Satellite system and method for performing real-time spectroscopic analysis

Abstract

The present invention relates to a satellite system for performing real-time spectroscopic analysis, wherein environmental changes occurring on the surface of the Earth are monitored, and collected spectroscopic data is analyzed and optimized in real time by a satellite itself and transmitted to a ground station, thereby reducing reliance on image processing at the ground station and providing useful insights into the detection of natural resources and environmental dynamics of the Earth. The present invention may comprise a spectral camera, a preprocessor, an extractor, an analyzer, and a communicator.

Description

Satellite system and method for performing real-time spectral analysis

[0001] The present invention relates to a satellite technology for performing real-time spectral analysis, and more particularly, to a satellite system and method for performing real-time spectral analysis by monitoring environmental changes occurring on the Earth's surface, analyzing and optimizing the collected spectral data in real time on the satellite itself, and transmitting it to a ground station, thereby reducing dependence on image processing at the ground station and providing useful insights into the detection of the Earth's natural resources and environmental dynamics.

[0002] Currently, hundreds of artificial satellites around the world are performing their specific tasks in the sky above the Earth for purposes such as meteorology, communications, broadcasting, agriculture, space development, and military.

[0003] Meanwhile, hyperspectral images are defined by three characteristics: many spectral bands, continuous spectral bands, and narrow wavelengths. The difference between hyperspectral images and multispectral images is that the spectral reflectance values ​​of Landsat ETM+ multispectral images are obtained from six discrete wavelength bands, whereas AVIRIS hyperspectral images show spectral reflectance values ​​extracted from 224 continuous bands with a wavelength width of approximately 10 nm in the wavelength range of 400–2,500 nm. In other words, hyperspectral images can be defined as data that can obtain the complete spectral reflectance curve of a surface object corresponding to each pixel that composes the image.

[0004] These hyperspectral images are mounted on aircraft, satellites, and ground vehicles, and are used for target identification at high altitudes or on the ground. Specifically, hyperspectral imagery is primarily used in geology, such as identifying mineral resource distribution and identifying rock types. In the field of vegetation, extensive research is being conducted on estimating plant biochemical components (chlorophyll, carbon, nitrogen, lignin, etc.).

[0005] Korean Patent No. 10-2540762 describes a system that receives images captured by satellites directly from a ground station, analyzes and processes the images, and identifies or monitors targets. However, when receiving data from multiple satellites, the sheer volume of data received from these satellites requires significant processing time. This leads to a drawback: a higher reliance on the ground station leads to inefficient data processing.

[0006] The problem to be solved by the present invention is to provide an artificial satellite system and method that performs real-time spectral analysis to monitor environmental changes occurring on the Earth's surface and to detect the Earth's natural resources by analyzing and optimizing collected spectral data in real time.

[0007] The problem to be solved by the present invention is to provide a satellite system and method for performing real-time spectral analysis, which can reduce dependence on image processing at a ground station by detecting the Earth's natural resources through real-time analysis and optimization on the satellite itself.

[0008] The problem to be solved by the present invention relates to an artificial satellite system and method for performing real-time spectral analysis that provides useful insights into the detection of the Earth's natural resources and environmental dynamics.

[0009] An artificial satellite system for performing real-time spectral analysis according to one embodiment of the present invention may include a spectral camera for photographing the Earth using a multispectral sensor, a hyperspectral sensor, and a hyperspectral sensor, a preprocessor for performing quality correction and data normalization on images taken by each sensor, an extractor for extracting spectral signatures of a plurality of unit bands from each of the preprocessed images, an analyzer for detecting and identifying an object using the extracted spectral signature, and a communicator for transmitting the processing result of the analyzer to a ground station.

[0010] In one embodiment, the quality correction performed by the preprocessor may include at least one of radiation correction, atmospheric correction, and noise removal.

[0011] In one embodiment, the analyzer further includes an image enhancer that fuses spatial context into the extracted spectral signature, corrects atmospheric disturbances in the fused spectral signature, and removes noise and artifacts to improve image quality, and the analyzer can detect and identify an object using the corrected spectral signature.

[0012] In one embodiment, the analyzer may utilize any one of a Spectral Angle Mapper (SAM), an Adaptive Cosine Estimator (ACE), and Support Vector Machines (SVM) for detection and identification of objects.

[0013] In one embodiment, the system may further include a location tracker that determines a location within the image by predicting changes in the location of the identified object over time.

[0014] In one embodiment, the ground station can process the processing results of the analyzer to generate a 2D or 3D image.

[0015] In one embodiment, the object of detection and identification may be at least one of the Earth's natural resources and environmental dynamics.

[0016] In one embodiment, a real-time spectral analysis method for a small artificial satellite system according to another embodiment of the present invention may include a step in which a spectral camera photographs the Earth using a multispectral sensor, a hyperspectral sensor, and a hyperspectral sensor, a step in which a preprocessor performs quality correction and data normalization on images of each of the photographed sensors, a step in which an extractor extracts spectral signatures of a plurality of unit bands from each of the preprocessed images, a step in which an analyzer detects and identifies an object using the extracted spectral signature, and a step in which a communicator transmits a processing result of the analyzer to a ground station.

[0017] In one embodiment, the quality correction performed by the preprocessor may include at least one of radiation correction, atmospheric correction, and noise removal.

[0018] In one embodiment, the image enhancer further comprises a step of fusion of spatial context into the extracted spectral signature, correcting atmospheric disturbances and removing noise and artifacts from the fused spectral signature to enhance image quality, and the analyzer can detect and identify an object using the corrected spectral signature.

[0019] In one embodiment, the analyzer may utilize any one of a Spectral Angle Mapper (SAM), an Adaptive Cosine Estimator (ACE), and Support Vector Machines (SVM) for detection and identification of objects.

[0020] In one embodiment, the location tracker may further include a step of determining a location within the image by predicting a change in location over time of the identified object.

[0021] In one embodiment, the ground station may further include a step of processing the processing result of the analyzer to generate a 2D or 3D image.

[0022] In one embodiment, the object of detection and identification may be at least one of the Earth's natural resources and environmental dynamics.

[0023] According to an embodiment of the present invention, environmental changes occurring on the Earth's surface can be monitored, and the Earth's natural resources can be detected by analyzing and optimizing the collected spectral data in real time on an artificial satellite, thereby reducing dependence on image processing at a ground station.

[0024] FIG. 1 is a schematic diagram illustrating an artificial satellite system performing real-time spectral analysis according to Embodiment 1 of the present invention.

[0025] Figure 2 is a block diagram showing the configuration of the artificial satellite illustrated in Figure 1.

[0026] FIG. 3 is a schematic diagram illustrating an artificial satellite system performing real-time spectral analysis according to Embodiment 2 of the present invention.

[0027] FIG. 4 is a flowchart illustrating an artificial satellite system performing real-time spectral analysis according to Embodiment 3 of the present invention.

[0028] Hereinafter, several embodiments of the present invention will be described in detail using drawings. However, this is not intended to limit the present invention to any specific embodiment, and it should be understood that all transformations, equivalents, and substitutions that incorporate the technical spirit of the present invention are included within the scope of the present invention.

[0029] In this specification, singular expressions include plural expressions unless the context clearly indicates otherwise.

[0030] When a configuration in this specification is described as "having" or "composing" a sub-configuration, it is intended that the configuration may further include other configurations, rather than excluding other configurations, unless otherwise specifically stated.

[0031] In this specification, the terms “Unit,” “Module,” and “Component” mean a unit that processes at least one function or operation, and may be implemented by hardware, software, or a combination of hardware and software.

[0032] The term “connected” in this specification may mean, but is not limited to, that two components are directly connected, and may also mean that the components are connected via one or more other components positioned between the components.

[0033] <Example 1>

[0034] Example 1 relates to a technology that can reduce dependence on image processing at a ground station by automatically adjusting and optimizing spectral data acquired from a spectral camera (110) mounted on an artificial satellite (100) in real time on the artificial satellite (100) and then transmitting the same to a ground station.

[0035] Figure 1 is a schematic diagram illustrating an artificial satellite (100) system that performs real-time spectral analysis according to Embodiment 1 of the present invention.

[0036] In the artificial satellite (100) system of Example 1, the artificial satellite (100) monitors environmental changes occurring on the Earth's surface, and analyzes and optimizes the collected spectral data in real time on the artificial satellite (100) to detect the Earth's natural resources and provide useful insights into environmental dynamics.

[0037] The artificial satellite (100) transmits analyzed and optimized data to the ground station, and the ground station processes the data transmitted from the artificial satellite (100) to create a 2D or 3D image.

[0038] Figure 2 is a block diagram showing the configuration of the artificial satellite (100) illustrated in Figure 1.

[0039] As shown in Fig. 2, the artificial satellite (100) may include a spectroscopic camera (110), a preprocessor (120), an extractor (130), an analyzer (140), and a communication device (150). In addition, it may further include an image enhancer (160) and a position tracker (170).

[0040] In this embodiment, the artificial satellite (100) may be any one of a communication artificial satellite, a weather artificial satellite, a broadcasting artificial satellite, and a scientific artificial satellite.

[0041] The spectral camera (110) acquires spectral data by photographing the Earth using a multispectral sensor, a hyperspectral sensor, and a hyperspectral sensor. For reference, the spectral camera (110) can also photograph the Earth using only a hyperspectral sensor and a hyperspectral sensor.

[0042] Multispectral images captured by a multispectral sensor typically capture images from less than ten discrete bands within the visible and mid-infrared regions. Since the data is stored in a discrete format, independent data corresponding to each pixel is generated. In this embodiment, the multispectral sensor may include a multispectral sensor from Landsat. Multispectral imaging enables the classification of macroscopic objects based on their spectrum.

[0043] Hyperspectral images obtained by hyperspectral sensors typically capture images in more than 100 continuous bands across the visible, near-infrared, mid-infrared, and thermal infrared regions, and extract the entire spectrum from each pixel. Hyperspectral sensors disperse incident light, acquiring tens to hundreds of spectral data points for each pixel in a continuous, narrow wavelength range of an object on the ground. Hyperspectral sensors capture the unique optical properties of each substance, as well as the absorption and reflection characteristics of the substance. Hyperspectral imaging can classify objects based on the chemical composition of solids or liquids.

[0044] Ultra-hyperspectral images, obtained using hyperspectral sensors, typically capture images from over 1,000 continuous bands across the visible, near-infrared, mid-infrared, and thermal infrared regions. Hyperspectral imaging enables analysis of the chemical composition of gases.

[0045] The preprocessor (120) performs quality correction and data normalization on images (i.e., spectral data) captured by each of the multispectral sensor, hyperspectral sensor, and hyperspectral sensor.

[0046] Quality correction by the preprocessor (120) may include at least one of radiation correction, atmospheric correction, and noise removal.

[0047] Radiometric Correction refers to the process of correcting radiometric distortion that appears as a change in the brightness value of an image due to atmospheric and topographical influences, including the sun's elevation angle, the response characteristics of a probe, atmospheric conditions, and topographical factors, when acquiring an image from an artificial satellite (100).

[0048] Atmospheric Correction refers to the process of correcting distortion that occurs in an image when light waves or electromagnetic waves are absorbed or scattered by atmospheric particles when observing the Earth's surface with an artificial satellite (100).

[0049] For example, atmospheric correction may include a process of estimating the amount of scattering, absorption, and reflection corresponding to each wavelength and the intensity of incident solar light by inputting factors such as solar elevation angle and meteorological elements into an atmospheric model. In the process of acquiring a spectral image through an artificial satellite (100), since the sunlight passes through the Earth's atmosphere twice, it may be affected by scattering, absorption, reflection, etc. by the atmosphere, and this atmospheric influence may weaken the intensity of the sunlight incident on the spectral camera (110) sensor of the artificial satellite (100), which may ultimately affect the brightness of the image data. Therefore, atmospheric correction may be performed to secure an accurate artificial satellite (100) image.

[0050] In addition, quality correction by the preprocessor (120) may include geometric correction and orthogonal correction.

[0051] Geometric correction may include a process of projecting each location of a spectral image via an artificial satellite (100) onto a flat surface with appropriate x and y axes. The artificial satellite (100) data may have geometric errors due to the rotation of the Earth, the curvature of the Earth, the attitude and errors of the spectral camera (110) mounted on the artificial satellite (100), etc. Therefore, geometric correction may be performed to secure accurate artificial satellite (100) images.

[0052] Ortho-correction may involve the process of precisely linking satellite (100) image data and ground coordinates to produce quantitative information.

[0053] Image denoising improves image quality through noise removal filters or algorithms. For example, noise can be removed using techniques such as Gaussian filtering, median filtering, bilateral filtering, averaging filtering, median filtering, and adaptive filtering based on local image variance.

[0054] Image noise removal can also be achieved using the Minimum Noise Fraction transform (MNF). This algorithm reduces the number of bands in hyperspectral data and minimizes noise within the image.

[0055] The minimum noise ratio transform (MNTR) is a technique for extracting signals with spatial autocorrelation within an image. This allows for the separation of the effects of artificial image signals, such as noise, from the original image. For example, noise-induced image signals typically have unique spatial patterns clearly distinguishable from those of natural fluctuations and exhibit high spatial autocorrelation, making them easily extracted using the minimum noise ratio transform (MNTR).

[0056] The minimum noise ratio transform (MNTR) method applies a two-step principal component analysis (PCA). The first transform reduces and rescales the data's noise to produce uncorrelated data across bands. The second transform generates a noise-removed MNF eigenimage and a noise-included MNF eigenimage.

[0057] Principal component analysis (PCA), applied to the minimum noise ratio transformation method, is one of the methods for efficiently summarizing multidimensional data with correlations between variables into low-dimensional data.

[0058] The preprocessor (120) can reduce the dimensionality of the data while preserving relevant spectral information using techniques such as principal component analysis (PCA) or minimum noise ratio transformation (MNF).

[0059] Data normalization refers to scaling the data to train a model. Note that data standardization may also be performed simultaneously. Data normalization is also a method of adjusting the scale of data.

[0060] Normalization and standardization ensure quality, compatibility with subsequent processing steps, and fast optimization speed.

[0061] The extractor (130) extracts a spectral signature, which is a plurality of unit bands, from each preprocessed image (i.e., spectral data).

[0062] A spectral signature refers to the variation in a material's reflectance or emissivity with respect to wavelength. Spectral signatures are used to simplify the complex spectral representation of natural landscapes and the wavelength-dependent characteristics of an object or group of objects on the Earth's surface. A spectrum represents the amount of light of different wavelengths, indicating the amount of light reflected, absorbed, or transmitted by a substance. This indicates the amount of a specific color contained in that light. Because each substance and compound reacts differently to light, its spectral characteristics vary. These spectral signatures can be used to identify substances or compounds.

[0063] The analyzer (140) detects and identifies a target using the extracted spectral signature. Note that the target of detection and identification may be at least one of the Earth's natural resources and environmental dynamics.

[0064] For example, the analyzer (140) can detect and identify a target using either spectrometry or spectral unmixing.

[0065] Spectroscopy is a method of measuring the spectral intensity of a specific wavelength (or physical quantity) within a spectrum. It involves splitting light into monochromatic light using a spectrometer and measuring its intensity to investigate the distribution of light energy at each wavelength. This method can be used to obtain various physicochemical experimental data.

[0066] Spectral mixing analysis is a method of analyzing the proportion (fraction) of the substances that make up each pixel by utilizing the unique spectral reflectance characteristics of the various substances contained in each pixel. In a hyperspectral cube, a spatial pixel can be viewed as a quantitative mixture of reflected energy from one or more substances, which is called spectral mixing. Spectral mixing occurs when a single pixel has insufficient spatial resolution to capture the characteristics of an object or when the spectral resolution of a spectral sensor is low, causing the spectra from different substances to appear as the sum of their individual spectra. The method of analyzing such mixed pixels is called spectral unmixing. In this case, a single pure substance that makes up each pixel is called an endmember, and each pixel can be expressed as the sum of the composition proportions (fractions) of multiple endmembers (fractional abundance). Spectral mixing analysis is divided into linear spectral mixing analysis and nonlinear spectral mixing analysis depending on whether each spectral component affecting a pixel can be assumed independently.

[0067] Additionally, the analyzer (140) may use any one of a supervised classification technique (SAM, Spectral Angle Mapper), an adaptive cosine estimator (ACE, Adaptive Cosine Estimator), and a support vector machine technique (SVM, Support Vector Machines) to detect and identify the target.

[0068] Supervised classification technique (Spectral Angle Mapper, SAM) is a method in which the analyst designates classification items and training data for each classification item, and classifies the image data based on the characteristics of the training data. In supervised classification, the analyst selects the classification items and training data for each classification item in advance, and the image data is classified by determining which training data each pixel is most similar to based on the characteristics of the selected training data. Therefore, for classification, the image data must be analyzed to first set the classification items, and then the training data for each set classification item must be selected. Depending on the criteria for determining which pixel is most similar to which training data, the method is categorized into methods such as Minimum Distance, Parallelogram Method, Maximum Likelihood Method, Spectral Angle Mapper, and Neural Network Method.

[0069] The Adaptive Cosine Estimator (ACE) is an algorithm used to detect objects in hyperspectral images. To predict the performance of this algorithm in hyperspectral scenes, the adaptive cosine estimator utilizes a statistical model that combines a multivariate t-distribution for the background and a Gaussian distribution for the target.

[0070] Support vector machines (SVMs) are models that define a baseline or decision boundary for classification. They create a non-probabilistic binary linear classification model by determining which class new unclassified data belongs to within the boundary. The resulting classification model is represented as a boundary in the data-mapped space, and the SVM algorithm finds the boundary with the greatest width.

[0071] The communicator (150) transmits the processing result of the analyzer (140) to the ground station.

[0072] The communication device (150) can transmit data efficiently and reliably by using an optimized frequency band, modulation method, and data transmission protocol.

[0073] The ground station can process the processing results of the analyzer (140) transmitted from the artificial satellite (100) to perform additional analysis and visualization, and can also use this to create 2D or 3D images.

[0074] For example, a ground station may receive processing results from multiple artificial satellites (100) positioned adjacent to each other, and may collate the data to create a single 2D or 3D image.

[0075] Meanwhile, the image enhancer (160) fuses a space context into the spectral signature extracted from the extractor (130), corrects atmospheric disturbances in the fused spectral signature, and removes noise and artifacts to improve image quality.

[0076] Spatial context information refers to information from past and future points in time that is used to calculate the quality of the block at the current point in time.

[0077] The image enhancer (160) can utilize a machine learning algorithm, and the machine learning algorithm is used to improve the quality and visual clarity of the image.

[0078] The tracer (170) determines the position within the image by predicting the change in position over time of the object identified by the analyzer (140). For example, the tracker (170) determines the exact position of the object within an ultra-hyperspectral image and predicts its movement or change over time.

[0079] The tracker (170) can use centroid tracking and correlation tracking methods for accurate target tracking.

[0080] The center tracking method is a method of tracking a target by finding the center point of the target image from the target image signal value distinguished from the background signal, and, when the target moves, making the center point of the moved target come to the center of the tracking window.

[0081] The correlation tracking method memorizes the target image from the previous screen in the tracking window, finds the same image corresponding to the target image from the previous screen in the current screen (i.e., the image with the highest correlation with the target image), and tracks it by moving the tracking window by the amount of variation.

[0082] The correlation tracking method can utilize a correlation filter, and the correlation filter can include a KCF (Kernelized Correlation Filter) or a MOSSE (Minimum Output Sum of Squared Error) filter.

[0083] Although the present invention has been described above with reference to several embodiments, it will be understood by those skilled in the art that various modifications and changes can be made to the present invention without departing from the spirit and scope of the present invention as set forth in the claims below.

[0084] In addition, the invention regarding the method among the embodiments described above may be implemented as a program or as a computer-readable recording medium having the program stored thereon.

[0085] That is, the present invention can be implemented in the form of an application, and can be implemented as a software program that runs on a mobile terminal such as a smartphone or tablet PC running on Google's Android or Apple's IOS, or can be implemented as a software program that runs on a wearable device such as Google Glass, Apple Watch, Samsung Galaxy Watch, smart watch, or the like, or can be implemented as a software program that runs on a laptop PC or desktop PC running on Microsoft's Windows or Google's Chrome OS.

[0086] In addition, partial functions of the above-described device or system may be provided by being included in a computer-readable recording medium by tangibly implementing a program of commands for implementing them. The computer-readable recording medium may include program commands, data files, data structures, etc., alone or in combination. Examples of the computer-readable recording medium include magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs and DVDs, magneto-optical media such as floptical disks, and hardware devices specially configured to store and execute program commands such as ROMs, RAMs, flash memories, and USB memories.

[0087] <Example 2>

[0088] Example 2 relates to a technology in which spectral data for each direct point of each of the unit satellites (101, 102, 103) is independently adjusted and optimized in real time and then transmitted to a host satellite (100A), and the host satellite (100A) collects data transmitted from the unit satellites (101, 102, 103) and transmits the data to a ground station.

[0089] FIG. 3 is a schematic diagram illustrating an artificial satellite system performing real-time spectral analysis according to Embodiment 2 of the present invention.

[0090] The unit artificial satellites (101, 102, 103) and the host artificial satellite (100A) of Example 2 have the same configuration as the artificial satellite (100 of FIG. 2) of Example 1, that is, they include a spectroscopic camera, a preprocessor, an extractor, an analyzer, a communicator, an image enhancer, and a position tracker, and a duplicate description is omitted.

[0091] As shown in Fig. 3, each unit artificial satellite (101, 102, 103) moves around the Earth from its respective deployed location and photographs its own direct point to obtain spectral data.

[0092] The unit satellites (101, 102, 103) independently preprocess and analyze the acquired spectral data and transmit it to the host satellite (100A).

[0093] The host satellite (100A) collects data transmitted from the unit satellites (101, 102, 103). For example, the host satellite (100A) can collect data in the same order as the order in which the unit satellites (101, 102, 103) are arranged adjacently.

[0094] The host satellite (100A) transmits the collected data to the ground station.

[0095] The ground station can process the processing results of the analyzer transmitted from the host satellite (100A) to perform additional analysis and visualization, and can also use this to create 2D or 3D images.

[0096] For reference, if data collected from the host satellite (100A) is used to create visualizations and 2D or 3D images, this process may be omitted at the ground station.

[0097] <Example 3>

[0098] Example 3 relates to a method for reducing dependence on image processing at a ground station by having the satellite itself adjust and optimize spectral data acquired from a spectral camera mounted on the satellite in real time and then transmit the data to the ground station.

[0099] FIG. 4 is a flowchart illustrating an artificial satellite system performing real-time spectral analysis according to Embodiment 3 of the present invention.

[0100] As shown in Fig. 4, the process may include a step of photographing the Earth (S110), a step of performing quality correction and data normalization (S120), a step of extracting a spectral signature (S130), a step of detecting and identifying an object (S150), and a step of transmitting the image to a ground station (S170). It may also further include a step of improving image quality (S140) and a step of determining a location within the image (S160).

[0101] Step 110 of photographing the Earth involves capturing spectral data by using a spectroscopic camera to photograph the Earth using a multispectral sensor, a hyperspectral sensor, and a hyperspectral sensor. Note that the spectroscopic camera can also photograph the Earth using only the hyperspectral sensor and the hyperspectral sensor.

[0102] The step of performing quality correction and data normalization (S120) is such that the preprocessor performs quality correction and data normalization on images (i.e., spectral data) captured by each of the multispectral sensor, hyperspectral sensor, and hyperspectral sensor.

[0103] Quality correction by the preprocessor may include at least one of radiation correction, atmospheric correction, and noise removal.

[0104] Radiometric correction refers to the process of correcting the radiometric distortion that appears as changes in the brightness value of an image due to atmospheric and topographical influences, including the sun's elevation angle, the response characteristics of the probe, atmospheric conditions, and topographical factors, when acquiring images from an artificial satellite.

[0105] Atmospheric correction refers to the process of correcting distortions in images that occur when light waves or electromagnetic waves are absorbed or scattered by atmospheric particles when observing the Earth's surface with an artificial satellite.

[0106] In addition, quality correction by the preprocessor may include geometric correction and orthogonal correction.

[0107] Geometric correction may involve projecting each location in a spectral image from a satellite onto a flat surface with appropriate x and y axes. Satellite data can contain geometric errors due to factors such as the Earth's rotation, its curvature, and the attitude and orientation of the satellite's spectral camera. Therefore, geometric correction may be performed to ensure accurate satellite imagery.

[0108] Orthorectification may involve the process of precisely linking satellite image data with ground coordinates to produce quantitative information.

[0109] Image denoising improves image quality through noise removal filters or algorithms. For example, noise can be removed using techniques such as Gaussian filtering, median filtering, bilateral filtering, averaging filtering, median filtering, and adaptive filtering based on local image variance.

[0110] Image noise removal can also be achieved using the Minimum Noise Fraction transform (MNF). This algorithm reduces the number of bands in hyperspectral data and minimizes noise within the image.

[0111] The minimum noise ratio transform (LMTR) is a technique for extracting signals with spatial autocorrelation within an image. This allows for the separation of the effects of artificial image signals, such as noise, from the original image. For example, noise-induced image signals typically have unique spatial patterns clearly distinguishable from those of natural fluctuations and exhibit high spatial autocorrelation, making them easily extracted using the minimum noise ratio transform (MNTR).

[0112] The minimum noise ratio transform (MNTR) method applies a two-step principal component analysis (PCA). The first transform reduces and rescales the data's noise to produce uncorrelated data. The second transform generates a noise-reduced MNF eigenimage and a noise-included MNF eigenimage.

[0113] Principal component analysis (PCA), applied to the minimum noise ratio transformation method, is one of the methods for efficiently summarizing multidimensional data with correlations between variables into low-dimensional data.

[0114] Preprocessors can reduce the dimensionality of the data while preserving relevant spectral information using techniques such as principal component analysis (PCA) or minimum noise ratio transform (MNF).

[0115] Data normalization refers to scaling the data to train a model. Note that data standardization may also be performed simultaneously. Data normalization is also a method of adjusting the scale of data.

[0116] Normalization and standardization ensure quality, compatibility with subsequent processing steps, and fast optimization speed.

[0117] In the step of extracting a spectral signature (S130), the extractor extracts a spectral signature, which is a plurality of unit bands, from each preprocessed image (i.e., spectral data).

[0118] A spectral signature refers to the variation in a material's reflectance or emissivity with respect to wavelength. Spectral signatures are used to simplify the complex spectral representation of natural landscapes and the wavelength-dependent characteristics of an object or group of objects on the Earth's surface. A spectrum represents the amount of light of different wavelengths, indicating the amount of light reflected, absorbed, or transmitted by a substance. This indicates the amount of a specific color contained in that light. Because each substance and compound reacts differently to light, its spectral characteristics vary. These spectral signatures can be used to identify substances or compounds.

[0119] The step (S140) of improving image quality is such that the image enhancer fuses a space context into the spectral signature extracted from the extractor, corrects atmospheric disturbances in the fused spectral signature, and removes noise and artifacts to improve image quality.

[0120] Spatial context information refers to information from past and future points in time that is used to calculate the quality of the block at the current point in time.

[0121] Image enhancers can utilize machine learning algorithms, which are used to improve the quality and visual clarity of images.

[0122] The step of detecting and identifying the target (S150) involves detecting and identifying the target using the extracted spectral signature of the analyzer. The target of detection and identification may be at least one of the Earth's natural resources and environmental dynamics.

[0123] Spectroscopy is a method of measuring the spectral intensity of a specific wavelength (or physical quantity) within a spectrum. It involves splitting light into monochromatic light using a spectrometer and measuring its intensity to investigate the distribution of light energy at each wavelength. This method can be used to obtain various physicochemical experimental data.

[0124] Spectral mixing analysis is a method of analyzing the proportion (fraction) of the substances that make up each pixel by utilizing the unique spectral reflectance characteristics of the various substances contained in each pixel. In a hyperspectral cube, a spatial pixel can be viewed as a quantitative mixture of reflected energy from one or more substances, which is called spectral mixing. Spectral mixing occurs when a single pixel has insufficient spatial resolution to capture the characteristics of an object or when the spectral resolution of a spectral sensor is low, causing the spectra from different substances to appear as the sum of their individual spectra. The method of analyzing such mixed pixels is called spectral unmixing. In this case, a single pure substance that makes up each pixel is called an endmember, and each pixel can be expressed as the sum of the composition proportions (fractions) of multiple endmembers (fractional abundance). Spectral mixing analysis is divided into linear spectral mixing analysis and nonlinear spectral mixing analysis depending on whether each spectral component affecting a pixel can be assumed independently.

[0125] Additionally, the analyzer may utilize any one of a supervised classification technique (Spectral Angle Mapper, SAM), an Adaptive Cosine Estimator (ACE), and a Support Vector Machine technique (SVM, SVM) for detection and identification of the target object.

[0126] Supervised classification technique (Spectral Angle Mapper, SAM) is a method in which the analyst designates classification items and training data for each classification item, and classifies the image data based on the characteristics of the training data. In supervised classification, the analyst selects the classification items and training data for each classification item in advance, and the image data is classified by determining which training data each pixel is most similar to based on the characteristics of the selected training data. Therefore, for classification, the image data must be analyzed to first set the classification items, and then the training data for each set classification item must be selected. Depending on the criteria for determining which pixel is most similar to which training data, the method is categorized into minimum distance, parallelogram, maximum likelihood, spectral angle mapper, and neural network methods.

[0127] The Adaptive Cosine Estimator (ACE) is an algorithm used to detect objects in hyperspectral images. To predict the performance of this algorithm in hyperspectral scenes, the adaptive cosine estimator utilizes a statistical model that combines a multivariate t-distribution for the background and a Gaussian distribution for the target.

[0128] Support vector machines (SVMs) are models that define a baseline or decision boundary for classification. They create a non-probabilistic binary linear classification model by determining which class new unclassified data belongs to within the boundary. The resulting classification model is represented as a boundary in the data-mapped space, and the SVM algorithm finds the boundary with the greatest width.

[0129] Step (S160) of determining a position within an image involves determining a position within an image by predicting changes in the position of an object identified by the analyzer over time using a tracker. For example, the tracker determines the precise position of an object within an ultra-hyperspectral image and predicts its movement or changes over time.

[0130] The tracker can use centroid tracking and correlation tracking methods for accurate target tracking.

[0131] The center tracking method is a method of tracking a target by finding the center point of the target image from the target image signal value distinguished from the background signal, and, when the target moves, making the center point of the moved target come to the center of the tracking window.

[0132] The correlation tracking method memorizes the target image from the previous screen in the tracking window, finds the same image corresponding to the target image from the previous screen in the current screen (i.e., the image with the highest correlation with the target image), and tracks it by moving the tracking window by the amount of variation.

[0133] The correlation tracking method can utilize a correlation filter, and the correlation filter can include a KCF (Kernelized Correlation Filter) or a MOSSE (Minimum Output Sum of Squared Error) filter.

[0134] In the step of transmitting to the ground station (S170), the communication device transmits the processing result of the analyzer to the ground station.

[0135] The communication device can transmit data efficiently and reliably by using optimized frequency bands, modulation methods, and data transmission protocols.

[0136] The ground station can process the analysis results received from the satellite for further analysis and visualization, and can also use them to create 2D or 3D images.

[0137] For example, a ground station can receive processing results from multiple satellites positioned adjacent to each other and combine the data to create a single 2D or 3D image.

Claims

1. A spectral camera that photographs the Earth using multispectral sensors, hyperspectral sensors, and hyperspectral sensors; A preprocessor that performs quality correction and data normalization on images taken by each sensor; An extractor that extracts spectral signatures, which are multiple unit bands, from each of the above preprocessed images; An analyzer that detects and identifies a target using the extracted spectral signature; and A communicator that transmits the processing results of the above analyzer to the ground station An artificial satellite system that performs real-time spectral analysis including .

2. In paragraph 1, The quality correction performed by the above preprocessor includes at least one of radiation correction, atmospheric correction, and noise removal. A satellite system that performs real-time spectral analysis.

3. In paragraph 1, It further includes an image enhancer that improves image quality by fusing spatial context to the above-extracted spectral signature, correcting atmospheric disturbances, and removing noise and artifacts from the fused spectral signature. The above analyzer detects and identifies the target object using the corrected spectral signature. A satellite system that performs real-time spectral analysis.

4. In paragraph 3, The above analyzer uses any one of Spectral Angle Mapper (SAM), Adaptive Cosine Estimator (ACE), and Support Vector Machines (SVM) for detection and identification of the target. A satellite system that performs real-time spectral analysis.

5. In paragraph 3, Further comprising a position tracker for determining a position in an image by predicting a change in the position of the identified object over time. A satellite system that performs real-time spectral analysis.

6. In paragraph 1, The above ground station is characterized in that it processes the processing result of the analyzer to generate a 2D or 3D image. A satellite system that performs real-time spectral analysis.

7. In any one of paragraphs 1 to 6, The object of the above detection and identification is at least one of the Earth's natural resources and environmental dynamics. A satellite system that performs real-time spectral analysis.

8. In a real-time spectral analysis method for a small artificial satellite system, A step in which a spectral camera takes pictures of the Earth using a multispectral sensor, a hyperspectral sensor, and a hyperspectral sensor; A step in which a preprocessor performs quality correction and data normalization on the images captured by each sensor; A step in which an extractor extracts spectral signatures, which are multiple unit bands, from each of the preprocessed images; A step in which an analyzer detects and identifies a target object using the extracted spectral signature; and A step in which a communicator transmits the processing results of the above analyzer to a ground station. A real-time spectroscopic analysis method for a small artificial satellite system including a .

9. In paragraph 8, The quality correction performed by the above preprocessor includes at least one of radiation correction, atmospheric correction, and noise removal. A real-time spectral analysis method for small satellite systems.

10. In paragraph 8, The image enhancer further includes a step of improving image quality by fusing spatial context to the extracted spectral signature, correcting atmospheric disturbances, and removing noise and artifacts from the fused spectral signature. The above analyzer detects and identifies the target object using the corrected spectral signature. A real-time spectral analysis method for small satellite systems.

11. In paragraph 10, The above analyzer uses any one of Spectral Angle Mapper (SAM), Adaptive Cosine Estimator (ACE), and Support Vector Machines (SVM) for detection and identification of the target. A real-time spectral analysis method for small satellite systems.

12. In paragraph 10, The position tracker further comprises a step of determining the position in the image by predicting the change in position over time of the identified object. A real-time spectral analysis method for small satellite systems.

13. In paragraph 8, The above ground station further includes a step of processing the processing result of the above analyzer to generate a 2D or 3D image. A real-time spectral analysis method for small satellite systems.

14. In any one of paragraphs 8 to 13, The object of the above detection and identification is at least one of the Earth's natural resources and environmental dynamics. A real-time spectral analysis method for small satellite systems.

Images