Mineral identification method and device based on hyperspectral camera and storage medium

Abstract

The invention discloses a mineral recognition method and device based on a hyperspectral camera and a storage medium, relates to the technical field of image processing, and aims to improve mineral recognition efficiency and recognition accuracy. Comprising the following steps: in response to an identification signal of a to-be-identified mineral in a target area, determining a load parameter of a spaceborne hyperspectral camera required for identifying the to-be-identified mineral; shooting the target area by using a satellite-borne hyperspectral camera with a load parameter to obtain a satellite-borne remote sensing image of the target area, and performing quality enhancement processing on the satellite-borne remote sensing image; selecting a preset number of reference spectral bands in a preset spectral set, determining a pure pixel index, and extracting an end member spectrum in the satellite-borne remote sensing image after quality enhancement based on the reference spectral bands and the pure pixel index; and matching the end member spectrum with standard end member spectrums corresponding to different minerals in a preset end member spectrum library, and determining the types of the minerals in the target area based on a matching result.

Description

Technical Field

[0001] This invention relates to the field of image processing technology, and in particular to a mineral identification method, apparatus, and storage medium based on a hyperspectral camera. Background Technology

[0003] Currently, mineral identification is usually done through manual detection. However, manual detection is time-consuming and labor-intensive, and due to the varying skill levels of the personnel, it can lead to missed or incorrect identification of minerals. Summary of the Invention

[0004] This invention provides a mineral identification method, device, and storage medium based on a hyperspectral camera, which mainly improves the efficiency and accuracy of mineral identification.

[0005] According to a first aspect of the present invention, a mineral identification method based on a hyperspectral camera is provided, comprising: In response to the identification signal of the mineral to be identified in the target area, determine the payload parameters of the spaceborne hyperspectral camera required to identify the mineral to be identified; The target area is photographed using a spaceborne hyperspectral camera with the aforementioned payload parameters to obtain a spaceborne remote sensing image of the target area. The spaceborne remote sensing image is then enhanced to obtain a quality-enhanced spaceborne remote sensing image. A preset number of reference spectral bands are selected from the preset spectral set, and the purity pixel index is determined. Based on the reference spectral bands and the purity pixel index, endmember spectra are extracted from the enhanced spaceborne remote sensing image. The endmember spectrum is matched with the standard endmember spectra corresponding to different minerals in a preset endmember spectrum library, and the mineral types in the target region are determined based on the matching results.

[0006] According to a second aspect of the present invention, a mineral identification device based on a hyperspectral camera is provided, comprising: The determination unit is used to determine the payload parameters of the spaceborne hyperspectral camera required to identify the mineral to be identified in response to the identification signal of the mineral to be identified in the target area. An enhancement unit is used to capture images of the target area using a spaceborne hyperspectral camera with the payload parameters, to obtain a spaceborne remote sensing image of the target area, and to perform quality enhancement processing on the spaceborne remote sensing image to obtain a quality-enhanced spaceborne remote sensing image. The extraction unit is used to select a preset number of reference spectral bands from a preset spectral set and determine the clean pixel index. Based on the reference spectral bands and the clean pixel index, it extracts endmember spectra from the enhanced spaceborne remote sensing image. The matching unit is used to match the endmember spectrum with the standard endmember spectra corresponding to different minerals in a preset endmember spectrum library, and determine the mineral types in the target region based on the matching results.

[0007] According to a third aspect of the present invention, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the above-described mineral identification method based on a hyperspectral camera.

[0008] According to a fourth aspect of the present invention, a computer device is provided, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the above-described mineral identification method based on a hyperspectral camera.

[0009] The present invention provides a mineral identification method, apparatus, and storage medium based on a hyperspectral camera. Compared with current methods of mineral identification through manual detection, the present invention, before mineral identification, ensures the quality of remote sensing image acquisition and storage by rationally designing the payload parameters of the spaceborne hyperspectral camera, thereby guaranteeing the subsequent mineral identification effect. By enhancing the quality of the remote sensing image, noise in the image can be removed, improving image quality and thus enhancing the accuracy of subsequent mineral identification. By extracting endmember spectra using reference spectral bands and purity pixel indices, mixed pixels can be automatically filtered to ensure that the extracted endmember spectra have high purity. Finally, mineral identification is performed based on high-purity endmember spectra, which can improve the accuracy of mineral identification. By matching the endmember spectra with standard spectra in a preset endmember spectral library to identify minerals, the physical interpretability of the classification results can be ensured, further improving the accuracy of mineral identification. Attached Figure Description

[0010] The accompanying drawings, which are included to provide a further understanding of the invention and form part of this application, illustrate exemplary embodiments of the invention and, together with their description, serve to explain the invention and do not constitute an undue limitation thereof. In the drawings: Figure 1 A flowchart of a mineral identification method based on a hyperspectral camera provided by an embodiment of the present invention is shown; Figure 2 This diagram illustrates a standard reference spectral library provided by an embodiment of the present invention. Figure 3 This invention provides a spectral curve comparison diagram according to an embodiment of the invention; Figure 4 A flowchart of another mineral identification method based on a hyperspectral camera provided by an embodiment of the present invention is shown; Figure 5 A schematic diagram of a mineral identification device based on a hyperspectral camera provided in an embodiment of the present invention is shown. Figure 6 A schematic diagram of another mineral identification device based on a hyperspectral camera provided in an embodiment of the present invention is shown; Figure 7 A schematic diagram of the physical structure of a computer device provided in an embodiment of the present invention is shown. Detailed Implementation

[0011] The present invention will be described in detail below with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in the present application can be combined with each other.

[0012] Currently, the method of identifying minerals through manual detection is time-consuming and labor-intensive, and due to the varying skill levels of the detection personnel, it can lead to missed or incorrect identification of minerals.

[0013] To address the aforementioned problems, embodiments of the present invention provide a mineral identification method based on a hyperspectral camera, such as... Figure 1 As shown, the method includes: 101. In response to the identification signal of the mineral to be identified in the target area, determine the payload parameters of the spaceborne hyperspectral camera required to identify the mineral.

[0014] The target area can be any area where mineral identification is required; the mineral to be identified can be any type of mineral such as copper ore, gold ore, or iron ore.

[0015] In this embodiment of the invention, to accurately identify minerals, it is first necessary to rationally design the payload parameters of the spaceborne hyperspectral camera. Based on this, step 101 specifically includes: determining the absorption characteristic data of various minerals, the regional characteristic data of the target area, and the mineral identification requirement data; determining the spectral range of the spaceborne hyperspectral camera based on the absorption characteristic data, the regional characteristic data, and the mineral identification requirement data; determining reference ground features in the target area, and determining the spectral characteristics of the reference ground features, and determining the spectral correlation of adjacent bands in the reference hyperspectral data of the reference ground features; determining the image compression parameters of the spaceborne hyperspectral camera based on the spectral characteristics of the ground features and the spectral correlation; acquiring the performance indicators, image acquisition requirements, and reference remaining payload parameters of multiple reference image acquisition devices; and determining the camera remaining payload parameters of the spaceborne hyperspectral camera based on the performance indicators, image acquisition requirements, and reference remaining payload parameters of each reference image acquisition device, wherein the camera remaining payload parameters are payload parameters other than the spectral range and the image compression parameters.

[0016] Among them, absorption characteristic data can be the geographical depth at which each element in the mineral to be identified exhibits absorption characteristics. For example, quartz in copper ore has weak absorption characteristics at 1400, 1900, and 2100 nm; chlorite has absorption characteristics at 900 nm and 2200-2300 nm; and sericite typically has... Absorption characteristics are present in the 2200-2300nm range, while dolomite exhibits extremely weak absorption characteristics in the 400-700nm range. Regional feature data includes: regional topographic data, regional weather data, regional vegetation area, and vegetation types. Mineral identification requirements include the mineral type to be identified and the identification accuracy. Reference ground features can be information on vegetation, buildings, etc. on the ground in the target area. Spectral characteristics of ground features include the reflection, absorption, emission, and transmission characteristics of ground features to electromagnetic waves of different bands.

[0017] Specifically, the method for determining the spectral range of a spaceborne hyperspectral camera includes: determining the absorption feature vector corresponding to the absorption feature data, the regional feature vector corresponding to the regional feature data, and the demand feature vector corresponding to the mineral identification demand data; performing feature-level cross processing on the absorption feature vector, the regional feature vector, and the demand feature vector to obtain a feature cross vector; performing element-level cross processing on the absorption feature vector, the regional feature vector, and the demand feature vector to obtain an element cross vector; performing low-order cross processing on the absorption feature vector, the regional feature vector, and the demand feature vector to obtain a low-order cross vector; transforming the feature cross vector, the element cross vector, and the low-order cross vector using a preset transformation function to obtain a spectral cross feature vector; and inputting the spectral cross feature vector into a preset spectral range prediction model to predict the spectral range, thereby obtaining the camera spectral range of the spaceborne hyperspectral camera.

[0018] Specifically, feature extraction models (such as CNN models) are used to extract the absorption feature vectors corresponding to the absorption feature data, the region feature vectors corresponding to the region feature data, and the demand feature vectors corresponding to the mineral identification demand data. Then, to fully utilize the relationships between the data, extract more latent features, and simultaneously handle both high-order and low-order processing to make the data utilization more efficient and the subsequent prediction results more accurate, meeting the needs of practical application scenarios, it is necessary to perform cross-processing on the absorption feature vectors, region feature vectors, and demand feature vectors. For example, if the absorption feature vector is (a1, a2), the region feature vector is (b1, b2), and the demand feature vector is (c1, c2), the specific cross-processing method includes: performing feature-level cross-processing between different feature vectors, that is, after performing a Hadamard product on all elements of the vectors, a convolution transformation is performed under a certain weight w1 to obtain the feature cross vector f(w1 × (a1 × b1 × c1, a2 × ... Simultaneously, element-level crosses are performed on all feature vector data. This involves performing a Hadamard product on each element of the vectors, assigning different weight values ​​w2 and w3 to each product, and then performing a linear transformation to obtain the element-level cross vector f(w2×a1×b1×c1, w3×a2×b2×c2). Furthermore, low-order cross processing is performed on all feature vectors, and the result of the cross processing is assigned a weight coefficient w4, followed by a linear transformation to obtain the low-order cross vector f(w4(a1,a2,b1,b2,c1,c2)). Finally, a preset transformation function is used to combine the above feature cross vectors, element-level cross vectors, and low-order cross vectors to obtain the spectral cross feature vector. It should be noted that the preset transformation function Wie is determined according to actual needs, and the above examples are merely illustrative and do not limit the embodiments of this application. Therefore, by cross-processing the absorption feature vector, regional feature vector, and demand feature vector, different features can be automatically or explicitly combined to generate new feature combinations. These combined features may contain complex nonlinear relationships between the original features, enabling the model to capture more refined and richer information in the data. In other words, it can make full use of the relationships between various data, extract more latent features, and take into account both high-order and low-order processing, making the data utilization more efficient and the resulting spectral range results more accurate, thus meeting the needs of practical application scenarios.Furthermore, to improve the prediction accuracy of the preset spectral range prediction model, it is necessary to train and construct the preset spectral range prediction model. Based on this, the method includes: constructing a preset initial spectral range prediction model; acquiring a sample dataset, wherein the sample dataset includes regional feature data, mineral identification requirement data, and absorption feature data of the target area collected by a sample camera with spectral range labels; dividing the sample dataset into a training set and a test set; training the preset initial spectral range prediction model using the training set; and testing the trained preset initial spectral range prediction model using the test set; finally, the trained preset initial spectral range prediction model that meets the test conditions is used as the preset spectral range prediction model. Specifically, in the model training process, the preset initial spectral range prediction model is first constructed, and then the sample dataset is acquired. It is ensured that the dataset contains all necessary files. The data is converted to a format that the preset initial spectral range prediction model can understand, and finally, the model is trained and tested. Specifically, the dataset can be divided first: the sample dataset can be divided into a training set and a test set using random or specific strategies (such as stratified sampling). Then, the model is trained using the training set, and the trained model is tested using the test set to evaluate its performance on unseen data. Calculate and record metrics such as precision and recall on the test set. If the model performance does not meet requirements, return to the training phase for further iterations or adjustments. This process yields a prediction model with a preset spectral range that meets the requirements. Finally, the spectral cross-validation directly outputs the spectral range of the spaceborne hyperspectral camera. For example, to design a hyperspectral camera capable of identifying copper and gold mines, the camera's spectral range needs to cover the entire visible-near-infrared band, such as 400-2500 nm.

[0019] Furthermore, in determining the image compression process, firstly, widely distributed and spectrally stable ground features within the target area, such as target minerals (e.g., spodumene, montmorillonite), are selected as reference ground features. Then, the spectral characteristics of the reference ground features, such as spectral slope, reflectance value, rate of change of reflectance between adjacent bands, absorption peaks, and reflection peaks, are determined in each band. Simultaneously, the reflectance of the i-th band and the (i+1)-th band in the reference hyperspectral data are determined, and the covariance of the reflectance of the i-th band and the (i+1)-th band is calculated. The standard deviation of the i-th band is multiplied by the standard deviation of the (i+1)-th band, and the ratio of the covariance to the multiplication result is taken as the spectral correlation between the i-th band and the (i+1)-th band. Thus, the spectral correlation between any adjacent bands can be calculated in the above manner. Then, the reference hyperspectral data is divided into bands according to the spectral correlation, that is, the highly correlated spectra are grouped into the same group, and the low-correlation spectra are grouped into the same group, resulting in multiple sets of reference hyperspectral data. Finally, based on the spectral characteristics of ground objects, different compression parameters are set for different sets of hyperspectral data, that is, the image compression parameters of the spaceborne hyperspectral camera are obtained. For example, PCA compression is used for the 1.8-2.0μm band, retaining the first 3 principal components, and lossless compression is used for the 1.41μm absorption peak band. If the acquired remote sensing image data is large, it is inconvenient to transmit the data to the ground. To solve this problem, the data is compressed and stored, and then transmitted to the ground for spectral data calculation and imaging to restore the spectral information of the acquired images. Applying compression technology to the hyperspectral camera reduces the amount of data acquired on-board, decreasing storage space and transmission bandwidth, while ensuring the integrity and usability of spectral features and other information in the acquired remote sensing image data. Furthermore, compressing and storing the data after acquisition also reduces the use of some processing components in the hyperspectral camera, lightening its overall weight and size, thus achieving weight reduction. After acquiring image data, the acquired image data is compressed and stored using components such as the camera's embedded optical system, retaining only key spectral information. Artificial intelligence algorithms are introduced. After the ground receiving system receives the compressed and stored image data, the AI ​​algorithm is used to solve and reconstruct the image, restoring the spectral features and other information of the acquired images. To ensure the accuracy of the AI ​​algorithm's calculations, we plan to use a mineral spectral library for training to improve its calculation capabilities.

[0020] Furthermore, in setting the remaining payload parameters, the performance indicators, image acquisition requirements, and reference remaining payload parameters of multiple reference image acquisition devices are first determined. These multiple reference image acquisition devices include, but are not limited to, airborne hyperspectral cameras, on-orbit spaceborne hyperspectral cameras, and cameras mounted on on-orbit CubeSats. Performance indicators include, but are not limited to, sensitivity, signal-to-noise ratio, shutter speed, readout rate, and integration density. Image acquisition requirements include, but are not limited to, image acquisition quality and image acquisition range. Reference remaining payload parameters include, but are not limited to, spatial resolution, focal length, aperture, field of view, and geometric distortion. Camera remaining payload parameters include, but are not limited to, spatial resolution, focal length, aperture, field of view, and geometric distortion. In other words, the payload parameters of the spaceborne hyperspectral camera can be set by combining the payload parameters of multiple high-precision recording cameras. Therefore, this embodiment of the invention can ensure image acquisition quality by setting the payload parameters for the spaceborne hyperspectral camera.

[0021] 102. The target area is photographed using a spaceborne hyperspectral camera with payload parameters to obtain a spaceborne remote sensing image of the target area. The spaceborne remote sensing image is then enhanced to obtain the enhanced spaceborne remote sensing image.

[0022] In this embodiment of the invention, after acquiring spaceborne remote sensing images using a spaceborne hyperspectral camera, the images need to undergo preprocessing such as correction to obtain high-quality spaceborne remote sensing images. Analyzing these high-quality spaceborne remote sensing images can improve the accuracy of mineral identification.

[0023] 103. Select a preset number of reference spectral bands from the preset spectral set and determine the purity pixel index. Based on the reference spectral bands and the purity pixel index, extract endmember spectra from the enhanced spaceborne remote sensing images.

[0024] In this embodiment of the invention, after image quality enhancement, it is also necessary to extract endmember spectra from the enhanced image. Based on this, step 103 specifically includes: sorting each pixel in the enhanced satellite remote sensing image based on the clean pixel index of each pixel, and selecting a preset number of pixels as initial endmembers in each sorted pixel; calculating the spectral matching degree between each initial endmember and the spectrum corresponding to the reference spectral band; selecting a target endmember in each initial endmember based on the spectral matching degree, and using the spectrum of the target endmember as the endmember spectrum in the satellite remote sensing image.

[0025] Among them, the purity pixel index reflects the "extremeness" of a pixel in the multidimensional spectral space. The higher the value, the more likely the pixel is to be a pure endmember.

[0026] Specifically, the preset spectral set stores standard spectra corresponding to various standard minerals (the standard spectra include reference spectral bands). Principal component analysis is performed on the enhanced image to extract the first 5 principal components. Random vector projection is then performed: N random unit vectors (e.g., N=1000) are generated, and each pixel is projected onto these vectors. Extreme values ​​are counted: the number of times a pixel becomes an extreme value (maximum / minimum) in all projections is recorded; this number is the pure pixel index. All pixels in the spaceborne remote sensing image are sorted from high to low according to their pure pixel index values, generating a sorted list. Starting from the top of the sorted list, the first M pixels are selected as initial endmembers. The angle between the initial endmember spectrum and the spectrum corresponding to the reference spectral band is calculated using spectral angle mapping. The matching degree is determined based on the angle; the smaller the angle, the higher the matching degree. Then, initial endmembers with a matching degree greater than a preset matching degree threshold (the preset matching degree threshold is set according to actual needs) are selected as target endmembers. Finally, the spectrum of the target endmember is used as the endmember spectrum in the spaceborne remote sensing image. An endmember represents the spectral curve of the remote sensing image.

[0027] 104. Match the endmember spectra with the standard endmember spectra corresponding to different minerals in the preset endmember spectral library, and determine the mineral types in the target area based on the matching results.

[0028] In this embodiment of the invention, various spectral matching algorithms, such as binary encoding, spectral angle mapping, and spectral feature fitting, can be used to match the extracted endmember spectral curves with the spectra of typical minerals in a preset endmember spectral library, thereby identifying the type of the extracted endmember spectrum and thus obtaining the mineral type. For example, the weights of the three identification results are all set to 1, with a total score of 3. The closer the total score is to 3, the higher the matching degree between the extracted endmember spectrum and the specific typical mineral in the preset endmember spectral library. Finally, based on the score, the mineral type of the endmember mineral is determined according to the mineral with the highest score. If there are multiple minerals with high scores, it is necessary to combine auxiliary information such as the color of the sampled mineral itself and the collection area for judgment. This embodiment of the invention extracts endmember spectra by referencing spectral bands and purity pixel index, which can automatically filter mixed pixels to ensure that the extracted endmember spectra have high purity. Finally, mineral identification is performed based on high-purity endmember spectra, which can improve the accuracy of mineral identification. By matching the endmember spectrum with the standard spectrum in the preset endmember spectral library to identify minerals, the physical interpretability of the classification results can be ensured, further improving the accuracy of mineral identification.

[0029] Furthermore, to increase the accuracy of mineral identification, the accuracy of the identification results can be evaluated. To assess the correctness and feasibility of the proposed spaceborne remote sensing image data processing method, a standard hyperspectral camera was used to acquire and process spectral images of some minerals with typical spectral characteristics, constructing a standard reference spectral library. The spectral curves of the corresponding minerals in the standard reference spectral library were compared with those of the identified minerals, and the accuracy of mineral identification was determined based on the comparison results. For example, the standard hyperspectral instrument is a Resonon pushbroom hyperspectral camera equipped with a Pika XC2 camera. The Resonon pushbroom hyperspectral camera performs scanning imaging through a moving imaging platform, offering higher spectral resolution than frame-type cameras. This camera is widely used in desktop and field systems. The Pika XC2 camera used has an imaging band range of 400-1000nm, exhibiting extremely high spectral resolution, very low barrel and trapezoidal distortion, and extremely low stray light. After acquiring the unknown mineral to be sampled, the mineral surface is pre-processed: surface sediment, debris, and other impurities are removed to avoid interference from external contaminants in spectral acquisition. Simultaneously, the laboratory hyperspectral imager and sampling lamp were started to perform lamp preheating and initial calibration of the hyperspectral imager. This included equipment leveling, exposure parameter (integration time) calibration, and dark current acquisition to ensure the acquired images were free of geometric distortion and radiometric bias. After the hyperspectral imager parameters were adjusted, the mineral was placed on the sampling platform, and its position was adjusted so that it was directly below the hyperspectral imager's field of view. The camera was then used to take pictures. After the mineral sampling was completed, the image data of the acquired mineral was exported as an HDR image data file and a corresponding RAW header file. After obtaining the mineral image data acquired by the hyperspectral imager, professional remote sensing image processing software was used to preprocess the image data. The image data preprocessing steps are as follows: First, using a band processing tool (band math), the DN value (digital number) of the acquired image was converted into reflectance based on the quantization value of the hyperspectral image using the following formula (taking a quantization value of 255 as an example).

[0030] Here, b1 represents a single-band image, and the float function is used to preserve the decimal precision of reflectance. The image is then linearly stretched by a percentage, cropping the extreme values ​​at both ends of the histogram by 0.5%-1.0%, dynamically stretching the remaining pixel values ​​to the display range. This effectively suppresses the interference of noise, light spots, and other outliers on the overall image contrast, highlighting the details of the acquired image. Next, based on the location of the collected minerals in the image, a Region of Interest (ROI) is selected, and the average spectrum within the selected ROI is calculated to obtain the spectral curve of the minerals in the acquired image. When selecting the ROI, the selected region needs to be moderate; it should not be too large or too small. An overly large region will weaken the characteristic features of the mineral spectrum, while an overly small region will inevitably result in extreme values. After obtaining the spectral curve of the collected minerals by selecting the ROI and calculating the average spectrum within the region, the spectral curve of the minerals is subjected to envelope removal processing. Envelope removal can eliminate the interference of background trends in the image, highlight the spectral characteristics of the collected minerals, and make the mineral spectral curves comparable, laying the foundation for subsequent verification processing methods. For example, 12 common mineral specimens were selected: dolomite, hematite, olivine, calcite, kaolinite, pyrite, jaundice, apatite, siderite, chlorite schist, soapstone, and fibrous gypsum. The constructed standard reference spectral library is as follows: Figure 2 As shown in the figure. Six minerals—hematite, calcite, kaolinite, pyrite, chlorite, and soapstone—were selected for comparison. The spectral curve comparison is shown in the figure below. Figure 3 As shown in the comparison chart, the processing accuracy of the spaceborne data proposed in this embodiment of the invention is high, and the identified minerals are consistent with the corresponding minerals in the standard reference spectral library.

[0031] According to the mineral identification method based on a hyperspectral camera provided by the present invention, compared with the current method of mineral identification through manual detection, the present invention can ensure the quality of remote sensing image acquisition and storage by reasonably setting the payload parameters of the spaceborne hyperspectral camera before mineral identification, thereby ensuring the subsequent mineral identification effect; by enhancing the quality of the remote sensing image, noise in the image can be removed, improving image quality and thus improving the accuracy of subsequent mineral identification; by extracting endmember spectra by referencing spectral bands and purity pixel index, mixed pixels can be automatically filtered to ensure that the extracted endmember spectra have high purity, and mineral identification can be performed based on high-purity endmember spectra, thereby improving the accuracy of mineral identification; by matching the endmember spectra with standard spectra in a preset endmember spectral library to identify minerals, the physical interpretability of the classification results can be ensured, further improving the accuracy of mineral identification.

[0032] Furthermore, to better illustrate the above-described mineral identification process, and as a refinement and extension of the above embodiments, this invention provides another mineral identification method based on a hyperspectral camera, such as... Figure 4 As shown, the method includes: 201. In response to the identification signal of the mineral to be identified in the target area, determine the payload parameters of the spaceborne hyperspectral camera required to identify the mineral.

[0033] Specifically, the payload parameters of the spaceborne hyperspectral camera, such as spectral range, compression parameters, signal-to-noise ratio (SNR), and geometric distortion, are determined. For example, the performance standards for a spaceborne hyperspectral camera should meet the following requirements: spectral range of 400 nm-2450 nm; spectral bandwidth of better than 20 nm; instantaneous field of view better than 3 mrad; SNR measured under laboratory calibration conditions, with a peak SNR better than 1000:1; and center wavelength position accuracy better than 2 nm. For a spaceborne hyperspectral camera, the spatial resolution is 30 m, the spectral range is 400-2500 nm, the spectral resolution after subdivision is ≤5 nm for visible-near infrared (VNIR) and ≤10 nm for short-wave infrared (SWIR), the number of spectral channels is 330, the swath width is 60 km, the total field of view is approximately 4.9°, the center wavelength accuracy is better than ±0.5 nm for visible-near infrared and better than ±1.0 nm for short-wave infrared, the pushbroom imaging mode is used, the quantization bit is 12 bits, and the design lifetime is 5. Year; the Ziyuan-1 02D satellite has an orbital altitude of 778km, a spatial resolution of 30m, a spectral range of 400-2500nm, a spectral resolution of 10nm for visible-near infrared (VNIR) and 20nm for short-wave infrared (SWIR), 166 spectral channels (76 for VNIR and 90 for SWIR), a swath width of 60km, a total field of view of about 4.5°, a center wavelength accuracy better than ±1nm, and also adopts a pushbroom imaging mode with a quantization bit depth of 12 bits. Considering the unique data storage method of the camera used, its weight can be reduced, and it is expected to be similar in weight to the CubeSat hyperspectral camera, such as an estimated weight of around 4 kg; the satellite orbital altitude is set at around 500 km, in a sun-synchronous orbit; the spatial resolution is consistent with that of the CubeSat; the number of spectral channels is planned to be 120 (VNIR60+SWIR60); the spectral resolution is consistent with that of the onboard hyperspectral camera, VNIR 10 nm, SWIR 15 nm; swath width is 18 km; total field of view is around 2°; center wavelength accuracy is better than ±0.5 nm for VNIR and better than ±1.0 nm for SWIR; pushbroom imaging mode is adopted; quantization bit depth is 12 bits; and the designed service life is 2 years.

[0034] 202. Use a spaceborne hyperspectral camera with payload parameters to capture images of the target area to obtain spaceborne remote sensing images of the target area.

[0035] Specifically, the spaceborne hyperspectral camera is designed based on the payload parameters, and the designed spaceborne hyperspectral camera is used to acquire images of the target area.

[0036] 203. Obtain the digital elevation model of the shooting area corresponding to the satellite remote sensing image, determine the terrain occlusion factor and sky view factor of the shooting area based on the digital elevation model, and perform radiometric correction on the satellite remote sensing image based on the terrain occlusion factor and sky view factor to obtain the radiometrically corrected satellite remote sensing image.

[0037] In this embodiment of the invention, to improve the image quality of spaceborne remote sensing images, radiometric correction of the spaceborne remote sensing images is first required. During radiometric correction, the terrain occlusion factor needs to be determined first. Therefore, step 203 specifically includes: determining the terrain slope, terrain aspect, solar azimuth angle, solar altitude angle, and critical slope of the shooting area based on the digital elevation model; and determining whether the shooting area meets the preset conditions for calculating the terrain occlusion factor based on the terrain aspect, solar azimuth angle, terrain slope, and critical slope. The method for determining whether the shooting area meets the preset conditions for calculating the terrain occlusion factor includes: determining whether the absolute difference between the terrain aspect and the solar azimuth angle is greater than a predetermined value. A threshold is set, and / or it is determined whether the terrain slope is greater than the critical slope. If the absolute difference is greater than the preset threshold, and / or the terrain slope is greater than the critical slope, then the shooting area is determined to meet the preset conditions for calculating the terrain occlusion factor. If the absolute difference is less than or equal to the preset threshold, and the terrain slope is less than or equal to the critical slope, then the shooting area is determined to not meet the preset conditions for calculating the terrain occlusion factor. If the shooting area meets the preset conditions for calculating the terrain occlusion factor, then based on the terrain slope, terrain aspect, solar azimuth angle, and solar altitude angle, evaluation parameters for the terrain occlusion factor are determined. Based on the evaluation parameters, the terrain occlusion factor of the shooting area is determined. Among them, the preset threshold is set according to actual needs; the terrain slope is the ground tilt angle at the location of the pixel; the terrain aspect is the orientation of the pixel's slope; the solar azimuth angle is the direction of the projection of sunlight on the horizontal plane; the solar altitude angle is the angle between sunlight and the horizontal plane; and the critical slope is the threshold for occlusion determination.

[0038] Specifically, the terrain shading factor is used to quantify the shading effect of terrain on direct solar radiation. If shading exists in the target area, the terrain shading factor needs to be determined. The evaluation parameters of the terrain shading factor are then calculated using the following formula. :

[0039] in, For solar altitude angle, For terrain slope, For the solar azimuth angle, This refers to the slope of the terrain. Furthermore, if... If the terrain occlusion factor is 0, then the terrain occlusion factor is 0. Then the terrain occlusion factor equals .

[0040] Simultaneously, it is also necessary to determine the sky visibility factor. Based on this, the method includes: determining a global digital elevation model, wherein the global digital elevation model includes a digital elevation model of the shooting area and a digital elevation model of the adjacent areas corresponding to the shooting area, and determining the center point coordinates of the target pixel in the shooting area and the neighborhood pixel coordinates of the adjacent areas based on the global digital elevation model; dividing the sky hemisphere into multiple directions and generating a unit direction vector for each direction; emitting a virtual ray from the center point coordinates to the direction indicated by each unit direction vector; determining the intersection point of the corresponding virtual ray with the adjacent area; determining the distance between the center point coordinates and each intersection point; and determining the sky visibility factor of the shooting area based on the distance.

[0041] Specifically, a high-resolution digital elevation model (DEM) must include the target pixel and its surrounding neighborhood. Three-dimensional coordinate transformation: Convert the two-dimensional raster data of the DEM into a three-dimensional coordinate system (such as UTM projection + elevation) to determine the center point coordinates of each pixel. Divide the sky hemisphere into multiple directions. For each sampling direction, generate a unit direction vector. From the center point o of the target pixel towards each sky direction A virtual line of sight (virtual ray) is emitted, and the intersection of the line of sight and the surrounding terrain is detected. Then, the occlusion evaluation parameters are determined according to the following formula. :

[0042] in, This represents the distance between the center point coordinates and the intersection point. (This is a parameter used for occlusion evaluation.) If the value is greater than the preset occlusion threshold (the preset occlusion threshold is set according to actual needs), then it is determined that there is occlusion in the direction indicated by the unit direction vector. If the occlusion evaluation parameter is... If the value is less than or equal to a preset occlusion threshold, then it is determined that the direction i indicated by the unit direction vector is not occluded; if occlusion exists, then... Set to 0; if there is no obstruction, then... Set it to 1. Then calculate the sky visibility factor according to the following formula. :

[0043] Where N is the total number of unit direction vectors. For direction The horizon function, Let be the zenith angle (solar altitude angle) in the i-th direction. Let be the solar azimuth angle in the i-th direction. This represents the increment of the solar solid angle in the i-th direction.

[0044] Furthermore, after calculating the terrain occlusion factor and the sky visibility factor, radiometric correction is performed on the spaceborne remote sensing image according to the following formula:

[0045] in, This is the radiation correction factor. The original radiation value. Fill in the blank with the value of scattered radiation. Where is the effective angle of solar incidence, and S is the terrain shading factor. The sky visibility factor is ultimately determined using the radiometric correction coefficient. Perform radiometric correction on the image.

[0046] In another embodiment of the present invention, the radiation correction module can be called in the ENVI software toolbar, the radiation calibration tool can be selected, and the FLAASH settings can be applied in the settings interface to ensure that the parameters are consistent with those of the subsequent atmospheric correction. The save path can be set, and the radiation correction can be completed by clicking OK.

[0047] 204. Based on the spatial matching relationship between the digital elevation model and the radiometrically corrected spaceborne remote sensing image, determine the average altitude of the shooting area, the meteorological data of the shooting area, the camera attribute information of the spaceborne hyperspectral camera, and the center wavelength of the radiometrically corrected spaceborne remote sensing image. Based on the average altitude, meteorological data, camera attribute information, and center wavelength, perform atmospheric correction on the radiometrically corrected spaceborne remote sensing image to obtain the atmospherically corrected spaceborne remote sensing image.

[0048] The meteorological data includes temperature, rainfall, humidity, and other data; the camera attribute information includes distortion coefficient, focal length, signal-to-noise ratio, camera type, camera height, and other data.

[0049] Specifically, for example, since atmospheric correction is sensitive to surface elevation, the average elevation of the study area must be calculated first, and then the core correction operation must be performed: the average elevation will use the 900m resolution DEM data included in the ENVI software. The statistical data will be calculated using the Statistics module in the toolbar, selecting the included 900m resolution data, and using the radiometrically corrected image as the input file. Based on the spatial matching relationship between the DEM and the image, the average elevation of the study area will be automatically calculated and the elevation result will be output. Afterwards, the FLAASH Atmosphere Correction tool in the Radiometric Correction module of the toolbar will be used. In the top bar, set the output image path; in the middle bar, select camera attributes such as Sensor Type, and set the Sensor Altitude, Ground Elevation, and Pixel Size; in the bottom bar, based on the image's shooting location, set the Atmospheric Model and enable Water Retrieval, setting the center wavelength to the water vapor absorption sensitive band (1135nm) in the hyperspectral data, leaving other settings as default; finally, select Advanced settings, disable Use TiledProcessing, leave other settings as default, and click Apply to complete atmospheric correction. It should be noted that the above examples are merely illustrative and do not specifically limit the embodiments of the present invention.

[0050] 205. Determine the histograms of each band in the atmospherically corrected spaceborne remote sensing image. Based on the histograms of each band and the shooting process information of the spaceborne hyperspectral camera, determine the anomalous image bands in the atmospherically corrected spaceborne remote sensing image. Determine the number and clustering degree of the anomalous image bands. Based on the number and clustering degree of the bands, determine the corrected half-width of the anomalous image bands. Based on the corrected half-width, select reference image bands around the anomalous image bands. Based on the band values ​​of the reference image bands, perform bad line processing on the anomalous image bands to obtain the bad line processed spaceborne remote sensing image.

[0051] Among them, shooting process information refers to information such as the frequency band in which hardware failures occur during the shooting process.

[0052] Specifically, when processing a 10-band remote sensing image (bands 1-10), a preliminary inspection reveals bad lines (abnormal bands) in bands 3 and 7 caused by hardware malfunctions. The histograms of each band are then examined using the "statistical tools" of image analysis software (such as ENVI or QGIS). For example, band 3 shows abnormally high values ​​in rows 120-130 (normal range 50-150, abnormal value reaches 500), and band 7 shows abnormally low values ​​in rows 200-210 (normal range 200-300, abnormal value as low as 20). Combined with hardware logs, it is confirmed that the sensor experienced a temporary malfunction in the readout circuits of bands 3 and 7 during imaging, causing data distortion in the corresponding rows. Therefore, abnormal image bands can be identified using the above method. For example, the abnormal area in band 3 is concentrated in rows 120-130 (11 rows in total), with a high degree of abnormal value clustering; the correction half-width is set to 2 (i.e., replacing it with the average of the two rows before and after). The abnormal area in band 7 is rows 200-210 (11 rows in total), but the outlier values ​​are scattered. The correction half-width is set to 1 (replacing the value with the average of the preceding and following rows). For outliers in rows 120-130 (e.g., a value of 500 in row 125), with a half-width of 2, the average of the normal values ​​in rows 123-127 (two rows before and after) (e.g., 140 in row 123, 145 in row 124, 138 in row 126, and 135 in row 127) is taken as (140+145+138+135) / 4 = 139.5, and this average is used to replace the 500 in row 125. In other words, the average value is used to replace the bad lines in the abnormal band. After replacement, the values ​​in rows 120-130 transition naturally with adjacent rows without abrupt changes. Therefore, the above method can be used to process bad lines in abnormal image bands.

[0053] In another embodiment of the present invention, first check which bands in the image are abnormal due to hardware failure. Then, use Raster Management - Raster Dicer - Replace bad lines in the toolbar, input the abnormal bands, and set the half-width to the average value. This is generally set according to the number and clustering degree of outliers (usually 1-3). Finally, select Confirm to complete the removal of bad lines.

[0054] 206. Identify the abnormal bands in the corresponding bands of the bad line processed spaceborne remote sensing image, and determine the occurrence period of the abnormal bands based on their positions. Based on the total number of bands and the occurrence period of the bands in the bad line processed spaceborne remote sensing image, remove the abnormal bands in the bad line processed spaceborne remote sensing image to obtain the strip-removed spaceborne remote sensing image.

[0055] Specifically, by stretching the histogram of a specific band in the image display software, alternating bright and dark vertical stripes were observed in the image. The stripe intervals were measured: for example, starting from row 1, a brightness abrupt change occurred every 16 rows (e.g., row 1 is bright, row 17 is dark, row 33 is bright, and so on). This confirmed that the stripe period was 16 rows, caused by inconsistent responses from all detectors (assuming the hyperspectral camera has 16 detector units). Statistical verification: The mean values ​​of rows 1-16 and 17-32 were extracted, revealing a significant difference between the two sets of mean values ​​(e.g., the first set's mean was 150, and the second set's mean was 120), confirming the existence of stripe periodicity. The stripe removal tool was then activated: Tool path: In ENVI, QGIS, or professional remote sensing software, click: Raster Management → Stripe Removal. Parameter settings: Number of Detectors: Enter the stripe period number 16 (i.e., the number of sensor detector units or the stripe repetition period). The image is divided into multiple sub-blocks according to the striping period (16 rows) (e.g., rows 1-16, rows 17-32, etc.). A histogram is calculated for each sub-block, and its grayscale distribution is adjusted to match the global histogram. For example, the dark stripes in rows 17-32 are linearly stretched so that their mean is consistent with rows 1-16, while preserving local details, thus achieving stripe removal from spaceborne remote sensing images. Alternatively, in the stripe removal tool's toolbar, select Raster Management, then Destripe, enter the stripe occurrence period in the Number of Detectors field, and set the save path to complete the stripe removal process.

[0056] 207. Dimensionality reduction processing is performed on the spaceborne remote sensing image after strip removal to obtain a quality-enhanced spaceborne remote sensing image.

[0057] Specifically, the minimum noise fractionation method is used to reduce the dimensionality of the data. For example, in the Transform module of the dimensionality reduction software toolbox, select Forward MNF Estimate Noise Statistics, save the output path, and perform MNF (Minimum Noise Fraction) transformation to reduce the dimensionality of the data.

[0058] 208. Select a preset number of reference spectral bands from the preset spectral set and determine the purity pixel index. Based on the reference spectral bands and the purity pixel index, extract endmember spectra from the enhanced spaceborne remote sensing images.

[0059] Specifically, the endmember spectra of the image are extracted using the Pixel Purity Index from the spectral module in the toolbox of the endmember spectral extraction software. A preset number of spectral bands are selected from the spectral subset (the remaining bands represent noise after MNF transformation) to extract pure pixels from the image. For example, the iteration count is set to 10,000, and the threshold coefficient is set to 3. Then, by creating a new region of interest, only pixels with a PPI greater than 10 are retained as endmember spectra.

[0060] 209. Match the endmember spectra with the standard endmember spectra corresponding to different minerals in the preset endmember spectral library, and determine the mineral types in the target area based on the matching results.

[0061] In this embodiment of the invention, for mineral identification, it is necessary to match the endmember spectrum with the standard endmember spectra corresponding to different minerals in a preset endmember spectral library. Therefore, step 209 specifically includes: setting a preset band threshold for each spectral band in the endmember spectrum, generating a binary spectrum of the endmember spectrum based on the preset band threshold, calculating the Euclidean distance between the binary spectrum and the standard binary spectrum of the standard endmember spectrum, determining the binary matching result between the endmember spectrum and the standard endmember spectrum based on the Euclidean distance, determining the spectral angle between the endmember spectrum and each standard endmember spectrum, and determining the... The spectral angle matching result between the endmember spectrum and the standard endmember spectrum is determined; the spectral feature data of the endmember spectrum and the standard spectral feature data of each standard endmember spectrum are determined, and based on the spectral feature data and the standard spectral feature data, spectral feature fitting matching is performed between the spectral feature data and each standard endmember spectrum to obtain the spectral feature fitting matching result; the weight coefficients corresponding to the binary matching result, the spectral angle matching result, and the spectral feature fitting matching result are determined respectively, and based on the weight coefficients, the binary matching result, the spectral angle matching result, and the spectral feature fitting matching result are weighted and summed to obtain the comprehensive matching result.

[0062] The preset band thresholds are set according to actual needs; the spectral feature data include the reflectance, radiance, absorption band position and width, spectral slope, and spectral shape of each pixel in the remote sensing image in different bands.

[0063] Specifically, a threshold (e.g., average spectral brightness) is set for each band value; values ​​above the threshold are recorded as 1, and values ​​below are recorded as 0. A binary sequence is generated. The encoded sequences of the test spectrum and the standard endmember spectrum in the library are compared based on the minimum distance algorithm. The smaller the Euclidean distance, the higher the similarity (the higher the matching degree), thus obtaining the binary matching result. The spectrum is treated as a multi-dimensional spatial vector. The generalized angle similarity between the test spectrum and the standard endmember spectrum is calculated. The smaller the angle, the higher the matching degree, thus obtaining the spectral angle matching result. The eigenvectors corresponding to the spectral feature data and the standard eigenvectors corresponding to the standard spectral feature data are determined. Based on the eigenvectors and standard eigenvectors, the cosine similarity between the spectral feature data and each of the standard endmember spectra is calculated, thus obtaining the spectral feature fitting matching result. Finally, the binary matching result, the spectral angle matching result, and the spectral feature fitting matching result are weighted and summed to obtain the comprehensive matching result. Finally, the mineral species corresponding to the standard endmember spectrum with the highest comprehensive matching degree is selected as the mineral species in the target area.

[0064] According to another mineral identification method based on a hyperspectral camera provided by the present invention, compared with the current method of mineral identification through manual detection, the present invention, before mineral identification, can ensure the quality of remote sensing image acquisition and storage by rationally designing the payload parameters of the spaceborne hyperspectral camera, thereby ensuring the subsequent mineral identification effect; by enhancing the quality of the remote sensing image, noise in the image can be removed, improving image quality and thus improving the accuracy of subsequent mineral identification; by extracting endmember spectra through reference spectral bands and purity pixel index, mixed pixels can be automatically filtered to ensure that the extracted endmember spectra have high purity, and mineral identification can be performed based on high-purity endmember spectra, thereby improving the accuracy of mineral identification; by matching the endmember spectra with standard spectra in a preset endmember spectral library to identify minerals, the physical interpretability of the classification results can be ensured, further improving the accuracy of mineral identification.

[0065] Furthermore, as Figure 1 In specific implementation, embodiments of the present invention provide a mineral identification device based on a hyperspectral camera, such as... Figure 5 As shown, the device includes: a determining unit 31, an enhancing unit 32, an extracting unit 33, and a matching unit 34.

[0066] The determining unit 31 can be used to determine the payload parameters of the spaceborne hyperspectral camera required to identify the mineral to be identified in response to the identification signal of the mineral to be identified in the target area.

[0067] The enhancement unit 32 can be used to take pictures of the target area using a spaceborne hyperspectral camera with the payload parameters, to obtain a spaceborne remote sensing image of the target area, and to perform quality enhancement processing on the spaceborne remote sensing image to obtain a quality-enhanced spaceborne remote sensing image.

[0068] The extraction unit 33 can be used to select a preset number of reference spectral bands in a preset spectral set and determine the clean pixel index. Based on the reference spectral bands and the clean pixel index, it can extract endmember spectra from the enhanced spaceborne remote sensing image.

[0069] The matching unit 34 can be used to match the endmember spectrum with the standard endmember spectra corresponding to different minerals in a preset endmember spectrum library, and determine the mineral types in the target region based on the matching results.

[0070] In specific application scenarios, in order to determine the payload parameters of the spaceborne hyperspectral camera required for identifying the minerals to be identified, the determining unit 31 can be specifically used to determine the absorption characteristic data of various minerals, the regional characteristic data of the target area, and the mineral identification requirement data; based on the absorption characteristic data, the regional characteristic data, and the mineral identification requirement data, determine the spectral range of the spaceborne hyperspectral camera; determine the reference ground features of the target area, determine the spectral characteristics of the reference ground features, and determine the spectral correlation of adjacent bands in the reference hyperspectral data of the reference ground features; based on the spectral characteristics of the ground features and the spectral correlation, determine the image compression parameters of the spaceborne hyperspectral camera; acquire the performance indicators, image acquisition requirements, and reference remaining payload parameters of multiple reference image acquisition devices; and based on the performance indicators, image acquisition requirements, and reference remaining payload parameters of each reference image acquisition device, determine the camera remaining payload parameters of the spaceborne hyperspectral camera, wherein the camera remaining payload parameters are payload parameters other than the spectral range and the image compression parameters.

[0071] In specific application scenarios, in order to enhance the quality of remote sensing images, such as... Figure 6 As shown, the enhancement unit 32 includes a radiation correction module 321, an atmospheric correction module 322, a bad line processing module 323, a stripe removal module 324, and a dimension reduction module 325.

[0072] The radiometric correction module 321 can be used to acquire a digital elevation model of the shooting area corresponding to the satellite remote sensing image, determine the terrain occlusion factor and sky view factor of the shooting area based on the digital elevation model, and perform radiometric correction on the satellite remote sensing image based on the terrain occlusion factor and the sky view factor to obtain the radiometrically corrected satellite remote sensing image.

[0073] The atmospheric correction module 322 can be used to determine the average altitude of the shooting area based on the spatial matching relationship between the digital elevation model and the radiometrically corrected spaceborne remote sensing image, determine the meteorological data of the shooting area, the camera attribute information of the spaceborne hyperspectral camera, and the center wavelength of the radiometrically corrected spaceborne remote sensing image, and perform atmospheric correction on the radiometrically corrected spaceborne remote sensing image based on the average altitude, the meteorological data, the camera attribute information, and the center wavelength to obtain the atmospherically corrected spaceborne remote sensing image.

[0074] The bad line processing module 323 can be used to determine the histograms of each band of the atmospherically corrected spaceborne remote sensing image, and based on the histograms of each band and the shooting process information of the spaceborne hyperspectral camera, determine the abnormal image bands in the atmospherically corrected spaceborne remote sensing image, determine the number of bands and the degree of band clustering of the abnormal image bands, determine the corrected half-width of the abnormal image bands based on the number of bands and the degree of band clustering, select reference image bands around the abnormal image bands based on the corrected half-width, and perform bad line processing on the abnormal image bands based on the band values ​​of the reference image bands to obtain the spaceborne remote sensing image after bad line processing.

[0075] The strip removal module 324 can be used to identify abnormal stripes in the stripes corresponding to the bad line processed spaceborne remote sensing image, and determine the strip occurrence period of the abnormal stripes based on the position of the abnormal stripes. Based on the total number of bands in the bad line processed spaceborne remote sensing image and the strip occurrence period, the abnormal stripes in the bad line processed spaceborne remote sensing image are removed to obtain the strip-removed spaceborne remote sensing image.

[0076] The dimensionality reduction module 325 can be used to perform dimensionality reduction processing on the spaceborne remote sensing image after strip removal to obtain the spaceborne remote sensing image with enhanced quality.

[0077] In specific application scenarios, to determine the terrain occlusion factor and sky visibility factor of the shooting area, the radiometric correction module 321 can be specifically used to determine the terrain slope, terrain aspect, solar azimuth, solar altitude angle, and critical slope of the shooting area based on the digital elevation model; based on the terrain aspect, solar azimuth, terrain slope, and critical slope, determine whether the shooting area meets the preset conditions for calculating the terrain occlusion factor. The method for determining whether the shooting area meets the preset conditions for calculating the terrain occlusion factor includes: determining whether the absolute difference between the terrain aspect and the solar azimuth is greater than a preset threshold, and / or determining whether the terrain slope is greater than the critical slope. If the absolute difference is greater than the preset threshold, and / or the terrain slope is greater than the critical slope, then the shooting area is determined to meet the preset conditions for calculating the terrain occlusion factor. If the absolute difference is less than or equal to the preset threshold, and the terrain slope is less than or equal to the critical slope, then the shooting area is determined not to meet the preset conditions for calculating the terrain occlusion factor. The preset conditions for the shooting area are as follows: If the shooting area meets the preset conditions for calculating the terrain occlusion factor, then the evaluation parameters of the terrain occlusion factor are determined based on the terrain slope, terrain aspect, solar azimuth angle, and solar altitude angle. Based on the evaluation parameters, the terrain occlusion factor of the shooting area is determined. The sky viewing angle factor of the shooting area is determined based on the digital elevation model, including: determining the global digital elevation model, wherein the global digital elevation model includes the digital elevation model of the shooting area and the digital elevation model of the adjacent areas corresponding to the shooting area, and determining the center point coordinates of the target pixel in the shooting area and the neighborhood pixel coordinates of the adjacent areas based on the global digital elevation model; dividing the sky hemisphere into multiple directions and generating a unit direction vector for each direction, emitting a virtual ray from the center point coordinates to the direction indicated by each unit direction vector, and determining the intersection point of the corresponding virtual ray with the adjacent area; determining the distance between the center point coordinates and each intersection point, and determining the sky viewing angle factor of the shooting area based on the distance.

[0078] In specific application scenarios, in order to extract endmember spectra from enhanced spaceborne remote sensing images, the extraction unit 33 includes a sorting module 331 and a calculation module 332.

[0079] The sorting module 331 can be used to sort each pixel in the enhanced satellite remote sensing image based on the clean pixel index of each pixel in the enhanced satellite remote sensing image, and select a preset number of pixels as initial end pixels in each sorted pixel.

[0080] The calculation module 332 can be used to calculate the spectral matching degree between each initial endmember and the spectrum corresponding to the reference spectral band, and based on the spectral matching degree, select a target endmember in each initial endmember, and use the spectrum of the target endmember as the endmember spectrum in the spaceborne remote sensing image.

[0081] In specific application scenarios, in order to match the endmember spectrum with the standard endmember spectra corresponding to different minerals in the preset endmember spectrum library, the matching unit 34 includes a binary matching module 341, a spectral angle matching module 342, a spectral feature matching module 343, and a weighting module 344.

[0082] The binary matching module 341 can be used to set a preset band threshold for each spectral band in the endmember spectrum, generate a binary spectrum of the endmember spectrum based on the preset band threshold, calculate the Euclidean distance between the binary spectrum and the standard binary spectrum of the standard endmember spectrum, and determine the binary matching result between the endmember spectrum and the standard endmember spectrum based on the Euclidean distance.

[0083] The spectral angle matching module 342 can be used to determine the spectral angle between the endmember spectrum and each of the standard endmember spectra, and based on the spectral angle, determine the spectral angle matching result between the endmember spectrum and the standard endmember spectrum.

[0084] The spectral feature matching module 343 can be used to determine the spectral feature data of the endmember spectrum and the standard spectral feature data of each standard endmember spectrum, and based on the spectral feature data and the standard spectral feature data, perform spectral feature fitting and matching between the spectral feature data and each standard endmember spectrum to obtain the spectral feature fitting and matching result.

[0085] The weighting module 344 can be used to determine the weight coefficients corresponding to the binary matching result, the spectral angle matching result, and the spectral feature fitting matching result, respectively. Based on the weight coefficients, the binary matching result, the spectral angle matching result, and the spectral feature fitting matching result are weighted and summed to obtain a comprehensive matching result.

[0086] In specific application scenarios, in order to determine the camera spectral range of the spaceborne hyperspectral camera based on absorption feature data, regional feature data, and mineral identification requirement data, the determining unit 31 includes a determining module 311, a cross module 312, and a spectral range prediction module 313.

[0087] The determining module 311 can be used to determine the absorption feature vector corresponding to the absorption feature data, the regional feature vector corresponding to the regional feature data, and the demand feature vector corresponding to the mineral identification demand data, respectively.

[0088] The cross-processing module 312 can be used to perform feature-level cross-processing on the absorption feature vector, the regional feature vector, and the demand feature vector to obtain a feature cross-vector; perform element-level cross-processing on the absorption feature vector, the regional feature vector, and the demand feature vector to obtain an element cross-vector; perform low-order cross-processing on the absorption feature vector, the regional feature vector, and the demand feature vector to obtain a low-order cross-vector; and use a preset transformation function to transform the feature cross-vector, the element cross-vector, and the low-order cross-vector to obtain a spectral cross-feature vector.

[0089] The spectral range prediction module can be used to input the spectral cross feature vector into a preset spectral range prediction model to predict the spectral range and obtain the camera spectral range of the spaceborne hyperspectral camera.

[0090] It should be noted that other corresponding descriptions of the functional modules involved in the mineral identification device based on a hyperspectral camera provided in this embodiment of the invention can be found in [reference needed]. Figure 1 The corresponding description of the method shown will not be repeated here.

[0091] Based on the above, Figure 1 Accordingly, this embodiment of the invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, performs the following steps: responding to an identification signal of a mineral to be identified in a target area, determining the payload parameters of a spaceborne hyperspectral camera required to identify the mineral; using the spaceborne hyperspectral camera with the payload parameters to capture an image of the target area, obtaining a spaceborne remote sensing image of the target area, and performing quality enhancement processing on the spaceborne remote sensing image to obtain a quality-enhanced spaceborne remote sensing image; selecting a preset number of reference spectral bands in a preset spectral set and determining a purity pixel index; extracting endmember spectra from the quality-enhanced spaceborne remote sensing image based on the reference spectral bands and the purity pixel index; matching the endmember spectra with standard endmember spectra corresponding to different minerals in a preset endmember spectral library, and determining the mineral type in the target area based on the matching results.

[0092] Based on the above, Figure 1 The method shown and as Figure 5 The embodiment of the device shown in the invention also provides a physical structure diagram of a computer device, such as... Figure 7As shown, the computer device includes: a processor 41, a memory 42, and a computer program stored in the memory 42 and executable on the processor. Both the memory 42 and the processor 41 are mounted on a bus 43. When the processor 41 executes the program, it performs the following steps: in response to an identification signal of a mineral to be identified in a target area, it determines the payload parameters of the spaceborne hyperspectral camera required to identify the mineral; it uses the spaceborne hyperspectral camera with the payload parameters to capture an image of the target area, obtaining a spaceborne remote sensing image of the target area, and performs quality enhancement processing on the spaceborne remote sensing image to obtain a quality-enhanced spaceborne remote sensing image; it selects a preset number of reference spectral bands from a preset spectral set and determines the purity pixel index; based on the reference spectral bands and the purity pixel index, it extracts endmember spectra from the quality-enhanced spaceborne remote sensing image; it matches the endmember spectra with standard endmember spectra corresponding to different minerals in a preset endmember spectral library, and based on the matching results, it determines the mineral type in the target area.

[0093] Through the technical solution of this invention, before mineral identification, the payload parameters of the spaceborne hyperspectral camera are reasonably set to ensure the acquisition and storage quality of remote sensing images, thereby guaranteeing the subsequent mineral identification effect. By enhancing the quality of the remote sensing images, noise in the images can be removed, improving image quality and thus enhancing the accuracy of subsequent mineral identification. By extracting endmember spectra using reference spectral bands and purity pixel indices, mixed pixels can be automatically filtered to ensure that the extracted endmember spectra have high purity. Finally, mineral identification is performed based on high-purity endmember spectra, which can improve the accuracy of mineral identification. By matching the endmember spectra with standard spectra in a preset endmember spectral library to identify minerals, the physical interpretability of the classification results can be ensured, further improving the accuracy of mineral identification.

[0094] It is obvious to those skilled in the art that the modules or steps of the present invention described above can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Optionally, they can be implemented using computer-executable program code, thereby storing them in a storage device for execution by a computing device. In some cases, the steps shown or described can be performed in a different order than those presented herein, or they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Thus, the present invention is not limited to any particular combination of hardware and software.

[0095] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A mineral identification method based on a hyperspectral camera, characterized in that, include: In response to the identification signal of the mineral to be identified in the target area, determine the payload parameters of the spaceborne hyperspectral camera required to identify the mineral to be identified; The target area is photographed using a spaceborne hyperspectral camera with the aforementioned payload parameters to obtain a spaceborne remote sensing image of the target area. The spaceborne remote sensing image is then enhanced to obtain a quality-enhanced spaceborne remote sensing image. A preset number of reference spectral bands are selected from the preset spectral set, and the purity pixel index is determined. Based on the reference spectral bands and the purity pixel index, endmember spectra are extracted from the enhanced spaceborne remote sensing image. The endmember spectrum is matched with the standard endmember spectra corresponding to different minerals in a preset endmember spectrum library, and the mineral types in the target region are determined based on the matching results.

2. The method according to claim 1, characterized in that, The determination of the payload parameters of the spaceborne hyperspectral camera required to identify the mineral to be identified includes: The absorption characteristic data of various minerals, the regional characteristic data of the target area, and the mineral identification requirement data are determined. Based on the absorption characteristic data, the regional characteristic data, and the mineral identification requirement data, the spectral range of the spaceborne hyperspectral camera is determined. The reference ground features of the target area are determined, and the spectral characteristics of the reference ground features are determined, as well as the spectral correlation of adjacent bands in the reference hyperspectral data of the reference ground features. Based on the spectral characteristics of the ground features and the spectral correlation, the image compression parameters of the spaceborne hyperspectral camera are determined. The performance indicators, image acquisition requirements, and reference remaining payload parameters of multiple reference image acquisition devices are obtained. Based on the performance indicators, image acquisition requirements, and reference remaining payload parameters of each reference image acquisition device, the camera remaining payload parameters of the spaceborne hyperspectral camera are determined. The camera remaining payload parameters are payload parameters other than the spectral range and the image compression parameters.

3. The method according to claim 1, characterized in that, The process of enhancing the quality of the spaceborne remote sensing image to obtain the enhanced spaceborne remote sensing image includes: A digital elevation model of the shooting area corresponding to the satellite remote sensing image is obtained. Based on the digital elevation model, the terrain occlusion factor and the sky view factor of the shooting area are determined. Based on the terrain occlusion factor and the sky view factor, the satellite remote sensing image is radiometrically corrected to obtain the radiometrically corrected satellite remote sensing image. Based on the spatial matching relationship between the digital elevation model and the radiometrically corrected spaceborne remote sensing image, the average altitude of the shooting area is determined, as are the meteorological data of the shooting area, the camera attribute information of the spaceborne hyperspectral camera, and the center wavelength of the radiometrically corrected spaceborne remote sensing image. Based on the average altitude, the meteorological data, the camera attribute information, and the center wavelength, atmospheric correction is performed on the radiometrically corrected spaceborne remote sensing image to obtain the atmospherically corrected spaceborne remote sensing image. The histograms of each band of the atmospherically corrected spaceborne remote sensing image are determined. Based on the histograms of each band and the shooting process information of the spaceborne hyperspectral camera, the anomalous image bands in the atmospherically corrected spaceborne remote sensing image are determined. The number of bands and the degree of band clustering of the anomalous image bands are determined. Based on the number of bands and the degree of band clustering, the correction half-width of the anomalous image bands is determined. Based on the correction half-width, reference image bands are selected around the anomalous image bands. Based on the band values ​​of the reference image bands, bad line processing is performed on the anomalous image bands to obtain the spaceborne remote sensing image after bad line processing. Anomalous bands are identified in the bands corresponding to the satellite remote sensing image after bad line processing. Based on the position of the anomalous bands, the occurrence period of the anomalous bands is determined. Based on the total number of bands in the satellite remote sensing image after bad line processing and the occurrence period of the bands, the anomalous bands in the satellite remote sensing image after bad line processing are removed to obtain the satellite remote sensing image after band removal. The spaceborne remote sensing image after stripe removal is subjected to dimensionality reduction processing to obtain the enhanced quality spaceborne remote sensing image.

4. The method according to claim 3, characterized in that, Determining the terrain occlusion factor of the shooting area based on the digital elevation model includes: Based on the digital elevation model, the terrain slope, terrain aspect, solar azimuth angle, solar altitude angle, and critical slope of the shooting area are determined. Based on the terrain aspect, the solar azimuth angle, the terrain slope, and the critical slope, it is determined whether the shooting area meets the preset conditions for calculating the terrain occlusion factor. The method for determining whether the shooting area meets the preset conditions for calculating the terrain occlusion factor includes: determining whether the absolute difference between the terrain aspect and the solar azimuth angle is greater than a preset threshold, and / or determining whether the terrain slope is greater than the critical slope. If the absolute difference is greater than the preset threshold, and / or the terrain slope is greater than the critical slope, then the shooting area is determined to meet the preset conditions for calculating the terrain occlusion factor. If the absolute difference is less than or equal to the preset threshold, and the terrain slope is less than or equal to the critical slope, then the shooting area is determined not to meet the preset conditions for calculating the terrain occlusion factor. If the shooting area meets the preset conditions for calculating the terrain occlusion factor, then the evaluation parameters of the terrain occlusion factor are determined based on the terrain slope, the terrain aspect, the solar azimuth angle, and the solar altitude angle, and the terrain occlusion factor of the shooting area is determined based on the evaluation parameters. The sky visibility factor of the shooting area is determined based on the digital elevation model, including: A global digital elevation model is determined, wherein the global digital elevation model includes a digital elevation model of the shooting area and a digital elevation model of the adjacent areas corresponding to the shooting area, and the center point coordinates of the target pixel in the shooting area and the neighborhood pixel coordinates of the adjacent areas are determined based on the global digital elevation model; The sky hemisphere is divided into multiple directions, and a unit direction vector is generated for each direction. Virtual rays are emitted from the coordinates of the center point to the direction indicated by each unit direction vector, and the intersection point of the corresponding virtual ray with the adjacent region is determined. Determine the distance between the center point coordinates and each of the intersection points, and based on the distances, determine the sky viewing angle factor of the shooting area.

5. The method according to claim 1, characterized in that, The step of extracting endmember spectra from the enhanced spaceborne remote sensing image based on the reference spectral band and the clean pixel index includes: Based on the clean pixel index of each pixel in the enhanced satellite remote sensing image, each pixel in the enhanced satellite remote sensing image is sorted, and a preset number of pixels are selected as initial end pixels in each sorted pixel. Calculate the spectral matching degree between each initial endmember and the spectrum corresponding to the reference spectral band. Based on the spectral matching degree, select a target endmember in each initial endmember and use the spectrum of the target endmember as the endmember spectrum in the spaceborne remote sensing image.

6. The method according to claim 1, characterized in that, The step of matching the endmember spectrum with the standard endmember spectra corresponding to different minerals in a preset endmember spectral library includes: A preset band threshold is set for each spectral band in the endmember spectrum, and a binary spectrum of the endmember spectrum is generated based on the preset band threshold. The Euclidean distance between the binary spectrum and the standard binary spectrum of the standard endmember spectrum is calculated, and the binary matching result between the endmember spectrum and the standard endmember spectrum is determined based on the Euclidean distance. Determine the spectral angle between the endmember spectrum and each of the standard endmember spectra, and based on the spectral angle, determine the spectral angle matching result between the endmember spectrum and the standard endmember spectrum; The spectral feature data of the endmember spectrum and the standard spectral feature data of each standard endmember spectrum are determined. Based on the spectral feature data and the standard spectral feature data, spectral feature fitting and matching are performed between the spectral feature data and each standard endmember spectrum to obtain the spectral feature fitting and matching result. The weight coefficients corresponding to the binary matching result, the spectral angle matching result, and the spectral feature fitting matching result are determined respectively. Based on the weight coefficients, the binary matching result, the spectral angle matching result, and the spectral feature fitting matching result are weighted and summed to obtain the comprehensive matching result.

7. The method according to claim 2, characterized in that, The step of determining the camera spectral range of the spaceborne hyperspectral camera based on the absorption feature data, the regional feature data, and the mineral identification requirement data includes: Determine the absorption feature vector corresponding to the absorption feature data, the regional feature vector corresponding to the regional feature data, and the demand feature vector corresponding to the mineral identification demand data, respectively. The absorption feature vector, the regional feature vector, and the demand feature vector are subjected to feature-level cross processing to obtain a feature cross vector. The absorption feature vector, the regional feature vector, and the demand feature vector are subjected to element-level cross processing to obtain an element cross vector. The absorption feature vector, the regional feature vector, and the demand feature vector are subjected to low-order cross processing to obtain a low-order cross vector. The feature cross vector, the element cross vector, and the low-order cross vector are transformed using a preset transformation function to obtain a spectral cross feature vector. The spectral cross feature vector is input into a preset spectral range prediction model to predict the spectral range, thereby obtaining the camera spectral range of the spaceborne hyperspectral camera.

8. A mineral identification device based on a hyperspectral camera, characterized in that, include: The determination unit is used to determine the payload parameters of the spaceborne hyperspectral camera required to identify the mineral to be identified in response to the identification signal of the mineral to be identified in the target area. An enhancement unit is used to capture images of the target area using a spaceborne hyperspectral camera with the payload parameters, to obtain a spaceborne remote sensing image of the target area, and to perform quality enhancement processing on the spaceborne remote sensing image to obtain a quality-enhanced spaceborne remote sensing image. The extraction unit is used to select a preset number of reference spectral bands from a preset spectral set and determine the clean pixel index. Based on the reference spectral bands and the clean pixel index, it extracts endmember spectra from the enhanced spaceborne remote sensing image. The matching unit is used to match the endmember spectrum with the standard endmember spectra corresponding to different minerals in a preset endmember spectrum library, and determine the mineral types in the target region based on the matching results.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

10. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images