52 results about "Visual interpretation" patented technology

The invention discloses a TBM tunneling geological condition intelligent identification method and system fused with multiple modes. The method and system are used for identifying various geological condition scenes faced by a shield tunneling machine TBM during construction. The method comprises the steps that muck images and parameter information data are collected based on a shield tunneling machine industrial camera system and a sensor network, and data preprocessing operation is carried out; based on the collected data information, an improved Swin Transform model is provided, and multi-modal input of muck images and TBM parameters is fused; an improved multi-modal fusion model Swinin-T-M is integrated on the basis of a Dempster Shafer theory, the limitation and instability of a single model are improved, an integrated multi-modal deep learning model ESwinin-T-M is built, and the recognition accuracy and robustness of an algorithm are improved; and a visual interpretation technology is utilized to analyze the learning rule and key features of the model. According to the method, intelligent and accurate recognition of complex geological conditions in the TBM tunneling working process is achieved, and a feasible scheme is provided for detection of the construction environment of the shield tunneling machine.

The invention discloses an intelligent automatic drawing system and method, and relates to the technical field of agricultural remote sensing monitoring. The method comprises the following steps: acquiring a multi-temporal remote sensing image of a farmland area, performing preprocessing and spatial alignment, dividing the farmland into spatial units, extracting time sequence spectrum and vegetation index characteristics of the spatial units, and forming a visual characteristic vector sequence; analyzing the sequence by using unsupervised clustering, and automatically identifying the time sequence evolution trajectory categories representing different growth states; and finally, the system automatically marks the space units which are judged to be abnormal categories on a map, generates early warning and outputs a farmland abnormity thematic map. The system comprises a data preprocessing module, a data preprocessing module, a track classification module and an abnormal graph generation module. According to the invention, full-automatic and intelligent dynamic monitoring and abnormity identification of the farmland growth state are realized, the limitation of traditional single-temporal analysis or artificial visual interpretation is overcome, and the monitoring efficiency and objectivity are significantly improved.

The system comprises a laminated 5 diagnostic strip including a backing layer (105), an absorbent pad layer (106) configured for controlled capillary transport of milk, and a reagent deposit layer (107) defining multiple spatially separated dry reagent zones. The reagent zones include a pH indicator zone (101), a leukocyte esterase detection zone (102), a hemoglobin detection zone (103), and a nitrite detection zone (104), each adapted 10 for operation in the milk matrix. A calibrated sample application point (108) receives a predefined volume of untreated milk and fluidically communicates the sample to the reagent zones via the absorbent pad layer. Colorimetric reactions develop in a result window area (109), enabling simultaneous visual interpretation without analytical instruments, typically within three minutes. The invention further 15 includes a moisture-controlled storage container comprising desiccant elements (204, 205), a reference color chart (208), and a calibrated pipette (209), ensuring reagent stability and standardized use.

The application discloses a line-shaped slope map spot extraction method and system with a width less than or equal to 1 grid and a storage medium, and relates to the field of land resource management. Firstly, spot points are generated in a slope classification map to obtain all boundary node coordinate information of a polygon, all coordinate point combinations are calculated according to the coordinate information, all points are compared with existing nodes, then rectangular spots with a side length of two grid widths or more are found and marked in the spot area, the above process is repeated until all spots are traversed, finally, spots with qualified rectangles are removed, and the remaining spots are line-shaped slope map spots with a width less than or equal to 1 grid. The application does not need to rely on complex polygon geometry width calculation, avoids large errors and low efficiency problems of manual visual interpretation, provides a reliable basis for subsequent merging of spots according to the average allocation principle, guarantees the quality of slope classification map results, and supports the scientificity and accuracy of natural resource management decision-making.

The invention discloses a multiplex nucleic acid detection kit for swine acute diarrhea pathogens and application of the multiplex nucleic acid detection kit. The kit comprises an RAA (Reverse Amplification) primer pair, a specific crRNA (Ribonucleic Acid), a Cas12a protein, an NEBbuffer r2.1 and an ssDNA (Deoxyribose Nucleic Acid) reporter molecule, wherein the RAA primer pair, the specific crRNA, the Cas12a protein, the NEBbuffer r2.1 and the ssDNA reporter molecule are used for detecting the porcine epidemic diarrhea virus (PEDV), the porcine rotavirus (PoRV) and the porcine sapelovirus (PSV). According to the method, reverse transcription, amplification and Cas12a detection processes are coupled through an integrated system or a continuous reaction mode, the whole process is completed under the condition of constant temperature (37 DEG C), thermal cycle equipment is not needed, and only a constant-temperature heating device is needed. After the reaction is finished, a result can be observed through a portable blue light detector or naked eyes, so that on-site real-time visual interpretation is realized.

A unified framework system and method for a computer implemented artificial intelligent assistant to perform multiple tasks for remote sensing, includes a task input field for receiving a task identity, a global image encoder configured to receive a remote-sensing image and a user query and encode patch-level tokens at a high resolution via interpolation positional encodings, an MLP adapter configured to receive the patch-level tokens and adapt the tokens to language space, and a large language model configured to generate natural language responses interleaved with corresponding object locations based on the language space and task specific prompts. The global image encoder further includes a region input field for receiving region location parameters. The framework system is configured to switch based on the input task identity between different types of remote sensing visual interpretation tasks.

This invention belongs to the field of molecular biology detection technology, specifically relating to a Nipah virus (NiV) nucleic acid detection kit combining RT-LAMP and immunochromatography, and its application. This kit combines the high sensitivity and specificity of loop-mediated isothermal amplification (LAMP) technology with the rapid and visual interpretation advantages of immunochromatographic test strips, enabling on-site, rapid, and accurate detection of NiV. The core components of this invention include: a set of specific primers designed for conserved regions of the NiV gene, an optimized RT-LAMP reaction system, and specially treated immunochromatographic test strips for detecting LAMP amplification products. This kit eliminates the need for reverse transcription of RNA virus nucleic acids and expensive thermal cycling equipment. The entire detection process is completed within 20–30 minutes, with a sensitivity of up to 300 copies / μL. This solution combines the high efficiency of RT-LAMP technology with an intuitive, equipment-free detection method, achieving truly rapid and accurate on-site diagnosis of NiV.

This invention discloses a method for assessing ecosystem service losses due to regional climate change. The region includes counties, cities, or important ecological function zones. The steps are as follows: S1. Collect remote sensing data and preprocess the data through band combination, geometric correction, and projection transformation. Select a standard color composite scheme, establish interpretation markers, and perform visual interpretation. S2. Calculate and statistically analyze the unit area yield value and equivalent factor of grain crops. The equivalent factor is 1 / 7 of the average market value of grain in the current year. S3. Statistically analyze land use types, calculate the magnitude of land use change, reflecting the changes in the total amount of different land use types and the overall trend of land use type changes. S4. Select vegetation cover as an indicator, and adjust the ecosystem service value based on two parameters: net primary productivity and vegetation cover. This invention is of great significance for formulating ecosystem management strategies under the background of future climate change.

The invention relates to the technical field of virus detection, in particular to a primer probe combination, a kit and a method for detecting G3 gene group pig bocavirus based on a constant-temperature RAA fluorescence method in a non-diagnostic purpose mode. Aiming at a conserved region of a PBoV G3 virus genome, a probe which is specially modified and designed is adopted and has a synergistic effect with an upstream primer as shown in SEQ ID NO.4 and a downstream primer as shown in SEQ ID NO.7. The primer probe combination is used for detecting the G3 gene group pig bocavirus by a constant-temperature RAA fluorescence method, the detection sensitivity and specificity can be remarkably improved, viral nucleic acid as low as 10.5 copies / mu L can be detected in an extremely short time, meanwhile, cross reaction with other pig bocavirus gene groups or common pig pathogens is effectively avoided, and the detection sensitivity and specificity are improved. And a dual detection mode of instrument quantification and blue light naked eye interpretation can be supported, so that the detection efficiency and the convenience of field application are effectively improved.

This invention relates to the field of intelligent remote sensing image recognition technology, and discloses a method for identifying ground features in multi-temporal multispectral images based on pixel image fingerprints. The method includes: calculating several feature index bands for the original multi-temporal multispectral images and integrating them into a structured data file; selecting sample points of various target ground features based on high-resolution reference images and visual interpretation to construct a training sample set; inputting the data into a deep autoencoder network for training, using the trained autoencoder for feature learning, generating a standard pixel image fingerprint database, and setting a high-confidence recognition threshold; traversing the structured data file to be identified, using the trained autoencoder for matching and recognition, and generating target ground feature recognition results; and outputting the target ground feature recognition product through reclassification mapping rules and configuration of visualization colors. This invention effectively solves the problem of identifying "different objects with the same spectrum" caused by the limited number of bands and insufficient spectral separability in multispectral remote sensing images.

The invention discloses a medical image diagnosis method, device and system and a storage medium, and the method comprises the steps: carrying out the preliminary diagnosis of a medical image through a medical context big language model, and generating a medical report; quantifying the credibility of each sentence in the medical report by using an uncertainty measurement algorithm to obtain a credibility score; dynamically marking report contents by using color gradient mapping, and highlighting high-risk diagnosis entries; generating a structured visual thermodynamic diagram for each sentence of report by utilizing an explainable method based on explicit attribution and an attention mechanism; a React-based front-end interaction framework is utilized to integrate visual interpretation and report editing functions, explainable diagnosis decision basis and risk entry early warning are provided, and doctors are assisted in visual and rapid targeted verification; a continuous learning framework is adopted, feedback of doctors is collected, and sentence-level correction and a flexible training strategy are introduced to continuously optimize diagnosis precision. According to the invention, the interpretability and continuous learning ability of medical image diagnosis are improved.

The present application relates to the field of railway engineering, and relates to a kind of extraction method of rock slope based on hyperspectral remote sensing image, the method includes obtaining hyperspectral remote sensing image;Linear stretching treatment is carried out based on hyperspectral remote sensing image, obtain first spectral image;Based on first spectral image, carry out hue analysis processing, obtain first classification chart, first classification chart includes the area meeting the hue feature of rock slope;Band analysis is used to process hyperspectral remote sensing image, obtain second spectral image;Based on first spectral image and second spectral image, processing is carried out, obtain second classification chart, second classification chart includes the area meeting the lithology brightness of rock slope;Hyperspectral remote sensing image is processed, obtain third classification chart;First classification chart, second classification chart and third classification chart are processed, obtain the range of rock slope, the present application solves the problem that traditional rock slope extraction method relies on artificial visual interpretation, low efficiency, realizes the accuracy of extracting rock slope range.

The invention discloses a multi-modal fusion traditional Chinese medicine obesity intelligent dialectical system and method, and belongs to the technical field of artificial intelligence medical treatment, and the system comprises a multi-modal data collection module, a manifold feature extraction module, a hierarchical attention manifold mechanism, a dialectical decision module, a visual interpretation module, and a system verification and optimization module. The system collects face, tongue and handprint image data of a patient, and feature manifolds of all modalities are constructed through manifold feature extraction; performing multi-scale decomposition on each feature manifold by using a hierarchical attention manifold mechanism, calculating an attention weight based on a manifold curvature, establishing a mapping relationship among different feature manifolds, and realizing knowledge migration; the syndrome differentiation decision-making module carries out obesity syndrome type discrimination based on the fused multi-modal feature manifold representation; and the visual interpretation module generates a decision path of a syndrome differentiation process, accurately models a nonlinear geometric structure of traditional Chinese medicine diagnosis characteristics by introducing a manifold theory in differential geometry, and realizes multi-modal fusion of face diagnosis, tongue diagnosis and handprint.

The invention discloses a building disaster-bearing body remote sensing recognition method based on angular point features, and relates to the technical field of building disaster-bearing body recognition, and the method comprises the following steps: S1, collecting high-resolution optical remote sensing images in a concentrated manner, and carrying out the preprocessing of the high-resolution optical remote sensing images; s2, extracting image angular points of the preprocessed high-resolution optical remote sensing image set; s3, determining newly added angular points according to the image angular points; and S4, generating an outer boundary according to the newly added angular points, and completing remote sensing identification of the building disaster-bearing body. According to the method, a ground object sample does not need to be manufactured to carry out model pre-training, the building disaster-bearing body can be recognized only by depending on changes of image corner point features, the omission phenomenon in manual visual interpretation is avoided, and the method can be suitable for two scenes of recognition of a newly-added building disaster-bearing body before a disaster and recognition of a damaged building disaster-bearing body after the disaster.

The invention discloses a lake surface temperature reconstruction and freeze-thaw phenology extraction method based on filtering-automatic identification. The method comprises the steps of obtaining thermal infrared remote sensing data of a target lake in historical time; dividing a freeze-thaw period and a non-freeze-thaw period of the target lake; for a non-freeze-thaw period, acquiring a lake surface temperature time sequence of the non-freeze-thaw period; aiming at the freezing and thawing period, performing nonlinear curve fitting, and determining a freezing starting day, a freezing ending day, an ablation starting day and an ablation ending day of the target lake; according to the method, the continuity and precision of the lake surface temperature time sequence in the non-freezing-thawing period are improved, so that the error of traditional data caused by cold deviation is effectively reduced; and four accurate phenological key dates, namely freezing start and end and ablation start and end, can be scientifically, automatically and quantitatively extracted according to the freezing and thawing period, so that the deviation and low efficiency of subjective visual interpretation are avoided.

The present application belongs to the field of biological detection, and particularly relates to a field visual rapid detection method for citrus Huanglongbing. The present application is based on MIRA-lateral flow test paper, takes the outer membrane protein (OMP) of Asian species of Huanglongbing bacteria of citrus as a target, designs and selects specific primers and probes, and prepares dry powder microspheres together with MIRA reaction enzymes, mixes the field rapid extraction plant DNA lysate, dry powder microspheres and Buffer microspheres in a closed chamber to react, adopts a miniature temperature control device to complete amplification under the condition of 42 DEG C for 20 minutes, and directly realizes visual interpretation of the associated test paper strip. The detection method has no cross reaction with other citrus pathogenic bacteria, the sensitivity test can detect the DNA concentration lower limit of 6.9x10 ‑3 ng / μL, the detection device is small in size and convenient to carry, the whole process does not need professional operation or precise instruments, the detection can be completed within 30 minutes, and the detection method is suitable for rapid detection of citrus Huanglongbing by grass-roots agricultural technicians, planting enterprises and fruit farmers.

The invention relates to the technical field of medical image intelligent processing and robot control, and discloses an interpretation intention-driven scanning robot control method, device, equipment and medium, and the method comprises the steps: obtaining a scanning image, motion state data and task interaction reference data, generating a visual feature set and an interpretation intention vector, and performing time sequence coding on the motion state data to obtain historical track representation, generating motion decision features based on cross-modal feature fusion, obtaining a high-level strategy instruction and a low-level control instruction through hierarchical decoding, and synthesizing a driving signal for controlling the scanning robot to execute scanning operation. According to the method, cross-modal fusion is carried out on the visual interpretation intention and the motion trail features, so that the robot can make an action decision based on the operation intention, blind scanning behaviors are reduced, the objectivity and stability of the scanning process are improved, and automatic scanning control more conforming to clinical interpretation logic is achieved.

The embodiment of the invention provides a space-time dual-channel mask interpretation method for a dynamic weighted graph neural network and related equipment, and belongs to the field of artificial intelligence and machine learning. The method comprises the following steps: acquiring a pre-trained dynamic graph neural network model, a dynamic weighted graph sequence and original prediction thereof; initializing a space importance mask matrix and a time importance mask vector; under the condition of fixed model parameters, the two masks are jointly adjusted through iterative optimization; wherein in each iteration, a gating signal is generated by using a mask, an original edge weight is modulated to construct a disturbance diagram, and disturbance prediction is obtained through model forward calculation; reversely optimizing the mask by minimizing a total loss function fused with fidelity loss, sparsity loss and time smoothness loss; and generating an importance interpretation tensor fusing the space-time dimension based on the optimized mask. According to the method, fine-grained, high-fidelity, smooth and visual interpretation of the dynamic weighted graph decision is realized through decoupled space-time mask collaborative learning and weight perception disturbance.

The invention discloses an illegal construction identification method based on aerial photograph image registration and difference target detection. The method comprises the following steps of high-precision time sequence aerial photograph image acquisition, precise image registration under POS constraint, multi-dimensional difference extraction and construction pattern spot screening, construction site feature target precise detection, and illegal construction judgment and compliance verification. According to the illegal construction identification method based on aerial photography image registration and difference target detection, automation of the whole process from image collection to alarm pushing is achieved, traditional manual patrol and visual interpretation are replaced, and the identification efficiency is improved; the alarm record comprises complete space, attribute and image parameter information, and an accurate and reliable law enforcement basis is provided for law enforcement officers; through the design of route overlapping degree control, dynamic threshold adjustment, multi-model compatibility and the like, the method is suitable for illegal construction supervision scenes of different scales and different terrains such as cities, villages and towns, remote areas and the like, and the universality is high.

The application discloses an erosion gully classification method and device, electronic equipment and a storage medium. The method comprises the following steps: acquiring satellite images of a research area, and performing geographical coordinate calibration on the satellite images; acquiring a digital elevation model of the research area based on the calibrated satellite images, and performing pretreatment on the digital elevation model, and generating a contour map according to the pretreated digital elevation model; dividing the research area into a plurality of small watersheds based on the contour map; performing visual interpretation operation on each small watershed, and performing erosion gully classification processing on the research area based on each small watershed after the visual interpretation operation and a preset classification criterion. The application solves the technical problem that the current erosion gully classification method cannot scientifically classify erosion gullies according to the development topography of the erosion gullies.

This invention relates to the field of computer-aided diagnostic technology and discloses a method and system for multi-scale feature mining and intelligent risk prediction of hemorrhagic transformation (HT) in CT images. The method includes: performing standardized preprocessing on acquired raw CT images to obtain standardized CT image data; using a pre-constructed deep learning feature extraction network, sequentially performing shallow feature extraction and multi-scale deep feature mining on the standardized CT image data to obtain deep, full-dimensional features characterizing the lesion area and the overall state of the brain parenchyma; inputting the deep, full-dimensional features into a risk prediction classifier to calculate the probability value characterizing the risk of hemorrhagic transformation after ischemic stroke; and generating an attention heatmap based on the attention mechanism of the deep learning feature extraction network for visual interpretation of the prediction basis, and outputting the probability value and the attention heatmap, thereby realizing accurate and intelligent prediction and visual interpretation of HT risk in CT images.

The invention relates to the technical field of remote sensing, in particular to a cultivated land non-grain intelligent identification method based on multi-modal satellite remote sensing data and integrated learning. Comprising the following steps: remote sensing data preprocessing: processing a multi-source satellite image by using a cloud platform to generate a time-space aligned data set; constructing a sample data set: based on a high-resolution historical image, establishing a food crop positive and negative sample library through artificial visual interpretation; and multi-dimensional feature engineering: extracting spectral features and time sequence features from the image according to the growth cycle features of the food crops, and performing feature wave band screening. Drawing a grain planting probability: training a classification model based on an ensemble learning strategy and outputting a probability result of grain crop identification; constraint calibration based on actual data: screening grain planting pixels according to local statistical data; and non-grain discrimination: determining whether the image is non-grain according to the time sequence change characteristics of the pixels, and finally outputting a non-grain mapping result. According to the scheme, the non-grain identification of the cultivated land can be realized quickly and accurately.

The application discloses a thyroid ultrasound image tumor segmentation classification and refined rating method, which comprises constructing a thyroid nodule automatic segmentation classification model, and the model comprises a thyroid region prior guided feature enhancement network PGFE_Net, the thyroid region prior guided feature enhancement network PGFE_Net, an improved generative adversarial network GANs, a Resnet50 classification network and a gradient weighted class activation heat map Grad-CAM; wherein the PGFE_Net and the GANs extract the size, shape, edge characteristics and contrast information with the surrounding tissue of the nodule of the input image, generate a nodule mask, and perform a size prediction task; the nodule mask image and the original ultrasound image are combined, Resnet50 is used for feature discrimination and classification, accurate classification is carried out according to a TI-RADS standard, and the Grad-CAM provides visual interpretation of model decision-making. The application can effectively improve the detection and classification accuracy of thyroid nodule malignant tumors, realizes refined rating of the nodule, and provides auxiliary diagnosis support for clinicians.

The invention relates to an electronic nose system for honey authenticity identification based on deep learning. The electronic nose system comprises a smell collection device, a sensor array module, a data processing module, an artificial neural network classification module and an SHAP interpretability analysis module. According to the system, the characteristics of volatile organic compounds (VOCs) of honey are collected through a multi-channel smell sensor, and the characteristics are preprocessed and then input into an artificial neural network model to realize true and false identification. And introducing an SHAP method to carry out global and local visual interpretation on model output, and realizing self-calibration and anomaly detection based on the contribution degree of the sensor. The system can be integrated and deployed on a honey filling production line, online monitoring and linkage control are achieved, and the system is suitable for various honey varieties and application scenes. Compared with an existing method, the method has the advantages of being high in detection accuracy, high in interpretability, flexible in deployment and high in automation degree.

This invention discloses an early identification method for tobacco field flooding based on UAV hyperspectral imaging, aiming to solve the technical problem that existing technologies struggle to accurately and timely estimate the extent of tobacco field flooding damage in its early stages. This invention utilizes a UAV hyperspectral imagery of flood-affected tobacco fields, constructs Regions of Interest (ROIs) based on visual interpretation results, and classifies the tobacco fields into three post-disaster surface types using a spectral angle matching algorithm: flooded tobacco fields, normal tobacco fields, and soil. The calculation results of the spectral similarity coefficient and spectral angle show good classification performance, reflecting the spectral curve characteristics of these three land types. Finally, the overall classification accuracy is evaluated using a confusion matrix and the Kappa coefficient, verifying the feasibility of using this method for classifying and identifying tobacco field flooding, and providing a new solution for accurate and rapid assessment of tobacco field flooding.

This invention provides an adaptive design method and system for folding tables based on stability assessment, belonging to the field of furniture design technology. The method includes: physical principle-driven parametric design, automatically generating initial three-dimensional coordinates of the table legs based on input geometric parameters and physical constraints, and performing a pre-assessment of design quality; multi-dimensional quantitative stability assessment, constructing a comprehensive geometric and mechanical assessment model; robust hierarchical automatic optimization, employing a three-layer cascaded optimization strategy—direct optimization, multi-starting-point parallel optimization, and heuristic rule improvement—to iteratively adjust the spatial coordinates of the table legs and improve stability; and multi-view visualization analysis, automatically generating multi-dimensional charts such as support point distribution maps, height distribution maps, three-dimensional structure diagrams, and optimization comparison diagrams, providing a visual interpretation of design features, stability indicators, and optimization effects, and outputting a comprehensive report. This invention can provide key technical support for intelligent furniture manufacturing and personalized customization.

The application discloses an enzymatic colorimetric encoding-decoding system and application for multiple biomarker weighted detection, and belongs to the field of biomedical detection. The enzymatic colorimetric encoding-decoding system can realize the mapping relationship between color and multidimensional biomarkers through direct visual interpretation or simple spectrum analysis. The application further provides an enzymatic colorimetric encoding-decoding method for multiple biomarker weighted detection. By integrating enzyme-catalyzed multicolor reaction, nucleic acid-driven signal amplification and optical encoding-decoding mechanism, the multidimensional biomarker information is converted into quantifiable and visually analyzable optical signals, realizing single-tube multidimensional joint detection, high information density output and rapid disease state discrimination. The application has been successfully applied to early screening of pancreatic cancer, and can be extended to dynamic monitoring of multiple markers of other cancers and infectious diseases in the future, providing an efficient and low-cost solution for precision medicine and digital medicine.

This invention discloses a method, device, and medium for constructing a dynamic sample database for a large-scale remote sensing interpretation model. It relates to the field of remote sensing image sample database construction technology. The method includes: constructing a land cover sample knowledge graph to clarify the interpretation marker information of each land cover sample; constructing a standardized land cover sample database and storing land cover sample labels by category; initializing land cover segmentation masks and labels based on the SAM3 visual interpretation model, and performing hierarchical quality checks on the land cover segmentation masks and labels; dynamically iteratively updating the database, retrieving land cover samples from the database to generate a training set, and fine-tuning or optimizing the interpretation model. This invention constructs a knowledge graph adapted to the large-scale remote sensing interpretation model, forming a unified and scalable category system and interpretation marker information. This system can adapt to different regions and task requirements, providing a unified semantic foundation for the training and validation of the large-scale remote sensing interpretation model and solving the problem of inconsistent category systems.

The application discloses a kind of small target detection model visual explainability method, comprising: S1. deep feature map is obtained by CNN forward transmission, and feature map corresponding gradient is obtained by specific class target classification score back propagation;S2. after being processed based on sequencing grouping, mask containing high fine-grained local sensitivity information is obtained by feature image pixel level gradient weighting;S3. fuzzy mask occlusion image is input into CNN, and global decision contribution is obtained by containing N times integral sampling forward transmission, and contribution importance score is calculated and mean is taken as group importance score;S4. class activation map is obtained by linear weighting to grouping mask, and importance score is used to weight mask, to obtain visual explanation chart.The application is based on global decision contribution and parallel computing grouping mask importance score, fuzzy boundary and integral method are carried out to processing to mask, improve the sensitivity and algorithm speed of global decision contribution to different inputs.