910 results about "Spectral image" patented technology

The invention relates to the technical field of image processing, in particular to an image acquisition and analysis method and system, and provides the following scheme: obtaining a visible light and near-infrared multispectral image, generating a spectral difference image, and performing weighted fusion to obtain a first image; segmenting a target region based on the fused saliency map, and calculating a pixel reflectance ratio; solving a color mapping matrix according to the reflectance ratio, and carrying out color correction on the target region to obtain a standardized feature image; and extracting characteristic parameters such as spectrums, colors and textures, inputting the characteristic parameters to a multi-branch convolutional neural network, fusing the characteristic parameters through an attention mechanism, and outputting a state classification result and a quantitative index. The cross-spectral imaging difference can be adaptively compensated, and the fusion precision and the analysis stability are improved.

The invention discloses a wharf shoreline terrain reconstruction method based on remote sensing data, and belongs to the technical field of three-dimensional terrain reconstruction. The method comprises the following steps: acquiring a multi-temporal multispectral remote sensing image, a synthetic aperture radar image, a laser radar point cloud and synchronous tidal water level data of a target wharf area; processing the multispectral image to generate an initial land and water segmentation map; carrying out tide correction by utilizing the tide water level data, fusing the multi-polarization characteristics of the synthetic aperture radar image, and correcting the ground feature type to obtain an accurate land and water boundary diagram; processing the laser radar point cloud to generate a digital surface model, extracting a shoreline contour with sub-pixel-level precision from the digital surface model, and endowing the shoreline contour with an elevation value; and superposing the shoreline contour with the elevation with the multispectral image to generate a three-dimensional terrain model. According to the method, the problems of low terrain reconstruction precision, large tide influence and insufficient automation degree of a traditional method in a complex wharf environment are solved, and high-precision and high-efficiency wharf shoreline three-dimensional automatic reconstruction is realized.

The invention discloses a regional termite detection method and system based on multispectral fusion, and the method comprises the steps: carrying a multi-mode spectrum collection device through a movable detection platform, and synchronously collecting a visible light image, a near-infrared hyperspectral image and Raman spectrum data of a to-be-detected region; preprocessing and feature extraction are carried out on the modal data, and the modal data are converted to a frequency domain to obtain frequency features; based on the physical and biochemical characteristics of the termites and the nests thereof, calculating the matching degree and the credibility of each modal feature, and performing adaptive weighted feature fusion according to the matching degree and the credibility to generate comprehensive spectral features; inputting the fusion features into a pre-trained termite identification and risk assessment model to realize precise identification, positioning and threat level assessment of termite individuals, termite paths and nests; and carrying out trajectory analysis on termite activities by combining a multi-target tracking algorithm, and giving out early warning based on identification and tracking results. According to the invention, early-stage, lossless and accurate detection and active early warning of termites are realized, and the detection efficiency and reliability are significantly improved.

The invention relates to a disease diagnosis and treatment method, system and device based on ear-nose-throat endoscope images and a medium, and the method comprises the steps: synchronously collecting dual-spectrum images through time sequence triggering, and solving the problem of shielding of an anatomical structure caused by mucus flow; a dynamic mucus displacement field is modeled through pixel gradient, and misjudgment of a traditional segmentation method on static lesions and dynamic secretions is eliminated; a deformable convolutional layer is adopted to correct the spatial offset of white light and a narrow-band image, and the mismatch of a multi-mode characteristic due to optical scattering is overcome; and finally, a real-time surgical navigation mark and a clinical treatment scheme are synchronously generated based on topological attributes of the focus probability graph, and a closed-loop link from image analysis to diagnosis and treatment decision is realized. According to the method, the functions of mucus interference suppression, cross-modal accurate registration and real-time diagnosis and treatment assistance are integrated in a breakthrough manner, and the focus recognition accuracy and clinical operation efficiency of the endoscope image are remarkably improved.

The invention relates to the field of optical nondestructive testing, and discloses a visualization method for detecting mango quality based on hyperspectrum, which comprises the following steps of: firstly, selecting mango samples, and simulating different mango ripening stages by dividing the mango samples into a plurality of groups of samples; acquiring a high-dimensional spectral image by using a hyperspectral imaging system, and measuring the hardness, the soluble solid content and the titratable acid content; correcting and extracting the high-dimensional spectral image to obtain and optimize the average reflection spectrum of the sample, and establishing a mango maturity classification model and a quality regression prediction model; performing characteristic wavelength contribution degree analysis and screening, and optimizing a quality regression prediction model by adopting a key wave band; the method is applied to each pixel point of a high-dimensional spectral image, a two-dimensional distribution diagram is generated, and mango quality visualization is achieved. The method solves the problems that in the prior art, rapid and lossless judgment of the mango ripening stage cannot be achieved, accurate prediction and visual distribution of key internal quality indexes cannot be achieved, and intelligent grading and quality control of picked fruits are not ideal.

The invention belongs to the technical field of surveying and mapping and geographic information processing, and discloses a geographic entity intelligent identification and reconstruction system based on multi-source surveying and mapping data. The system is composed of a multi-source data acquisition and preprocessing module, a cross-modal feature coding and fusion module, a structural atlas construction and spatial logical reasoning module, a deformable neural modeling module and a physical prior guided collaborative prediction and closed-loop optimization module. According to the method, a cross-modal feature coding and fusion module is arranged, and a modal attention mechanism is introduced to dynamically weight multi-source data, so that heterogeneous information such as a laser point cloud, an inclined image and a multispectral image is fused into a unified coding vector in a high-dimensional space; compared with feature extraction performed by using a static deep network in a comparison file, the method of the invention adopts a minimum residual function to perform modal weight training, has an adaptive feature integration capability, and effectively improves the accuracy of geographic entity recognition and the robustness of boundary segmentation in different scenes.

The invention discloses a hyperspectral image classification method for cross-domain small sample learning based on diffusion enhancement prototype knowledge distillation, and belongs to the technical field of hyperspectral image processing. The method comprises the following steps: extracting a neighborhood data cube, aligning spectrums, dividing a support set and a query set, and applying a mask and enhancing noise; executing domain adversarial denoising and reconstruction tasks, aligning feature distribution, and outputting a pre-training encoder; decoupling features, capturing spectrum-space global and local dependency relationships, and calculating similarity between a query set and a category prototype; constructing a distillation framework to realize knowledge migration; optimizing model parameters, and introducing a signal-to-noise ratio to enhance loss suppression noise; and performing feature extraction by using the optimized student model to generate a hyperspectral image classification result. According to the method, the problems of domain offset, intra-class feature dispersion, inter-class boundary fuzziness, noise interference and the like are solved, and the classification accuracy in a small sample scene is improved.

The invention relates to the technical field of precision agriculture and grape cultivation, and discloses a vineyard fertilization method based on unmanned aerial vehicle multi-source remote sensing data, and the method comprises the steps: obtaining a canopy multi-spectral image and three-dimensional structure data through an unmanned aerial vehicle; carrying out image preprocessing and finely extracting a grape canopy region; fusing the extracted spectral vegetation index, texture features and three-dimensional structure features, and constructing a multi-source feature vector; constructing and optimizing a nitrogen nutrition inversion model through a machine learning algorithm by utilizing actually measured nitrogen nutrition parameters; generating a nitrogen content distribution diagram based on a model inversion result, calculating the nitrogen deficiency amount and the recommended dressing pure nitrogen amount of each space unit in combination with critical nitrogen concentration diagnosis and target yield, and converting the nitrogen deficiency amount and the recommended dressing pure nitrogen amount into the use amount of a foliage spraying working solution; and generating a variable fertilization prescription map, and converting the variable fertilization prescription map into a nozzle flow control instruction executable by the unmanned aerial vehicle to realize on-demand accurate variable fertilization. According to the invention, closed-loop management from nitrogen nutrition monitoring to variable rate fertilization is realized, and the nitrogen fertilizer utilization efficiency and the fertilization accuracy are improved.

The invention discloses a reconstruction spectral imaging system based on an electromagnetic drive Fabry-Perot resonant cavity, and belongs to the field of spectral imaging. The system comprises an optical lens group, an electromagnetic driving device, a Fabry-Perot resonant cavity, a photoelectric detection array and a calculation unit, the cavity length of the Fabry-Perot resonant cavity can be adjusted through the electromagnetic driving device. The optical lens group receives light reflected or emitted by an external to-be-measured target scene and carries out collimation processing, and then the light enters the Fabry-Perot resonant cavity. The Fabry-Perot resonant cavity screens the incident light based on the resonance condition of the cavity length of the Fabry-Perot resonant cavity, and the screened light is emitted to the photoelectric detection array; the photoelectric detection array converts the incident light intensity into an electric signal and outputs the electric signal to the calculation unit; and a spectrum reconstruction model is built in the calculation unit, and a hyperspectral image of the to-be-measured target scene is obtained by using the spectrum reconstruction model based on the electric signal. The working range and the spectral imaging precision of the spectral imaging system based on the Fabry-Perot resonator are improved, the technological requirements are reduced, and the engineering applicability of the spectral imaging system based on the Fabry-Perot resonator is improved.

The invention discloses a photovoltaic panel surface pollution degree image evaluation system. The system comprises an image acquisition module, an image processing module, a multi-task evaluation module and a pollution degree determination module. Multi-spectral image data of the surface of a photovoltaic panel are collected, after radiation calibration and atmospheric correction processing, spectral reflectivity features and spatial texture features are extracted and fused, a multi-task deep learning model is input, a pollution type classification result and a power generation efficiency loss weight of each pixel region are output, and a pollution type classification result of each pixel region is obtained. And finally, generating a pollution type spatial distribution diagram, calculating the overall power generation efficiency loss percentage, and determining the pollution degree grade. According to the method, the problem that the mixed pollution type cannot be distinguished and the differentiation influence cannot be evaluated in the prior art is solved, accurate quantitative evaluation of the pollution degree is realized, and a reliable basis is provided for fine operation and maintenance of a photovoltaic power station.

The invention relates to the technical field of remote sensing image processing and computer vision, in particular to a hyperspectral remote sensing image reconstruction method based on iterative optimization and depth prior. According to the method, complementary information of a low-resolution multispectral image and a high-resolution panchromatic image is utilized, the spatial resolution and the spectral fidelity of the high-spectral image can be improved at the same time, and the complementary information of the low-resolution multispectral image and the high-resolution panchromatic image is effectively utilized; through a spectrum low-rank decomposition and alternate optimization strategy, the complexity of a high-dimensional optimization problem is reduced, and the stability of iterative solution is improved; and after the deep denoising network is introduced, the denoising capability and the feature expression capability of the subspace coefficient image are further enhanced, so that the generated hyperspectral image is superior to the existing method in the aspects of spatial details and spectral authenticity, and the hyperspectral image has relatively strong robustness and wide application prospects.

The invention provides a health state prediction method for an 8K display device, and relates to the technical field of display health prediction.The health state prediction method comprises the steps that a multi-modal feature vector is extracted from each pixel node of a multi-modal frequency spectrum image, a mutual information feature vector of the multi-modal feature vector is calculated, and subtle changes of cooperative work among different physical attributes in the nodes are captured; on the basis, each pixel node is further regarded as a node in a pixel node grid, the connection weight is quantized by calculating the norm of a mutual information feature vector of the pixel node and obtaining the similarity between adjacent nodes, then the attenuation rate of the connection weight of a continuous time window is calculated, the distribution entropy value of the attenuation rate is analyzed, and a stability degradation entropy index is obtained; trivial connection change information is condensed into a quantitative index representing the overall orderliness decline degree of the system, then time sequence prediction is carried out on an entropy index historical sequence, the expected time of the entropy index historical sequence reaching the dynamic control upper limit is calculated, and the crossing from current state monitoring to future trend prediction is achieved.

The invention discloses a graphite equipment weld defect intelligent detection and positioning system based on multispectral imaging, and particularly relates to the field of welding, and the system is characterized in that a multiband image acquisition unit acquires spectral image data of multiple bands and transmits the spectral image data to a multispectral data storage library; the spectral parameter analysis unit extracts characteristic parameters of the weld joint area and establishes a spectral characteristic parameter library; a spectrum abnormity identification unit locates a suspected defect area of spectrum abnormity; the weld defect positioning unit marks defect boundaries and ranges; the defect identification and grading unit calculates a defect severity index; the element component analysis unit is used for detecting the suspected defect position sample and analyzing the types and contents of trace elements in the suspected defect position sample; the defect and component data are integrated through the weld quality comprehensive evaluation unit, the quality grade and the rectification suggestion are generated, the problem that the defect cause cannot be positioned in the prior art is solved, data support is provided for follow-up maintenance and process optimization, and potential safety hazards of equipment operation and maintenance are reduced.

The invention relates to the technical field of industrial vision multispectral imaging detection, in particular to a defect detection method and system based on multispectral fusion imaging, and the method comprises the steps: synchronously collecting a multi-channel spectral image of the surface of an object to be detected, and generating an enhanced feature map fusing multispectral information; identifying and preliminarily marking a suspected defect region by applying a region segmentation algorithm based on anomaly detection; extracting a multi-dimensional spectral response curve of each region in each original spectral channel, and constructing a spectral feature vector; matching the vector with a standard template of a pre-established defect spectral feature database, and classifying and confirming defect types; and carrying out contour refined analysis, calculating the geometric dimension and position of the defect, and generating a structured detection report. According to the method, the sensitivity of defect detection and the classification accuracy are improved through accurate matching of multispectral fusion and spectral features, and automatic and high-precision defect identification is realized.

The invention discloses an insulator detection method and system based on an intelligent robot, and relates to the field of image detection, and the method comprises the steps: collecting an image through an inspection robot carrying a visible light and infrared camera, processing a visible light image through employing a dynamic correction algorithm fusing Retinex-HSV, and achieving the registration fusion of a multispectral image through employing an SURF feature point and an RANSAC algorithm; based on an improved Mask R-CNN network, precise segmentation of the insulator is achieved, and then defect features such as surface dirt, structural cracks and core rod deterioration are extracted hierarchically; and finally, generating a defect comprehensive probability by fusing multiple features through a D-S evidence theory, and automatically alarming when the defect comprehensive probability exceeds a threshold value. The method has the advantages that the data quality is improved through multi-spectral image adaptive acquisition, Retinex-HSV dynamic correction and SURF-RANSAC fusion, and efficient and accurate full-automatic defect detection of the power transmission line insulator is realized in combination with lightweight Mask R-CNN segmentation and D-S evidence theory multi-feature decision.

The invention discloses an end-to-end polarization hyperspectral image classification method and system, belongs to the technical field of deep learning and optical imaging, and solves the technical problems of low reconstruction process speed, limited precision, low information utilization rate in a classification process and weak feature extraction capability in the prior art. The method comprises the following steps: carrying out target shooting based on a snapshot type space coding hyperspectral polarization imaging system, and carrying out system coding on a shot image to obtain two-dimensional aliasing data; reconstructing the two-dimensional aliasing data into a polarization hyperspectral data cube by using the trained reconstruction network; training the classification network based on the polarization hyperspectral data cube to obtain a trained classification network; and performing joint fine tuning on the trained reconstruction network and the trained classification network to obtain a polarization hyperspectral image classification model, and performing polarization hyperspectral image classification by using the polarization hyperspectral image classification model. The method is used for realizing high-quality reconstruction and accurate classification of the polarization hyperspectral image.

The invention discloses an orchard unmanned aerial vehicle multispectral image noise restoration method based on a generative adversarial network, and relates to the technical field of multispectral image noise restoration, and the method comprises the steps: synthesizing a noisy sample containing stripes, photon particles, thermal noise, Rayleigh, shielding, compression artifacts, overexposure and waveband crosstalk on a noise-free reference image through a conditional generative adversarial network; a crosstalk matrix T obtained through optical calibration is introduced; based on the canopy / soil / sky semantic partition, main-cooperation-subdivision level superposition is implemented; generating hierarchical labels by taking energy conservation consistency and thermal noise correlation as thresholds; a branch and region self-adaptive repair module corresponding to classification is arranged in the image-to-image generative adversarial network, and joint loss training containing T regularization is adopted; and outputting the repaired image subjected to physical consistency checking. The method gives consideration to visual quality and spectral fidelity, inhibits cross-band pollution, improves the stability of indexes such as NDVI and NDRE, and is suitable for precision agricultural monitoring.

The invention discloses a hyperspectral image classification method based on frequency domain denoising and element gradient correction, and the method comprises the following steps: carrying out the preprocessing of all hyperspectral image data, and dividing an overall training sample set formed by the processed hyperspectral images into a training set and a verification set; constructing a sample weighting model based on frequency domain denoising and element gradient correction; in the training process, a parameterization frequency spectrum gating sensing transformation module is utilized to map features to a frequency domain through discrete Fourier transform, a learnable frequency spectrum response function is utilized to adaptively suppress spectrum jitter noise, and finally pure features are reconstructed. And automatically constructing a high-confidence pseudo-clean verification set based on a Gaussian mixture model and time domain consistency. According to the method, a time domain momentum updating mechanism is introduced, the variance of statistical estimation is effectively smoothed, random interference caused by training fluctuation is resisted, and the accuracy of pseudo clean set construction and the convergence stability of overall model training are further improved.

The invention provides a method and a device for measuring the solid-phase deposition amount of crude oil under high-temperature and high-pressure conditions, and relates to the technical field of oil-gas field development. Under the reservoir temperature and pressure of a target oil reservoir, ground degassed crude oil, associated gas and an on-site solid-phase sediment sample are mixed and dissolved; an in-situ crude oil sample capable of truly reflecting in-situ crude oil components is prepared; then, reducing the pressure to a target pressure point in a stepped manner at a constant temperature, and delimiting a deposition monitoring area; spectral image data of the area is obtained, spectral reflection reference rate is analyzed and calculated pixel by pixel, and a spectral distribution matrix is constructed; judging whether the solid-phase deposition reaches balance or not by calculating a space consistency index and a time convergence index of the matrix; finally, the solid-phase deposition amount is calculated after deposition equilibrium; according to the method, the problem of solid-phase deposition measurement distortion under the ultrahigh-pressure oil reservoir condition is solved, and interference-free, real-time monitoring and quantitative analysis in the deposition process are achieved.

The invention discloses an intelligent optical sensing system based on multispectral fusion, and relates to the technical field of optical sensing, and the system comprises a multispectral image collection module which is used for synchronously collecting original image data of a target scene under three spectral channels of visible light, near-infrared and short-wave infrared; the space-time registration and preprocessing module is used for carrying out high-precision space-time registration and radiation correction on the images of different spectrum channels; the self-adaptive feature extraction and fusion module is used for extracting multi-scale spectral features from the registered multi-spectral image and dynamically selecting and weighting a fusion strategy according to scene content; and the lightweight decision network module outputs a final target identification and state discrimination result. According to the technical scheme, the time-space consistency of multispectral data acquisition can be realized, the cross-band registration precision is improved, the robustness under the conditions of low contrast, shielding and severe weather is enhanced, meanwhile, the calculation complexity and memory occupation are greatly reduced, and the method is suitable for edge calculation equipment with limited resources.

The invention belongs to the technical field of optical lens detection, and relates to a lens detection method and device, computer equipment and a storage medium, and the method comprises the steps: detecting a lens based on a plurality of detection points of the lens, and obtaining a plurality of spectral images; acquiring position information of each detection point, acquiring a standard image according to the position information, matching the standard image with the spectral image, and determining defect information of the lens; predicting the performance of the lens according to the spectral image and the defect information, and determining performance prediction data of the lens; and generating a detection report of the lens according to the performance prediction data. The production efficiency and the product yield are improved, and the detection efficiency of the detection process is improved.

This invention discloses a method and system for hyperspectral image denoising based on a spatial-spectral joint self-attention mechanism, belonging to the field of image processing technology. First, based on the characteristics of hyperspectral images, a spatial-spectral joint self-attention mechanism network is constructed. The noisy hyperspectral image is used as input to the network to extract spatial-spectral features. Then, a global spectral self-attention mechanism is used to extract the band correlations of the hyperspectral image. Finally, the extracted spatial-spectral features are reconstructed using a multiple perceptron and residual connections to reconstruct a clean, noise-free hyperspectral image. The system includes a feature extraction subsystem, a non-local spatial self-attention subsystem, a global spectral self-attention subsystem, and an image reconstruction subsystem. This invention can effectively restore noisy hyperspectral images to obtain noise-free hyperspectral images. Compared to convolutional networks, it can better model long-range dependency information and has better adaptability to target hyperspectral images.

The invention relates to an FFT-based unsupervised domain adaptive hyperspectral image classification method and system. The method comprises the steps of obtaining a remote sensing hyperspectral image, performing preprocessing, cross-domain alignment operation and edge filling operation, and performing alignment operation by using a local neighborhood; fourier transform is carried out on the source domain image and the target domain image; phase information of the source domain image is reserved, amplitude information is replaced with amplitude information of the target domain image, and the injection intensity of the amplitude of the target domain image is controlled; reconstructing the source domain image fused with the target domain amplitude information through inverse Fourier transform to obtain a data enhanced image; a ResNet model is used to extract joint features about a local space and a spectrum; performing feature extraction by adopting a channel attention mechanism and a space attention mechanism based on Fourier transform to obtain adaptive re-calibration features; unsupervised domain adaptive classification is realized by adopting multiple task heads and conditional adversarial training; and performing classification by using the trained classification model. According to the invention, the accuracy of image classification is improved.

The invention discloses a regenerated wafer scanning method, and relates to the technical field of semiconductor wafer scanning, and the method comprises the steps: obtaining the identification information of a target wafer transmission box when the target wafer transmission box is detected; detecting the in-place state of the wafer through a light curtain sensor, and generating a slot mapping table; controlling a preset micro-rotation mechanism to carry out micro-rotation on the wafer at each slot position within a set angle range, and collecting an annular structure light image; constructing polar coordinates and calculating a first attitude angle of each wafer; driving a micro-rotation mechanism to adjust the wafer to a preset reference angle according to the first attitude angle to obtain a standard attitude wafer; the mechanical arms are driven to be sequentially conveyed to a scanning station; respectively controlling a rotary positioning module to rotate in a first angle range and a second angle range, and acquiring a coded image and a spectral image; the wafer code is analyzed, confidence verification is carried out, and a wafer process type label is identified; and warehousing and recording the wafer codes, the process labels and the transmission box identification information. According to the invention, the storage management efficiency of the regenerated wafer can be effectively improved.

The invention relates to the technical field of image processing, and discloses a tea maturity detection method and system, and the method comprises the steps: obtaining an original image of tea in a natural illumination imaging environment; generating a global illumination difference chart representing an ambient illumination state, and performing illumination compensation on the to-be-detected area of the tea in the original image to obtain a compensated image of the tea; performing gray scale range adjustment on the compensated image of the tea leaves to obtain an enhanced spectral image of the tea leaves; integrating the color and spectral characteristics of the specific spectral reflection characteristics of the tea leaves with different maturity degrees in the enhanced spectral image and the three-dimensional morphological characteristics reflecting the spreading degree and thickness of the tea leaves into the fusion image characteristics of the tea leaves; mapping the color spectrum emission characteristics and texture characteristics of the tea leaves in the enhanced spectrum image to a maturity discrimination knowledge base of the tea leaves to obtain a preliminary maturity result; performing confidence coefficient weighting on the preliminary maturity result to obtain a target maturity; according to the invention, the tea maturity detection efficiency can be improved.

The invention discloses a product image recognition method, device and equipment, a medium and a product, and relates to the technical field of artificial intelligence, and the method comprises the steps: obtaining a first multispectral image of a first product object, and carrying out the dynamic noise suppression and feature enhancement processing of the first multispectral image, and obtaining a second multispectral image; inputting the second multispectral image into a pre-trained image recognition model to obtain a defect probability thermodynamic diagram corresponding to pixel points in the first multispectral image and uncertainty confidence of defect positioning; determining a defect type and defect position information corresponding to the first multispectral image according to the defect probability thermodynamic diagram and the uncertainty confidence; wherein the image recognition model comprises a global feature extraction module based on a Vision Transform architecture and a local feature extraction module based on a dilated convolutional neural network. According to the technical scheme provided by the embodiment of the invention, the identification capability of sub-pixel-level microdefects can be improved, and the accuracy of a product image identification result is improved.

A camera system. The camera system includes a sensor array; an infrared (IR) illumination system configured to emit active IR light in an IR light sub-band; a spectral illumination system configured to emit active spectral light in a spectral light sub-band; one or more logic machines; and one or more storage machines. The storage machines hold instructions executable by the one or more logic machines to address the sensor array to acquire an ambient image; based on at least the ambient image, activate the IR illumination system and address the sensor array to acquire an actively-IR-illuminated depth image; and based on at least the actively-IR-illuminated depth image, activate the spectral illumination system and address the sensor array to acquire an actively-spectrally-illuminated spectral image.

The invention relates to a fig health status assessment method and system based on multi-modal data fusion, and the method comprises the steps: constructing a multi-modal feature vector through fusing the features of a multi-spectral image, a hyperspectral cube and a visible light image; then proposing a multi-modal fusion suitability evaluation model, pre-processing the multi-modal feature vector, and then carrying out deep fusion and feature association to obtain a fusion feature; and finally, according to the fusion features, outputting an adaptation score and disease and pest classification. Therefore, the characteristic representation integrity is obviously enhanced, the evaluation accuracy and the pest and disease classification precision are improved, and the robustness and the anti-interference capability are enhanced.

The invention relates to a fire area inversion method based on airborne dual-spectrum detection and depth estimation, and belongs to the field of unmanned aerial vehicle detection. The method comprises the following steps: acquiring a multi-dimensional data set of a dual-spectrum image, a temperature image, an unmanned aerial vehicle attitude and the like; constructing a multi-modal space collaborative perception segmentation network, combining temperature change characteristics with temperature space distribution characteristics of flames to generate temperature region distribution characteristics, and coupling absolute temperature and pixel information to enhance flame weak edge extraction; designing a temperature-guided space structure loss function TSSLoss, and combining gradient change consistency constraint and temperature weight constraint on a segmentation loss function for network training; and according to the unmanned aerial vehicle pose, the target depth and the fire area segmentation pixel area, an early fire area is derived in combination with an airspace transmission inversion formula, and an actual fire area is calculated. According to the method, multi-source information is effectively combined for physical constraint, and more accurate fire detection segmentation and fire area calculation can be realized.

The method comprises the following steps: S1, collecting unmanned aerial vehicle multispectral image data of a tea garden, and synchronously collecting measured values of chlorophyll and nitrogen content of tea leaves in a corresponding region; s2, preprocessing the multispectral image data, and constructing a training set and a test set; s3, an LSSCM-MobileTeNet detection model is constructed, the model takes MobileNetV3 as a reference framework, a multi-Stage Stem structure (Multi-Stage Stem) is introduced to carry out hierarchical feature extraction, and an original bottleneck layer is replaced with a lightweight spatial spectrum convolution module (LSSCM-Block); and S4, training the LSSCM-MobileTeNet detection model by using the constructed data set, processing the multispectral image of the area to be detected by using the trained model, and outputting predicted values of the chlorophyll content and the nitrogen content. The method has the advantages that RMSE and MAE are remarkably reduced, R2 is improved, and higher prediction precision is achieved; according to the method, the chlorophyll nitrogen content of the tea leaves can be rapidly detected on a large scale; the multi-stage Stem structure enhances the ability of the model to extract the hierarchical features of the multispectral image.