118 results about "Imaging interpretation" patented technology

The invention discloses a causal reasoning-based satellite-ground collaborative remote sensing image interpretation method and device, relates to the technical field of remote sensing image processing, and mainly aims to solve the problem that the inference requirement of environmental causes in a remote sensing image cannot be met in the prior art. Comprising the following steps: a ground end obtains initial interpretation features obtained by performing initial interpretation on multi-modal remote sensing data by a satellite-borne end, and interprets the initial interpretation features based on a first interpretation model of which model training is completed to obtain core interpretation features; determining causal variables based on the core interpretation features, and constructing a causal variable graph based on the causal variables; determining a cause label and an interpretation result of the label area according to the causal variable graph, and obtaining a verification result corresponding to the cause label and the interpretation result; and based on the knowledge distillation and the verification result, determining learning core features of the first interpretation model for updating training, so that a second interpretation model in the satellite-borne end coordinates the learning core features fed back based on the ground end.

The invention discloses a breast cancer recurrence risk prediction method and system based on multi-modal data missing interpolation and gene interpretability enhancement. The method comprises the following steps: firstly, dynamically generating and complementing features of a missing mode by matching a generative adversarial network with a mode missing mask matrix; then, a feature screening mechanism driven by gene information is introduced, through a multi-task learning network, image feature extraction is supervised by using a gene expression tag in a model training process, and image features highly associated with recurrence-related genes are screened out; and finally, fusing the complemented multi-modal time sequence characteristics by adopting Transform, and outputting a recurrence risk probability. According to the method, the robust prediction performance can be realized under the condition of data missing, and meanwhile, image interpretation with a molecular biology basis is provided for the feature screening process of the model, so that the reliability and clinical acceptability of the whole system are enhanced.

The invention discloses a rain and sewage pipeline defect detection method and system based on distributed optical fibers. The method comprises the following steps: 1, laying sensing optical fibers in a pipeline; 2, calibrating an excitation position; 3, collecting and demodulating optical fiber vibration data; 4, performing defect identification and defect positioning on data processed by an intelligent algorithm; vibration signals are sent out through the excitation device, signal collection is carried out in combination with the distributed optical fiber sensing module, efficient recognition and accurate positioning of pipeline defects are achieved in cooperation with an intelligent detection algorithm, and hidden diseases which cannot be perceived by naked eyes can be recognized. By means of an intelligent defect recognition algorithm, signals collected by the optical fiber sensing module are deeply analyzed, the positions and the disease degrees of the defects in the pipeline can be automatically judged, dependence on manual image interpretation is avoided, and automation and intelligentization of pipeline defect detection are achieved. The system has the advantages of being high in adaptability, wide in coverage range, efficient, accurate, high in intelligent processing capacity, low in cost, high in efficiency and the like.

The invention discloses a multi-mode element learning remote sensing reconnaissance target identification method guided by a detection mechanism, and belongs to the field of computer vision and remote sensing image processing. The core of the method is as follows: multi-modal feature extraction and fusion guided by a detection mechanism: aiming at the electromagnetic scattering characteristic of an SAR image, the geometric texture characteristic of a visible light image and the thermal radiation characteristic of an infrared image, respectively designing a special feature extraction network, and introducing a cross-modal fusion module based on an attention mechanism, feature enhancement expression of physical information complementation is realized; according to the meta learning training normal form, an optimization-based meta learning device is constructed, and training is performed on a large number of multi-modal remote sensing task sets, so that the model obtains the capability of quickly learning new target categories from a small number of samples. According to the method, a physical detection mechanism and data-driven meta learning are combined, and the problems that a traditional method is low in recognition precision and poor in generalization ability in a complex reconnaissance scene with scarce samples, variable target types and difficulty in SAR image interpretation are effectively solved.

The invention relates to a remote sensing visual language large model training method and device based on unified reinforcement learning, and belongs to the technical field of artificial intelligence and remote sensing image processing, and the method comprises the steps: carrying out the preprocessing of an input remote sensing image and a text instruction, and extracting visual features and text features; performing modal alignment, and inputting the modal alignment result into a pre-trained large language model for supervised instruction fine adjustment to obtain a basic model; constructing a multi-dimensional deterministic unified reward module based on a truth value; and performing enhanced fine tuning on the basic model, calculating a reward value output in the group by using a unified reward module, and updating model parameters based on relative advantages to obtain an optimized remote sensing visual language large model. According to the method, a deterministic reward module is adopted, a value network is abandoned through group relative strategy optimization, the calculation cost is reduced, task-level indexes are directly optimized through a multi-dimensional unified reward function, and the precision and output normalization of remote sensing image interpretation are improved.

The invention provides a remote sensing image interpretation method based on large model reasoning and tool enhancement, which comprises the following steps: generating an initial reasoning node of a reasoning tree by using a large language model as an active decision maker based on system cue words and input remote sensing image features; based on the current inference tree state, the active decision maker executes inference and generates a new inference node, so that the inference tree is expanded; based on node scores, adopting a pruning strategy to eliminate reasoning paths with low scores; repeating the steps until a reasoning termination condition is met, and selecting a path with the highest accumulated score from the reasoning tree as an optimal reasoning path; and key information is extracted and organized into a structured intelligence report to be output. According to the method, a complete interpretation chain of'thinking-tool-observation 'is explicitly modeled through a tree reasoning mechanism, so that the generation logic of each intelligence conclusion is clear and visible, and the problems of trust crisis and insufficient interpretability caused by'black box' decision of a traditional deep learning model are solved.

The invention provides a graph structure guided remote sensing image interpretation method and system, a terminal and a medium, and the method comprises the steps: carrying out the interpretation through employing a visual language model based on an obtained remote sensing image and expert priori knowledge; the model training process comprises the steps of obtaining a graph structure based on a vector graph corresponding to a remote sensing image, and rapidly synthesizing a large number of samples for training by randomly modifying the graph structure; model training is carried out based on the synthesized training remote sensing images and the corresponding training graph structures and training vector diagrams, graph structure features obtained by the training graph structures are adopted as training priori knowledge, a large number of image features obtained by the training remote sensing images are interpreted, and feedback optimization is carried out with the training vector diagrams as labels. The visual language model adopted by the invention can obtain the spatial information and semantic information of the remote sensing image, avoids inaccurate interpretation caused by spatial relationship deficiency, and has a better interpretation effect.

The invention discloses a visual prompt multi-modal large model for multi-source remote sensing image interpretation, which belongs to the technical field of crossing of remote sensing and computer vision and comprises a multi-modal content coding and integration module, a cross-domain first-stage fusion training module, a pixel level visual positioning module and a large language model. The model supports the interpretation of a remote sensing image after the remote sensing image is arbitrarily amplified and reduced, and has flexible multi-granularity vision and language interaction capability. According to the model, a large language model is used as an interface, and multi-modal content integration including multi-sensor images, visual prompts and text instructions is achieved. In addition, two types of space tasks of anaphora understanding and visual positioning are unified into a visual prompt learning framework, and comprehensive and flexible multi-granularity understanding of remote sensing data is promoted.

The invention discloses a farmland protection forest area windproof effect evaluation method based on a land-gas coupling model, and relates to the technical field of agricultural science, high-precision topographic data of a target farmland protection forest area is collected and preprocessed, and spatial distribution information of a farmland protection forest is obtained by combining remote sensing image interpretation; on the basis of topographic data and spatial distribution information, topographic relief and earth surface coverage features are analyzed, and a refined underlying surface classification system is constructed; and configuring high-resolution grid parameters of a WRF mode, and coupling a Noah-MP land surface process model. According to the farmland protection forest area wind-proof effect evaluation method based on the land-gas coupling model, through combination of high-resolution topographic data and a multi-source remote sensing image, topographic relief and surface heterogeneity of a farmland protection forest area can be accurately depicted, and a WRF mode is coupled with a Noah-MP land surface process model; fine simulation of complex terrains and various underlying surfaces is realized, the limitation of traditional uniform underlying surface hypothesis is overcome, and the accuracy of windproof effect evaluation is remarkably improved.

The invention discloses an image interpretation method and device applied to natural resource investigation and monitoring. The method comprises the following steps: acquiring a to-be-interpreted remote sensing image and a current service scene; determining an interpretation model according to the current service scene; splitting the remote sensing image to be interpreted to obtain a plurality of remote sensing image slices; for each remote sensing image slice, inputting the remote sensing image slice into an interpretation model to obtain an interpretation result of the remote sensing image slice; and determining an interpretation result of the to-be-interpreted remote sensing image based on the interpretation result of the remote sensing image slice. The interpretation model can process remote sensing images in a unified mode, dependence on individual experience is eliminated, and it is ensured that interpretation results are stable and consistent. Moreover, the image is split into slices for parallel processing, so that the interpretation efficiency is greatly improved, and the time and cost of manual operation are reduced. The standardized algorithm of the model reduces the risk of misjudgment and missed judgment, can cope with complex scenes and detail requirements, and meets the increasingly refined monitoring requirements.

The invention provides a provincial domain level remote sensing AI big data natural resource intelligent monitoring system. The system comprises a data acquisition module, a data processing module, an intelligent interpretation module, a result verification module, a management application module and a communication interface module. The intelligent interpretation module integrates Segform semantic segmentation and a BIT change detection model, large-scale parallel interpretation is achieved by combining Redis cache management, celery task scheduling and MySQL process control under a GPU cluster architecture, batch image interpretation in a provincial domain range can be completed in a day-level period, linkage of interior work recheck and field work check is established through the achievement verification module, and large-scale parallel interpretation is achieved. And the management application module realizes the dynamic updating, thematic storage and cross-department sharing of the result, can quickly generate a change pattern spot distribution map, a cultivated land non-agrochemical and non-grain statistical table and a law enforcement check result, greatly shortens the interpretation period, and meets the requirement of dynamic supervision.

This invention provides an intelligent detection method and system for PI defects in chip packaging based on multimodal images, mainly relating to the field of chip packaging defect detection technology. The method comprises the following steps: First, at least two modal images of the chip packaging site are acquired, including a color image mainly characterizing the surface morphology of the chip and a fluorescence image mainly characterizing the physical properties of the chip surface material; then, the color image and fluorescence image to be detected are input into a multimodal target detection model to perform defect detection and obtain preliminary detection results; finally, based on the preliminary detection results and combined with the grayscale feature analysis of the fluorescence image, intelligent posterior decision-making is performed to obtain the intelligent detection result of PI defects in chip packaging. This invention combines the joint detection of color and fluorescence images with intelligent posterior decision-making, solving the problems of inaccurate detection of abnormal PI thickness defects, high dependence on manual image interpretation, and difficulty in automatically distinguishing multiple types of PI defects in existing technologies.

The present application relates to a kind of image processing methods for security inspection equipment, belong to security inspection technical field, it is difficult to meet the needs of image interpretation of image interpretation for the development of security inspection service and the image provided is difficult to meet the needs of third party integrated system problem.Method comprising the following steps: when the X-ray image of security inspection equipment is formed, the column data of the X-ray image formed is numbered in turn, and the row package image to which each column data belongs is judged;For each row package image, match the corresponding terminal of judging image, and send the column data of each row package image to the corresponding terminal of judging image, and the terminal of judging image is displayed according to the column data number received row package image, and centralized interpretation is carried out;The corresponding ray original image of each row package image is obtained, and each ray original image is uploaded to centralized storage;When third party integrated system needs to carry out image processing on row package image, corresponding ray original image is downloaded in centralized storage to carry out corresponding image processing.

The invention discloses a lung disease diagnosis and grading method based on quantitative CT image omics and deep learning. According to the method, through multi-dimensional information integration and model optimization, two core tasks of disease judgment and disease condition grading can be completed at the same time, and comprehensive support is provided for clinical diagnosis and treatment. The model sets feature priorities by referring to clinical diagnosis logic in the training process, the judgment result and the grading standard are completely matched with clinical general specifications, and the model can be directly applied to diagnosis and treatment decision-making without secondary conversion of doctors. After multi-center clinical verification and iterative optimization, the stability and the accuracy of the model are fully guaranteed, subjective errors caused by manual film reading can be effectively reduced, the diagnosis efficiency can be improved, the chronic obstructive pulmonary disease screening capability of primary medical institutions can be remarkably improved, early diagnosis and early treatment of more patients can be helped, and the clinical application prospect is wide. Therefore, the morbidity and disability rate of diseases are reduced, a clinical management path is optimized, and the overall disease burden is relieved.

The invention discloses a modeling method for the incidence relation between a deformation field and the erosion amount of a black soil erosion gully. According to the method, firstly, monitoring points are selected in black soil erosion gully areas with different landforms, high-precision equipment is used for collecting landform and hydrological data, and soil sample properties are analyzed; analyzing the deformation field, generating a digital elevation model, and determining the characteristics of the deformation field in combination with image interpretation; the erosion amount is calculated by adopting an improved erosion model, the deformation field and erosion amount data are integrated, and an incidence relation model is established by applying multiple methods. The model is verified through multi-source data fusion, and after improvement and perfection according to the result, the method is applied to black soil erosion gully dynamic monitoring, risk assessment and treatment planning. The method is comprehensive in data acquisition and deep in analysis, the model is accurate and effective, a scientific basis can be provided for black soil resource protection and sustainable utilization, and the problem of black soil erosion can be solved.

The invention discloses a post-earthquake disaster sensing emergency path planning method based on reinforcement learning, and the method employs the real disaster information provided by a high-resolution remote sensing image, combines the reinforcement learning theory, models the remote sensing image interpretation information into a reinforcement learning environment, enables a disaster sensing agent to carry out the continuous trial and error interaction with the environment, and achieves the emergency path planning of the post-earthquake disaster sensing. Learning an optimal post-earthquake emergency path planning strategy; according to the method, the problem that rich remote sensing data cannot be fully utilized in an existing post-earthquake emergency path planning method and the problem that disaster situations are not fully considered in the path planning process can be solved, in addition, the proposed post-earthquake emergency response framework is beneficial to optimizing rescue resource allocation, and the rescue efficiency is improved; according to the method, the path covering more disaster centers can be effectively planned in the middle and short distance scene, the rescue efficiency is maximized, and casualties and economic losses caused by earthquake disasters are greatly reduced.

The invention provides a remote sensing image interpretation method and system based on a large language model, and the method comprises the steps: obtaining a sample set, and the data in the sample set comprises user query data and a remote sensing image; preprocessing the sample; user query data is input into a task planning module for intention analysis, and a task plan is formed; according to the task plan, a knowledge enhancement module is adopted to obtain entities, relationships, attributes and high-correlation text fragments related to the task plan from the structured local knowledge base, and a tool calling instruction is generated based on the entities, the relationships, the attributes and the high-correlation text fragments; a reflection module is adopted to update and optimize the tool calling instruction; controlling a remote sensing tool module according to the updated tool calling instruction, and interpreting the preprocessed remote sensing image to obtain an interpretation result; and a two-stage training normal form based on supervision fine tuning and reinforcement learning is adopted to improve the structured and generalizable instruction generation capability of the task planning module in the remote sensing multi-tool link. According to the method, higher consistency and repeatability can be kept in a complex task chain and a multi-tool set, and the occurrence rate of mistaken calling, redundant calling, parameter errors, format errors and illusion is remarkably reduced.

This application provides a remote sensing image interpretation method, apparatus, and storage medium. The method is used for image interpretation via voice, including: acquiring a voice signal; parsing the voice signal; extracting the image to be interpreted from the voice signal; selecting a matching interpretation algorithm from a pre-built algorithm pool based on the voice signal; and interpreting the image based on the determined interpretation algorithm to obtain the interpretation result. This application can perform image interpretation by directly acquiring the voice signal; that is, the purpose of image interpretation can be achieved through voice control, eliminating the need for multiple manual clicks by the user and effectively improving the intelligence level of remote sensing image interpretation.

The invention discloses an SAR imaging system and method based on radio monitoring and GMTI composite guidance, a medium and equipment, and the system comprises a wide area monitoring module which is configured to carry out the wide area monitoring of a radiation source target in a target area through radio monitoring equipment, and outputs radio monitoring data; the slow-moving detection module is configured to detect a slow-moving target in a target area by using the GMTI equipment to obtain slow-moving target detection information; the SAR imaging module is configured to determine the position information of an interested SAR imaging target area according to the frequency, distance, angle and slow motion target detection information in the radio monitoring data so as to plan the route of the carrier and formulate an imaging working mode, and high-resolution SAR image data are obtained. The high-resolution SAR image data improves the positioning discrimination capability of the platform on the radiation source information.

Systems and techniques are disclosed for determining discrepancies between conditions and parameters associated with an image. An image processing model may generate output indicating predicted values for an image using the image as input, while a text processing model may generate output indicating corresponding predicted values for an image using textual data associated with the image as input. A comparison of the output data may be performed to determine discrepancies between values. Discrepancies that are sufficiently significant and relevant may be reported for additional analysis.

The invention provides a scheduling decision-making method and system for tasks in a hospital, terminal equipment and a medium. The method comprises the following steps: acquiring multi-source heterogeneous data; performing text analysis on the medical record data and the image interpretation data, and extracting disease semantic nodes, symptom semantic nodes and treatment scheme nodes to obtain a semantic node set; extracting timestamp information based on a time sequence of the vital sign data, the operation schedule information and the bed resource data; when matched semantic nodes and timestamp information exist, generating a node pair set with aligned time sequences; constructing an initial knowledge graph; obtaining sign image feature data of a patient, performing cross-modal correlation analysis, and updating the initial knowledge graph according to a cross-modal correlation analysis result; and searching the historical task scheduling strategy to obtain a target task scheduling strategy. According to the method, the timeliness of task scheduling in the hospital is improved, scheduling delay caused by disjunction of resource information is avoided, and the deviation of time accuracy is reduced.

The application discloses a fine-grained representation remote sensing image interpretation method and system based on Shearlet prior knowledge, adopts Shearlet transformation to extract high-frequency sub-band information of preprocessed remote sensing image data; uses Laplace pyramid and shearlet filter to obtain high-frequency components and multi-direction high-frequency sub-bands after directional splitting; through a spatial attention mechanism, sub-bands of the same level are adaptively fused to generate a detail-enhanced feature map; high-frequency components extracted through Shearlet transformation are prior to multi-scale and multi-level weighted fusion of deep residual blocks obtained through a deep residual network; a trained detail-enhanced backbone and different detection heads of multiple tasks are used to output accurate remote sensing image analysis results; through combination of Shearlet transformation and an advanced deep learning architecture, the accuracy and efficiency of remote sensing image interpretation are significantly improved, meanwhile, good model generalization ability and practicability are maintained, and it is expected that the application will promote the development of remote sensing image processing in a more intelligent and automatic direction.

Systems, devices, and methods operate on a fluorescence image to ensure a white balance correctness of the fluorescence image; perform a color analysis of the fluorescence image, including: converting the fluorescence image to an HSV color space, comparing a saturation parameter of the first converted image to a first saturation threshold condition and comparing a value parameter of the first converted image to a first value threshold condition, wherein the first value and saturation threshold conditions are based on a correlation between a first color and the first bacterial fluorescence signature, and identifying areas of the fluorescence image where the saturation parameter satisfies the first saturation threshold condition and the first value parameter satisfies the value threshold condition, discard the identified areas which are smaller than a size threshold, and outline the remaining identified areas on the fluorescence image, thereby to generate an overlay image.

A processor is configured to acquire a plurality of CT images derived by reconstructing a plurality of pieces of projection data having different energy levels, which are acquired by imaging a subject at a plurality of timings, at a plurality of predetermined energy levels, derive visibility information related to visibility of a target region included in each of the plurality of CT images, and determine an image interpretation energy level for interpreting the CT image among the plurality of energy levels based on the visibility information.

The application discloses a kind of coral reef area benthic biomass evaluation method, belong to marine ecological monitoring and evaluation technical field.The present application includes reef sample collection, reef coverage determination, reef effective attachment area measurement, benthic separation and collection, biological sample moisture treatment and weight weighing and six steps of biomass evaluation calculation.The present application is complete reef sample by stratified random sampling, and reef coverage is obtained using sample belt image interpretation, and effective attachment area is determined by measuring the length, width and height parameters of reef or three-dimensional laser point cloud, and the wet weight of species is weighed after separation, classification and standardization dehydration, and the benthic species biomass per unit area is calculated by combining reef coverage, species wet weight and effective attachment area, and the total benthic biomass per unit area and benthic species density can be further calculated.The present application is mainly used for benthic biomass evaluation in different habitat types of coral reef area, and can improve the accuracy and repeatability of evaluation results.

The application discloses a low-contrast image joint enhancement and super-resolution reconstruction method, which comprises the following steps: firstly, performing multi-scale Gaussian filtering on the low-contrast image; secondly, expanding the imaging dynamic range of the quotient image by using an S function; thirdly, performing color correction on the expansion result to obtain a contrast-improved image; and finally, optimizing the resolution and detail information of the contrast-improved image by a blind super-resolution reconstruction model constrained by a total variation to obtain a detail information and resolution improvement result. The low-contrast image joint enhancement and super-resolution reconstruction can help to overcome the influence of image adverse factors, such as low contrast / resolution and fuzzy details, and lay a foundation for image feature extraction and image interpretation.

The invention relates to the field of remote sensing image processing, and discloses a remote sensing image interpretation method and system based on cross-modal feature fusion. According to the method, a large language model serves as a teacher network, high-level semantic information extraction is carried out on preprocessed text modal data based on JSON Schema constraint to obtain structured semantic features, and then the structured semantic features are converted into teacher feature vectors through a feature encoder; performing feature extraction on the preprocessed visual modal data through a lightweight model included in the target student network to obtain a student feature vector, generating a gating weight based on the teacher feature vector through a gating fusion module, and performing bidirectional dynamic modulation on the student feature vector and the projected teacher feature vector based on the gating weight; semantic enhancement fusion features are obtained; and finally, based on the semantic enhancement fusion features, interpreting the remote sensing image through an interpretation layer in the target student network to obtain an interpretation result, so that the precision and robustness of multi-modal interpretation of the remote sensing image can be improved.

The invention belongs to the technical field of remote sensing image interpretation, and particularly relates to a remote sensing image segmentation method based on a wavelet frequency domain multi-modal fusion network. Comprising the following steps: step 1, building a frequency domain fusion module; step 2, building a spatial frequency domain wavelet attention fusion module; step 3, building a frequency enhancement fusion module; step 4, building a wavelet frequency domain multi-modal fusion network; according to the method, the problems of redundant information accumulation, insufficient feature expression ability and difficulty in effectively mining complementary information among modals caused by large modal difference in the existing spatial domain fusion method can be solved.

The invention discloses a full-view slice image interpretation method. Reasoning the full-view slice image through a trained neural network model to obtain first classification information used for representing a global classification result of the full-view slice image and second classification information used for representing a classification result of each region of interest in the full-view slice image, a splicing layer, a first branch model and a second branch model; and fusing the first classification information and the second classification information output by the trained neural network model to obtain an interpretation result of the full-view slice image. According to the embodiment of the invention, the accuracy of total-image grading interpretation is improved.

The invention discloses a river-related construction project batch construction conformity evaluation method and device, and the method comprises the steps: converting administrative permission information into a standardized permission space range, obtaining a remote sensing image interpretation pattern spot, and carrying out the coordinate unification; thirdly, screening out a candidate permission set associated with the pattern spot from the permission space range data by utilizing spatial retrieval, and further calculating a spatial relationship between the pattern spot and each permission in the candidate permission set to generate an associated permission set; and finally, respectively calculating conformity from three dimensions of type, position and area, and obtaining a conformity weighted score according to a preset weight so as to determine a final batch building conformity evaluation result of the pattern spots. According to the process, through a mathematical model and a space calculation technology, an original process depending on manual judgment is converted into objective numerical calculation, the evaluation efficiency and accuracy are greatly improved, errors caused by subjective factors are reduced, and a scientific and efficient decision basis is provided for supervision departments.