Automatic cropping and correction airborne hyperspectral data processing method and system

Abstract

This invention discloses an automatic cropping and correction method and system for airborne hyperspectral data processing. The automatic flight path cropping technology uses algorithmic analysis of GPS trajectory data, combined with information such as latitude, longitude, heading, and altitude, to extract and crop effective flight paths. On the one hand, it can remove redundant spectral data, significantly reduce the amount of subsequent calculations, and significantly improve the data processing speed, solving the problem of low efficiency in traditional manual cropping. On the other hand, by accurately extracting effective flight paths, it avoids the omission of effective data or redundant residues caused by manual operation, providing high-quality basic data for subsequent correction processing and ensuring the accuracy of overall data processing.

Description

Technical Field

[0001] This invention relates to the field of spectral processing, and more particularly to an airborne hyperspectral data processing method and system with automatic cropping and correction. Background Technology

[0002] Airborne hyperspectral data, with its high resolution and multi-band characteristics, plays an irreplaceable role in water quality monitoring, ecological assessment, and mineral exploration, providing key technical support for accurately acquiring information on the Earth's surface and water bodies. However, problems in the data cropping and geometric correction stages of current airborne hyperspectral data processing workflows severely restrict the full realization of its application effectiveness.

[0003] In flight path data processing, traditional methods rely on manual trimming of airborne hyperspectral data. This is not only inefficient but also susceptible to variations in operator experience, easily leading to incomplete extraction of effective flight path data or excessive redundant data. Redundant data directly increases the computational load of subsequent geometric correction, slowing down the overall processing speed and even causing deviations in the correction benchmark due to interference from invalid information. Conversely, the omission of effective data affects the completeness of the final analysis results, making it difficult to meet the stringent data accuracy requirements of scenarios such as water quality monitoring.

[0004] Meanwhile, in the geometric correction stage of airborne hyperspectral data, line-scan hyperspectral equipment relies excessively on high-precision IMU modules. While high-precision IMUs can guarantee correction accuracy to a certain extent, their high hardware costs significantly increase the deployment and maintenance costs of airborne systems, making them unaffordable for small-to-medium-scale applications or cost-sensitive scenarios, thus limiting the widespread adoption of airborne hyperspectral technology. These problems, combined, affect the accuracy and efficiency of airborne hyperspectral data applications in fields such as water quality monitoring, and the cost barrier hinders the widespread implementation of the technology, becoming a core bottleneck that urgently needs to be overcome in current airborne hyperspectral data processing.

[0005] In summary, there is a need for an automatic cropping and correction method and system for airborne hyperspectral data processing to address the shortcomings of existing technologies. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides an automatic cropping and correction method and system for airborne hyperspectral data processing. It aims to solve the problems of low processing efficiency, high cost, and difficulty in guaranteeing data accuracy in airborne hyperspectral data processing, which are caused by manual cropping of flight path data, incomplete extraction of effective data, or excessive redundant data. Furthermore, the correction process relies on a high-precision IMU module, resulting in high hardware costs. These factors combined affect the application effectiveness.

[0007] To achieve the above objectives, the present invention provides the following technical solution: an airborne hyperspectral data processing method for automatic cropping and correction, comprising the following steps:

[0008] Step S1: Airborne multi-source data acquisition. Data is collected according to the planned flight path using a hyperspectral camera, a visible light camera, GPS, and zenith light equipment. The navigation angle is calculated based on the changes in latitude and longitude at adjacent times. The zenith light equipment monitors changes in ambient light in real time. The exposure time of the hyperspectral camera is adjusted in real time according to the collected data. The reflectance is calculated in real time using zenith light data and hyperspectral data.

[0009] Step S2: Automatic flight strip trimming. The core flight path is identified by the trajectory clustering algorithm, and the flight path clusters that meet the conditions are selected as the target flight path segments to generate a complete working flight strip and complete the trimming.

[0010] Step S3: Geographic reference base construction. The visible light image is preprocessed to synthesize an orthophoto map. GPS information is used to assign geographic coordinates to each pixel of the base map to form a geographic reference base map.

[0011] Step S4: Orthorectified hyperspectral data generation. The effective flight strip data is divided into small blocks, and preliminary matching is performed by combining GPS and orthorectified base map. Feature points are extracted and matched to obtain the transformation matrix. The transformation matrix is ​​applied to each wavelength dimension to generate an orthorectified hyperspectral image with UTM coordinates.

[0012] Optionally, in step S1, data is collected according to the planned route in the following way:

[0013] The hyperspectral camera, zenith light, and GPS collect data at the same frequency. The GPS obtains the current timestamp and latitude / longitude information in real time. Each frame of the hyperspectral camera and zenith light corresponds to a set of GPS data. Each visible light image obtains the corresponding GPS data through the relationship between the acquisition frequencies and the timestamp.

[0014] Optionally, the heading angle in step S1 is derived in the following way:

[0015] Based on the temporal changes in latitude and longitude at adjacent moments, the heading angle of each frame of data is derived using the spherical azimuth formula, and the result is converted into an angle value from 0° to 360°. The inherent parameters of the equipment are then exported as data for automatic flight strip cutting and geometric correction.

[0016] Optionally, the core route is identified in step S2 in the following way:

[0017] Set the spatial neighborhood radius and minimum number of points for the adaptive operation route spacing. By calculating the spatial density distribution of trajectory points, spatially continuous clustered trajectory points with a number of points greater than or equal to the minimum number of points in the neighborhood are divided into the same track cluster. Three track clusters containing the longest continuous trajectory points are selected. The average heading angle of each cluster is calculated. Starting from the longest route, if there is a track cluster whose heading is ±180° opposite to any of the other two clusters, then the cluster is selected as the reference route.

[0018] Optionally, in step S2, the selection of eligible track clusters as target route segments can be achieved through the following methods:

[0019] Using a defined baseline route as a reference, traverse all remaining track clusters, filter out track clusters whose length exceeds a set threshold and whose heading angle deviates, mark them as target route segments, and exclude track clusters with excessive deviation or insufficient length.

[0020] Optionally, step S2 further includes fine-tuning of the flight strips, through the following methods:

[0021] By combining the average elevation and latitude and longitude changes of each track cluster to calculate the velocity parameters, discrete points are eliminated. A sliding window is set by linear fitting to check the slope stability of the trajectory within the window and the distance between adjacent points. Points with abnormal slopes or distances exceeding a certain range of average distance are removed from the target labels, and continuous trajectory segments with stable motion are retained.

[0022] Optionally, in step S2, a complete work strip is generated and trimmed, in the following way:

[0023] Integrate all target tag line segments that have passed the fine-tuning screening to generate a complete and effective operational flight strip. Based on business needs, cut out all track segments within the specified tag range from the complete flight strip, filter out the effective strip cutting range, and complete the automatic flight strip cutting.

[0024] Optionally, the geographic reference base map in step S3 is formed in the following way:

[0025] The image distortion caused by lens optical distortion is eliminated based on the camera's own parameters. Then, the brightness and contrast of the images are adjusted under different lighting conditions. Cross-image matching is performed by comparing the similarity of feature points between different images. The geometric transformation relationship between images is calculated based on the matching results. After aligning multiple images, the pixel gray values ​​of the overlapping areas are smoothed to form an orthophoto map.

[0026] By using the GPS positioning information associated with each image when it was captured, each pixel of the orthophoto map is assigned a corresponding geographic coordinate, forming a geographic reference map with a unified spatial benchmark.

[0027] Optionally, the orthophoto hyperspectral data generation in step S4 is performed in the following manner:

[0028] Step A1: Hyperspectral data block processing. The effective flight band hyperspectral data is divided into several data blocks according to a preset size. The GPS positioning information corresponding to the middle frame of each data block is extracted and converted into UTM coordinates. The image corresponding to each small block on the orthophoto map is cropped by combining the geographic pixel size of the hyperspectral image and the visible light orthophoto map. The UTM coordinates corresponding to the four corners of the cropped hyperspectral block can be calculated by using the UTM coordinates of the center block and the geographic pixel size. The pixel position is found by using the UTM coordinates, and then a slightly larger area is cropped according to the geographic range of the data block.

[0029] Step A2: Image matching and transformation matrix acquisition: Select three channels from the segmented hyperspectral data to synthesize a visible light image or select a channel with a high signal-to-noise ratio as the matching image. Extract feature points from the visible light orthophoto image blocks corresponding to the geographic coordinates of the obtained hyperspectral image blocks and perform feature point matching. Obtain the transformation matrix based on the matching relationship and save it.

[0030] Step A3: Orthorectified hyperspectral image generation. Traverse all hyperspectral data blocks, apply the corresponding spatial transformation matrix to the two-dimensional data array of each wavelength dimension, and generate a complete orthorectified hyperspectral image with UTM geographic coordinate attributes, thus completing the accurate association between spectral data and geographic space.

[0031] An airborne hyperspectral data processing system with automatic cropping and correction, employing the aforementioned airborne hyperspectral data processing method with automatic cropping and correction, is characterized by comprising a data acquisition module, an automatic flight strip cropping module, a geographic reference benchmark construction module, and an orthophoto hyperspectral data generation module.

[0032] The data acquisition module is responsible for collecting data using a multi-source sensor system, including a hyperspectral camera, a visible light camera, GPS, and zenith light equipment, according to a predetermined flight path.

[0033] The automatic flight strip trimming module is used to automatically identify core routes based on trajectory clustering algorithms, select eligible flight track clusters as target route segments, and complete the fine-grained selection and final trimming of flight strips.

[0034] The geographic reference benchmark construction module is used to preprocess visible light images to eliminate optical distortion, adjust brightness and contrast, synthesize an orthophoto base map, and use GPS information to assign geographic coordinates to each pixel of the base map to form a unified spatial benchmark.

[0035] The orthorectified hyperspectral data generation module is used to segment the obtained effective flight strip data into small blocks and perform preliminary matching with GPS and orthorectified base maps. Then, transformation matrices are obtained through feature point extraction and matching. Finally, these matrices are applied to the data in each wavelength dimension to generate orthorectified hyperspectral images with UTM coordinates.

[0036] The beneficial effects of this invention are:

[0037] 1. In this invention, the automatic flight track clipping technology uses algorithmic analysis of GPS trajectory data, combined with information such as latitude, longitude, heading, and altitude, to extract and clip effective flight strips. On the one hand, it can remove redundant spectral data, significantly reduce the amount of subsequent calculations, and significantly improve the data processing speed, solving the problem of low efficiency in traditional manual clipping. On the other hand, by accurately extracting effective flight strips, it avoids the omission of effective data or redundant residues caused by manual operation, providing high-quality basic data for subsequent correction processing and ensuring the accuracy of overall data processing.

[0038] 2. In this invention, a spectral data correction technology based on GPS and orthophotos is adopted. It does not rely on a high-precision IMU module. It acquires photos through a visible light camera and processes them into orthophotos. It combines GPS data for projection transformation and registration, and then achieves geometric calibration through a transformation matrix. Compared with the traditional correction method that relies on IMU modules, it not only significantly reduces hardware costs, making the hyperspectral data processing solution more economical and widely applicable, but also reduces redundant information interference during the correction process by combining the accurate data automatically cropped in the early stage, further improving the geometric accuracy and reliability of the spectral data.

[0039] 3. In this invention, the synergistic effect of automatic track cropping and GPS-orthophoto correction technology forms an efficient and complete processing flow. The high-quality data after automatic cropping provides more accurate input for the correction process, reducing the sources of error in the correction process. The low-cost advantage of the correction technology enables this processing solution to be applied in more scenarios. The combination of the two not only improves the efficiency and accuracy of data processing, but also lowers the application threshold, which helps airborne hyperspectral data play a greater role in water quality monitoring, ecological assessment and other fields. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of a method flow of the present invention.

[0041] Figure 2 This is an automatic exposure adjustment diagram according to the present invention.

[0042] Figure 3 This is a schematic diagram of a system structure according to the present invention. Detailed Implementation

[0043] To more clearly illustrate the technical solutions in the embodiments of the invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0044] like Figure 1 and Figure 2 As shown, an airborne hyperspectral data processing method with automatic cropping and correction includes the following specific components:

[0045] Step 1: Airborne multi-source data acquisition.

[0046] The drone is equipped with a hyperspectral camera, a visible light camera, GPS, and zenith light. It collects data along a planned flight path, with the hyperspectral camera, zenith light, and GPS collecting data at the same frequency. The GPS obtains the current timestamp and latitude / longitude information in real time. Each frame from the hyperspectral camera and zenith light corresponds to a set of GPS data. The visible light camera collects data at a slower frequency, but each visible light camera image has corresponding GPS data, which can be correlated with other data by the relationship between the collection frequencies and the timestamp.

[0047] Input base map data, collect data according to the planned route using hyperspectral camera, visible light camera, GPS and zenith light equipment, identify water body data, compare with the water body data on the base map, and correct the hyperspectral data.

[0048] Then, based on the temporal changes in latitude and longitude at adjacent moments, the heading angle of each frame of data is derived using the spherical azimuth formula, and the result is converted into an angle value of 0°-360° (0° represents true north, increasing clockwise). At the same time, the inherent parameters of the equipment, such as the field of view of the hyperspectral sensor and the focal length of the visible light camera, are also exported to provide basic support for subsequent automatic flight strip trimming and geometric correction.

[0049] At the same time, such as Figure 2As shown, the zenith light sensor monitors ambient light changes in real time, and the control program automatically adjusts the exposure based on the collected data to ensure optimal signal-to-noise ratio (SNR) for the zenith light sensor data acquisition. The hyperspectral sensor also performs real-time automatic exposure through the control program, thereby obtaining spectral data with the best SNR. Algorithmic analysis of the acquired zenith and hyperspectral data yields real-time reflectance hyperspectral data. This processing removes ambient light intensity variations caused by cloud cover or moisture, resulting in more accurate reflectance data. The pixel digital quantization value (DN) is linearly related to the exposure time. By analyzing the zenith and hyperspectral data using algorithms, reflectance is calculated in real time, and ambient light intensity variations caused by clouds or moisture are removed, accurate reflectance data is obtained. This enables adaptive exposure adjustment, achieving real-time reflectance acquisition and real-time ambient light correction.

[0050] To ensure data alignment, a synchronous trigger controller is also introduced into the system, enabling the zenith spectral sensor and the hyperspectral sensor to collect data at the same time, thus ensuring that the data collected by the two devices are at the same moment when the ambient light changes.

[0051] The algorithm is calculated as follows:

[0052] First, calculate the ratio data1 between the hyperspectral data and the current hyperspectral exposure time, and collect zenith light data to make its wavelength consistent with the hyperspectral data;

[0053] Secondly, the ratio of zenith light data to zenith light exposure time, data2, is calculated to obtain data fluctuations that are eliminated by changes in ambient light.

[0054] Finally, the ratio of data1 to data2, data3, is calculated. The spectral data is then converted into reflectance data using the calibration model k+b. The reflectance data is equal to the ratio data3 multiplied by the slope k of the calibration model, plus the intercept b.

[0055] The calibration model can be derived through data collection and analysis in the laboratory. For example, we can derive the calibration model by collecting data from a known reflectivity plate.

[0056] Meanwhile, the corrected hyperspectral data and real-time calculated reflectance are used as a basis for the selection and trimming of flight strips.

[0057] Step 2: Automatic tape cutting.

[0058] (1) Trajectory Clustering and Core Track Identification: The DBSCAN equal-density clustering algorithm is used. The spatial neighborhood radius for the appropriate operation route spacing is set, such as 50m; and the minimum number of points is set, such as 10. By calculating the spatial density distribution of trajectory points, spatially continuous clustered trajectory points with a number of points greater than or equal to the minimum number of points in the neighborhood are divided into the same track cluster, and isolated points with a density lower than the threshold are excluded. On this basis, three track clusters containing the longest continuous trajectory points are selected, and the average heading angle of each cluster is calculated. Starting from the longest route, if there is a track cluster whose heading is ±180° opposite to any of the other two clusters, then the cluster is selected as the reference route to provide a reference standard for subsequent track matching, because there may be cases where the longest continuous route is from the take-off point to the collection point, especially in the case of collecting seawater.

[0059] (2) Track matching and marking: Using the established baseline route as a reference, all remaining track clusters are traversed, and track clusters with a length exceeding the baseline route by 30% and a heading angle deviation of ≤10° are selected. The deviation range includes deviations in the opposite direction of ±180° from the baseline heading. These track clusters that meet the conditions are uniformly marked as the target route segment. At the same time, track clusters with excessive deviation or insufficient length are excluded to ensure that the retained track clusters match the baseline route in terms of spatial scale and heading consistency, laying the foundation for subsequent refined screening.

[0060] (3) Fine-tuning of flight strips: First, combine the average elevation and latitude and longitude changes of each flight strip cluster to calculate the speed parameters and remove discrete points that have not entered the formal data collection stage to reduce the interference of non-operational trajectories on the results. Then, set a sliding window, such as 10 points, through linear fitting to check the slope stability of the trajectory within the window and the distance between adjacent points. Remove points with abnormal slopes or distances exceeding a certain range of the average distance from the target labels. Finally, retain continuous trajectory segments with stable motion to improve the accuracy and reliability of flight strips. Generally, the points that are screened out are unstable points at both ends.

[0061] (4) Flight strip extraction and trimming: Integrate all target tag line segments that have passed the fine-tuning to generate a complete and effective operational flight strip. Based on business needs, trim all track segments within the specified tag range from the complete flight strip to determine the trimming range of the effective strips and complete the entire process of automatic flight strip trimming.

[0062] Step 3: Geographic reference benchmark construction, i.e., orthophoto map generation; orthophoto map is created using visible light images. Figure 1One advantage is that visible light orthorectified images do not require IMU data, and another is that hyperspectral cameras use line scanning, which may result in missed scans. The image sequence acquired by the airborne visible light camera undergoes preprocessing. First, based on camera parameters such as focal length and principal point position, image distortion caused by lens optical distortion is eliminated. Then, the brightness and contrast of the images under different lighting conditions are adjusted to ensure a consistent grayscale distribution across all images. Next, feature points are extracted from each image. These feature points remain stable under scale changes. Cross-image matching is performed by comparing the similarity of feature points between different images, while incorrectly matched point pairs are eliminated to ensure matching accuracy. Based on the matching results, the geometric transformation relationship between images is calculated. After aligning multiple images, the pixel grayscale values ​​in overlapping areas are smoothed to eliminate stitching artifacts, ultimately synthesizing an orthorectified base map covering the work area. Finally, using the GPS positioning information (latitude and longitude) associated with each image capture, each pixel of the orthorectified base map is assigned corresponding geographic coordinates, forming a unified spatial reference base map, which serves as a reference benchmark for the geometric correction of the hyperspectral data.

[0063] Step 4: Generate orthophoto hyperspectral data.

[0064] (1) Hyperspectral data block processing: Block processing is because the flight path strip cannot maintain a constant speed throughout the actual flight, and the low accuracy of GPS leads to local image distortion. If one block of the same flight path is accurately matched, another block at a greater distance may not be matched so accurately. The effective flight path hyperspectral data after cropping in step 2 is divided into several data blocks according to a preset size, such as width × width × spectral dimension, for example, 640 × 640 × 300. The GPS positioning information corresponding to the middle frame of each data block is extracted and converted into UTM coordinates. The image corresponding to each small block on the orthophoto map is cropped by combining the geographic pixel size of the hyperspectral image and the visible light orthophoto map. The geographic pixel size of the hyperspectral image can be obtained by averaging the distance between two frames of GPS data of the flight path. The approximate UTM coordinates corresponding to the four corners of the cropped hyperspectral block can be calculated by using the UTM coordinates of the center block and the geographic pixel size. Since each pixel in the visible light orthophoto image corresponds to a UTM coordinate, the pixel position can be found by using the UTM coordinates. Then, a slightly larger area is cropped according to the geographic range of the data block, such as reserving 10%. Redundancy is matched because GPS accuracy may not be high; this step is a coarse screening that uses geographic information to improve matching speed.

[0065] (2) Image matching and transformation matrix acquisition: Select three channels of the segmented hyperspectral data to synthesize a visible light image or select a channel with a high signal-to-noise ratio as the matching image. Use the same method as the visible light orthophoto image block corresponding to the geographic coordinates of the hyperspectral image block obtained in (1), such as superpoint, to extract feature points and perform feature point matching. Obtain the transformation matrix according to the matching relationship and save it.

[0066] (3) Generation of ortho-hyperspectral image: Traverse all hyperspectral data blocks, apply the corresponding spatial transformation matrix to the two-dimensional data array of each wavelength dimension, and finally generate a complete ortho-hyperspectral image with UTM geographic coordinate attributes, so as to realize the accurate association between spectral data and geographic space.

[0067] An airborne hyperspectral data processing system with automatic cropping and correction, employing the aforementioned airborne hyperspectral data processing method with automatic cropping and correction, is characterized by comprising a data acquisition module, an automatic flight strip cropping module, a geographic reference benchmark construction module, and an orthophoto hyperspectral data generation module.

[0068] The data acquisition module is responsible for collecting data using a multi-source sensor system, including a hyperspectral camera, a visible light camera, GPS, and zenith light equipment, according to a predetermined flight path.

[0069] The automatic flight strip trimming module is used to automatically identify core routes based on trajectory clustering algorithms, select eligible flight track clusters as target route segments, and complete the fine-grained selection and final trimming of flight strips.

[0070] The geographic reference benchmark construction module is used to preprocess visible light images to eliminate optical distortion, adjust brightness and contrast, synthesize an orthophoto base map, and use GPS information to assign geographic coordinates to each pixel of the base map to form a unified spatial benchmark.

[0071] The orthorectified hyperspectral data generation module is used to segment the obtained effective flight strip data into small blocks and perform preliminary matching with GPS and orthorectified base maps. Then, transformation matrices are obtained through feature point extraction and matching. Finally, these matrices are applied to the data in each wavelength dimension to generate orthorectified hyperspectral images with UTM coordinates.

[0072] This invention's automatic flight track clipping technology uses algorithmic analysis of GPS trajectory data, combined with information such as latitude, longitude, heading, and altitude, to extract and clip effective flight strips. On the one hand, it removes redundant spectral data, significantly reducing subsequent computation and greatly improving data processing speed, thus solving the problem of low efficiency in traditional manual clipping. On the other hand, by accurately extracting effective flight strips, it avoids the omission of effective data or redundant residues caused by manual operation, providing high-quality basic data for subsequent correction processing and ensuring the accuracy of overall data processing.

[0073] This method employs a spectral data correction technique based on GPS and orthophotos, eliminating the need for a high-precision IMU module. It acquires images using a visible light camera, processes them into orthophotos, combines them with GPS data for projection transformation and registration, and then achieves geometric calibration through a transformation matrix. Compared to traditional correction methods that rely on IMU modules, this significantly reduces hardware costs, making the hyperspectral data processing solution more economical and widely applicable. Furthermore, by incorporating pre-cropped, precise data, the correction process reduces redundant information interference, further improving the geometric accuracy and reliability of the spectral data.

[0074] The synergy between automatic track cropping and GPS-orthophoto correction technology forms an efficient and complete processing workflow. The high-quality data after automatic cropping provides more accurate input for the correction process, reducing the sources of error in the correction process. Meanwhile, the low-cost advantage of the correction technology enables this processing solution to be applied in more scenarios. The combination of the two not only improves the efficiency and accuracy of data processing, but also lowers the application threshold, which helps airborne hyperspectral data play a greater role in water quality monitoring, ecological assessment and other fields.

[0075] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions or improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An airborne hyperspectral data processing method with automatic cropping and correction, characterized in that, Includes the following steps: Step S1: Airborne multi-source data acquisition. Data is collected according to the planned flight path using a hyperspectral camera, a visible light camera, GPS, and zenith light equipment. The navigation angle is calculated based on the changes in latitude and longitude at adjacent times. The zenith light equipment monitors changes in ambient light in real time. The exposure time of the hyperspectral camera is adjusted in real time according to the collected data. The reflectance is calculated in real time using zenith light data and hyperspectral data. Step S2: Automatic flight strip trimming. The core flight path is identified by the trajectory clustering algorithm, and the flight path clusters that meet the conditions are selected as the target flight path segments to generate a complete working flight strip and complete the trimming. Step S3: Geographic reference base construction. The visible light image is preprocessed to synthesize an orthophoto map. GPS information is used to assign geographic coordinates to each pixel of the base map to form a geographic reference base map. Step S4: Orthorectified hyperspectral data generation. The effective flight strip data is divided into small blocks, and preliminary matching is performed by combining GPS and orthorectified base map. Feature points are extracted and matched to obtain the transformation matrix. The transformation matrix is ​​applied to each wavelength dimension to generate an orthorectified hyperspectral image with UTM coordinates.

2. The airborne hyperspectral data processing method with automatic cropping and correction according to claim 1, characterized in that, In step S1, data is collected according to the planned route, using the following methods: The hyperspectral camera, zenith light, and GPS collect data at the same frequency. The GPS obtains the current timestamp and latitude / longitude information in real time. Each frame of the hyperspectral camera and zenith light corresponds to a set of GPS data. Each visible light image obtains the corresponding GPS data through the relationship between the acquisition frequencies and the timestamp.

3. The airborne hyperspectral data processing method with automatic cropping and correction according to claim 1, characterized in that, The heading angle in step S1 is derived in the following way: Based on the temporal changes in latitude and longitude at adjacent moments, the heading angle of each frame of data is derived using the spherical azimuth formula, and the result is converted into an angle value from 0° to 360°. The inherent parameters of the equipment are then exported as data for automatic flight strip cutting and geometric correction.

4. The airborne hyperspectral data processing method with automatic cropping and correction according to claim 1, characterized in that, In step S2, the core route is identified using the following methods: Set the spatial neighborhood radius and minimum number of points for the adaptive operation route spacing. By calculating the spatial density distribution of trajectory points, spatially continuous clustered trajectory points with a number of points greater than or equal to the minimum number of points in the neighborhood are divided into the same track cluster. Three track clusters containing the longest continuous trajectory points are selected. The average heading angle of each cluster is calculated. Starting from the longest route, if there is a track cluster whose heading is ±180° opposite to any of the other two clusters, then the cluster is selected as the reference route.

5. The airborne hyperspectral data processing method with automatic cropping and correction according to claim 4, characterized in that, In step S2, track clusters that meet the criteria are selected as target route segments in the following manner: Using a defined baseline route as a reference, traverse all remaining track clusters, filter out track clusters whose length exceeds the set threshold and whose heading angle deviation is ≤10°, mark them as target route segments, and exclude track clusters with excessive deviation or insufficient length.

6. The airborne hyperspectral data processing method with automatic cropping and correction according to claim 5, characterized in that, Step S2 also includes fine-tuning of the flight strip, through the following methods: By combining the average elevation and latitude and longitude changes of each track cluster to calculate the velocity parameters, discrete points are eliminated. A sliding window is set by linear fitting to check the slope stability of the trajectory within the window and the distance between adjacent points. Points with abnormal slopes or distances exceeding a certain range of average distance are removed from the target labels, and continuous trajectory segments with stable motion are retained.

7. The airborne hyperspectral data processing method for automatic cropping and correction according to claim 6, characterized in that, In step S2, a complete work tape is generated and trimmed, using the following method: Integrate all target tag line segments that have passed the fine-tuning screening to generate a complete and effective operational flight strip. Based on business needs, cut out all track segments within the specified tag range from the complete flight strip, filter out the effective strip cutting range, and complete the automatic flight strip cutting.

8. The airborne hyperspectral data processing method for automatic cropping and correction according to claim 1, characterized in that, The geographic reference base map in step S3 is formed in the following way: The image distortion caused by lens optical distortion is eliminated based on the camera's own parameters. Then, the brightness and contrast of the images are adjusted under different lighting conditions. Cross-image matching is performed by comparing the similarity of feature points between different images. The geometric transformation relationship between images is calculated based on the matching results. After aligning multiple images, the pixel gray values ​​of the overlapping areas are smoothed to form an orthophoto map. By using the GPS positioning information associated with each image when it was captured, each pixel of the orthophoto map is assigned a corresponding geographic coordinate, forming a geographic reference map with a unified spatial benchmark.

9. The airborne hyperspectral data processing method with automatic cropping and correction according to claim 1, characterized in that, The generation of orthophoto hyperspectral data in step S4 is achieved through the following method: Step A1: Hyperspectral data block processing. The effective flight band hyperspectral data is divided into several data blocks according to a preset size. The GPS positioning information corresponding to the middle frame of each data block is extracted and converted into UTM coordinates. The image corresponding to each small block on the orthophoto map is cropped by combining the geographic pixel size of the hyperspectral image and the visible light orthophoto map. The UTM coordinates corresponding to the four corners of the cropped hyperspectral block can be calculated by using the UTM coordinates of the center block and the geographic pixel size. The pixel position is found by using the UTM coordinates, and then a slightly larger area is cropped according to the geographic range of the data block. Step A2: Image matching and transformation matrix acquisition: Select three channels from the segmented hyperspectral data to synthesize a visible light image or select a channel with a high signal-to-noise ratio as the matching image. Extract feature points from the visible light orthophoto image blocks corresponding to the geographic coordinates of the obtained hyperspectral image blocks and perform feature point matching. Obtain the transformation matrix based on the matching relationship and save it. Step A3: Orthorectified hyperspectral image generation. Traverse all hyperspectral data blocks, apply the corresponding spatial transformation matrix to the two-dimensional data array of each wavelength dimension, and generate a complete orthorectified hyperspectral image with UTM geographic coordinate attributes, thus completing the accurate association between spectral data and geographic space.

10. An airborne hyperspectral data processing system with automatic cropping and correction, employing the airborne hyperspectral data processing method with automatic cropping and correction as described in any one of claims 1-9, characterized in that, It includes a data acquisition module, an automatic flight strip trimming module, a geographic reference benchmark construction module, and an orthophoto hyperspectral data generation module; The data acquisition module is responsible for collecting data using a multi-source sensor system, including a hyperspectral camera, a visible light camera, GPS, and zenith light equipment, according to a predetermined flight path. The automatic flight strip trimming module is used to automatically identify core routes based on trajectory clustering algorithms, select eligible flight track clusters as target route segments, and complete the fine-grained selection and final trimming of flight strips. The geographic reference benchmark construction module is used to preprocess visible light images to eliminate optical distortion, adjust brightness and contrast, synthesize an orthophoto base map, and use GPS information to assign geographic coordinates to each pixel of the base map to form a unified spatial benchmark. The orthorectified hyperspectral data generation module is used to segment the obtained effective flight strip data into small blocks, perform preliminary matching with GPS and orthorectified base maps, then obtain transformation matrices through feature point extraction and matching, and finally apply these matrices to the data of each wavelength dimension to generate orthorectified hyperspectral images with UTM coordinates.

Images