non-image multi-mode sensing mode, an environment adjustment decision mechanism with the user behavior intention as the core is constructed in combination with indoor space semantic mapping, environment adjustment is converted into active and directional service from passive triggering, and active environment adjustment facing the user behavior intention is achieved.

The present application relates to the technical field of heavy rain early warning, in particular to a heavy rain terminal early warning method and system based on deep learning, which realizes consistent determination of image features such as the overturning state of a rain gauge and rain curtain, water marks and the like by introducing an image perception and semantic recognition mechanism, effectively excludes non-natural triggering factors (such as artificial water pouring), and improves the reliability of rainfall perception data. A locally deployable double-branch deep learning model is adopted to fuse multi-source data such as rainfall, temperature and humidity, and terrain factors, cross-attention mechanism is used for feature extraction and return period prediction, and a threshold correction coefficient is dynamically generated in combination with historical events to realize adaptive adjustment of the early warning standard and effectively reduce the risk of false positives and false negatives. The system supports low-power local operation and executes early warning decisions, effectively improving the timeliness and accuracy of local early warning response.