10 results about "Spectral angle" patented technology

This invention discloses a method for detecting the aging state of asphalt pavement by integrating hyperspectral imagery and deep learning, relating to the field of deep learning technology. The method includes: acquiring and preprocessing hyperspectral images of the target road section to obtain surface reflectance images; dynamically identifying and masking vehicles and their shadows in the images using a spectral angle matching method to extract effective pixel spectral datasets; performing hybrid pixel decomposition on the effective pixel spectral datasets using a non-negative matrix factorization algorithm to obtain asphalt endmember spectra and aggregate endmember abundance maps; inputting the asphalt endmember spectra into a pre-trained oxidation degree assessment model to output asphalt oxidation degree values; overlaying the aggregate endmember abundance map with near-infrared band images and inputting it into a pre-trained exposure rate segmentation network to output a binary mask of aggregate exposure areas and calculate the aggregate exposure rate. This method solves the technical problem that existing technologies cannot quickly and non-destructively detect the aging state of asphalt pavement.

This invention proposes an intelligent landslide identification method based on changes in remote sensing semantic embedding. The method includes collecting annual remote sensing semantic embedding data, optical remote sensing imagery, digital elevation models, and land cover data for the study area; constructing physically feasible landslide identification areas constrained by both slope and land cover; calculating and standardizing the intensity of land cover semantic changes in adjacent years; generating annual vegetation index results based on growing season optical remote sensing imagery to produce landslide anomaly intensity results; extracting strong landslide candidate areas using adaptive thresholding; automatically generating potential and non-landslide samples, and optimizing the samples based on time-series normalized vegetation index and spectral angle measurements; and finally, using supervised classification methods such as random forest to achieve landslide identification and obtain the final landslide classification map. This invention realizes a complete intelligent landslide identification process, from anomaly identification of changes in remote sensing semantic embedding, automatic extraction of strong landslide candidate areas, automatic sample optimization, to supervised classification mapping.

This invention relates to a high-resolution remote sensing image enhancement method for forest ecological monitoring, belonging to the field of forest ecological remote sensing monitoring. It includes the following steps: unified grid registration and effective forest pixel selection are performed on Sentinel-2 multi-band surface reflectance data; a continuous canopy shadow intensity map integrating topographic incidence relationships and local canopy darkness is constructed; slow-release reflectance correction is performed by band grouping; a spectral-spatial collaborative residual enhancement network is constructed, extracting texture and band combination features through parallel spatial enhancement and spectral enhancement branches, respectively, and introducing a band group attention mechanism based on ecological function; the network is jointly trained using weighted reconstruction loss and spectral angle loss; finally, a panoramic enhanced image is generated through sliding window inference and weighted stitching of overlapping areas. This invention can effectively suppress deep shadow overcompensation and spectral distortion problems, and improve the detail representation of ecologically sensitive bands such as red edge and near-infrared.

This invention discloses a comprehensive visual super-resolution method based on neural networks. Through a two-stage optimization process of offline training and online deployment, it achieves real-time super-resolution reconstruction of remote sensing images on an edge computing platform. In the offline stage, a physics-data hybrid-driven training dataset is constructed, including optical MTF analysis, atmospheric turbulence simulation, and sinc artifact enhancement. A dynamic network, RAG-SRNet, is designed to adaptively switch the RRDB computation path based on the screen projection size, and training is performed using remote sensing-specific perception loss, RaGAN adversarial loss, and spectral angle loss. In the online stage, TensorRT INT8 quantization and mixed-precision optimization are employed. CUDA-OpenGL zero-copy memory mapping enables the reuse of memory addresses for the inference output buffer and texture objects. Asynchronous inference thread pools and hardware semaphore synchronization are used to achieve pipelined parallelism for tile requests, super-resolution computation, and texture rendering. Compared to existing methods, this invention significantly improves PSNR, increases frame rate by over 25%, and reduces CPU usage by 77%, providing an industrial-grade visual enhancement solution for aerospace simulation and autonomous driving.

This invention discloses a large-scene hyperspectral zero-shot classification method and system based on a visual language model. The method includes the following steps: generating an initial pseudo-label map for the hyperspectral image based on a visual language model; calculating the spectral angle between each pixel in the initial pseudo-label map and the spectral fingerprint vector; filtering the initial pseudo-label map based on the spectral angle to obtain an optimized pseudo-label map; training a hyperspectral segmentation network based on the hyperspectral image and its optimized pseudo-label map to establish a mapping relationship between the hyperspectral image and its land cover categories for land cover category classification. The hyperspectral segmentation network consists of an encoder, a decoder, and a classifier, wherein the encoder is composed of N-level cascaded multi-kernel convolutional residual modules. This invention aims to accurately identify and segment different land cover categories in complex hyperspectral images without manual annotation.

This invention relates to the field of image recognition technology and proposes a method and system for grading vegetable waste and adapting it to insect feed based on image recognition. The method includes: radiometric calibration of hyperspectral image data to obtain a spectral image cube; extracting chromatic space deviation feature values ​​and short-wave infrared absorption peak curvature feature values ​​from the cube; performing pixel-level matching to obtain the overall spoilage level; when the overall spoilage level is "fresh and tender," performing spectral angle matching between the reflectance of the soluble sugar feature band and the palatability benchmark spectral library of yellow mealworms to obtain a first set of feed processing parameters; when it is "semi-rotten," comparing the spectral characteristics of the lignin-to-cellulose ratio with the feeding preference spectral range of white-spotted flower beetle larvae to obtain a second set of feed processing parameters; when it is "deeply rotten," generating a third set of feed processing parameters; and outputting the corresponding parameter sets to a feed formulation terminal to obtain the final feed adaptation scheme. This invention can improve the efficiency of grading vegetable waste and adapting it to insect feed.

The application discloses a kind of training-free snapshot photometric stereo vision methods based on spectral compression imaging, comprising: (1) setting multiple different wavelength monochromatic light sources to simultaneously irradiate object from different directions, form high-dimensional spectral-angle data cube;(2) the spectral-angle data cube is modulated in space by coded aperture, along the spectrum dimension is sheared translation by using optical dispersion element, finally integrates on single-channel image sensor, generates single two-dimensional gray compressed measurement image;(3) inverse problem model is constructed, and multiple illumination image sequence is decoupled and recovered from single two-dimensional gray compressed measurement image using iterative optimization algorithm based on physical drive;(4) the surface normal map of the measured object is predicted by inputting multiple illumination image sequence into pre-trained photometric stereo vision network module.The application can realize low-cost, high data efficiency and high physical fidelity three-dimensional reconstruction by single snapshot.

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 relates to the field of quality inspection and discloses a real-time nut quality sorting system based on a hyperspectral camera. The system includes a conveying device, a hyperspectral imaging device, a data processing and analysis module, and an execution module. Hyperspectral imaging acquires a three-dimensional data cube of the nut raw material in a specific wavelength band. The data processing module preprocesses the data, extracts spectral and spatial morphological features, and employs a two-stage cascaded discrimination algorithm for quality determination: the first stage quickly removes morphologically abnormal products based on morphological rules; the second stage accurately identifies internal defects such as mold by calculating the spectral angle and fusing the first derivative and reflectance features into a mold characteristic index. The execution module is a high-pressure jet valve array, which achieves precise removal of defective products through spatiotemporal mapping and delay control. This invention enables simultaneous detection of the internal and external quality of nuts, significantly improving sorting accuracy and efficiency.

The application provides a kind of wetland vegetation change detection method and device, the method comprises: obtaining the Landsat image and high time resolution MODIS image in the set time period of the area to be monitored, and pretreatment is carried out;Reconstruct long time series, high density Landsat surface reflectivity data set using space-time fusion algorithm;By analyzing the measured spectral characteristics of wetland vegetation, an optimized wetland vegetation change detection index based on spectral angle mapping principle is constructed;Through the LandTrendr time series segmentation algorithm of fusing spatial context features, the accurate and automatic detection of large-scale, long time series wetland vegetation change is realized.The method uses cloud computing capability to process massive remote sensing data, effectively solves the problem of data missing and insufficient algorithm stability in large-scale, long time series wetland vegetation monitoring, and provides efficient technical support for wetland resource management, carbon sink function evaluation and "carbon neutralization" strategy.