Reconstruction method and system of water body hyperspectral image based on unmanned aerial vehicle carrying non-imaging sensor

Abstract

The application discloses a water body hyperspectral image reconstruction method and system based on a non-imaging sensor carried by a UAV. The application relates to the technical field of water surface hyperspectral image construction and solves the problem that the prior art cannot realize effective acquisition of UAV images in open water areas without land. The application carries a non-imaging hyperspectral detector based on a UAV platform, obtains spectral signals of each footprint uniformly distributed in the internal space of the water body in a non-imaging manner after planning a flight route, the wavelength range of the spectral signals is 200nm-1000nm, and the spectral resolution is 0.5nm-5nm; the spectral signals are converted into hyperspectral digital matrix text information containing spectral band data and corresponding spatial position information of each footprint; the hyperspectral digital matrix text information is indexed according to wavelength information, is converted into an image with a spatial data format, and is endowed with time attributes and spatial resolution information, so that reconstruction of the hyperspectral image of the water body in the area without land is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of water surface image construction, and in particular to a water body hyperspectral image reconstruction method and system based on a non-imaging sensor carried by a UAV. BACKGROUND

[0002] When a UAV is used to shoot an open water surface or a large area of a lake, it often needs to rely on land to realize the splicing of multiple images. When there is no land in the shot image, splicing cannot be realized under the existing technology. The core reason why water surface images cannot be effectively spliced is that the physical characteristics of the water surface scene and the limitations of the existing technology exist the following fundamental technical obstacles.

[0003] Existing splicing algorithms, such as SIFT (Scale Invariant Feature Transform Sift) and ORB (Oriented FAST and Rotated BRIEF), rely on texture features in images for same-name point matching. The water surface presents a mirror reflection characteristic in a windless state, and the texture information is extremely scarce. Even if there are waves, the dynamic texture generated by the random fluctuation of the waves, such as ripples, is also difficult to form stable feature points, resulting in that the algorithm cannot establish a reliable spatial correspondence. Generally, when the water surface accounts for more than 1 / 3 of a single image, the traditional two-dimensional mapping method directly fails. The attitude change of the UAV during flight will introduce perspective distortion. When the water surface lacks a reference system, the algorithm cannot eliminate this distortion through feature point constraints. Taking DJI ZhiTu as an example, when the image overlap rate is insufficient or the camera viewing angle difference is too large, the space three calculation will collapse due to the lack of effective control points. This problem is particularly prominent in wide water body scenes, because long flight lines will amplify cumulative errors. The intensity of water surface reflection changes dynamically with the solar elevation angle and the observation angle, especially the larger the observation angle, which makes it difficult to correct the radiation difference, resulting in the failure of traditional uniform light algorithms such as histogram matching. Therefore, it is necessary to propose a water body hyperspectral image reconstruction method and system based on a non-imaging sensor carried by a UAV to solve the above problems. SUMMARY

[0004] The purpose of the present application is to provide a water body hyperspectral image reconstruction method and system based on a non-imaging sensor carried by a UAV to solve the problem that the existing technology cannot realize effective acquisition of UAV images in open water areas without land.

[0005] In a first aspect, the present application provides a water body hyperspectral image reconstruction method based on a non-imaging sensor carried by a UAV, comprising:

[0006] Step one, based on the unmanned aerial vehicle platform carrying non-imaging hyperspectral detector, after planning the flight route, the spectral signal of each footprint uniformly distributed inside the water body space is obtained in a non-imaging way, the wave band range of the spectral signal is 200nm-1000nm, and the spectral resolution is 0.5nm-5nm;

[0007] Step two, the spectral signal obtained in step one is converted into hyperspectral digital matrix text information, and the hyperspectral digital matrix text information contains spectral wave band data and corresponding spatial position information of each footprint;

[0008] Step three, the hyperspectral digital matrix text information obtained in step two is indexed according to the wave band information, converted into an image with spatial data format, and given time attribute and spatial resolution information, so as to realize the reconstruction of the hyperspectral image of the water body in the unlanded area.

[0009] Further, step one includes:

[0010] The unmanned aerial vehicle is controlled to fly to a preset height from the water surface, the preset height is 60m, and the diameter of a single footprint is 1.5m-2m;

[0011] The non-imaging hyperspectral detector collects the radiation brightness value of the wave band of 200nm-1000nm with a spectral resolution of 0.5nm-5nm to form the radiation brightness value of each footprint, and when the flight height is 60m, the footprint radiation brightness value spatial distribution value with a diameter of 1.5m-2m is formed.

[0012] Further, the method further includes:

[0013] Based on the radiation brightness value of each footprint at a certain wavelength, spatial resampling is performed to obtain the spatial grid data of the radiation brightness value of the wavelength;

[0014] The spatial grid data of any one wavelength of 200nm-1000nm is reconstructed according to the requirement;

[0015] The spatial grid data of different wavelengths is arranged in the order of wavelength to form a hyperspectral data set.

[0016] Further, step two includes:

[0017] The spectral signal of each footprint is labeled with spatial position, and the corresponding relationship between the spectral wave band data and the footprint spatial coordinates is established;

[0018] The labeled spectral wave band data is arranged according to the spatial distribution rule to generate the hyperspectral digital matrix text information.

[0019] Further, the spectral signal of each footprint is marked with spatial position, and the corresponding relationship between spectral band data and footprint spatial coordinates is established, including:

[0020] Based on the GPS positioning data of the unmanned aerial vehicle, the latitude and longitude coordinates of each footprint are recorded.

[0021] The latitude and longitude coordinates are bound with the spectral band data of the corresponding footprint to form original data with spatial index.

[0022] Further, in step three, the spatial data format is tif format.

[0023] Further, in step three, according to the requirement, the specific band hyperspectral digital matrix text information is called to reconstruct the hyperspectral image of the water body in the non-land area.

[0024] In the second aspect, the application provides a water body hyperspectral image reconstruction system based on an unmanned aerial vehicle carrying a non-imaging sensor, comprising: an unmanned aerial vehicle, a non-imaging hyperspectral detector and a processor.

[0025] The unmanned aerial vehicle is used to carry the non-imaging hyperspectral detector based on the unmanned aerial vehicle platform.

[0026] The non-imaging hyperspectral detector is used to obtain the spectral signal of each footprint uniformly distributed in the space of the water body in a non-imaging manner after planning the flight route, the spectral signal has a band range of 200nm-1000nm and a spectral resolution of 0.5nm-5nm.

[0027] The processor is used to convert the spectral signal into hyperspectral digital matrix text information, the hyperspectral digital matrix text information contains spectral band data and corresponding spatial position information of each footprint; the hyperspectral digital matrix text information is indexed according to the band information, converted into an image with spatial data format, and given time attribute and spatial resolution information, so as to realize the reconstruction of the hyperspectral image of the water body in the non-land area.

[0028] The present application has the following beneficial effects: the water body hyperspectral image reconstruction method and system based on unmanned aerial vehicle carrying non-imaging sensors, by carrying hyperspectral detectors based on non-imaging unmanned aerial vehicle platform, the spectral signal of 200nm-1000nm band is obtained in the form of footprints and the image is reconstructed, effectively solving the image stitching failure problem caused by lack of water surface texture, perspective distortion constraint and radiation difference correction in the prior art in the open water area without land. Specifically, the method does not need to rely on land feature point matching, by directly obtaining the spectral signal uniformly distributed in the water body space and converting it into digital matrix text information, avoiding the dependence of traditional imaging stitching algorithm on texture features, overcoming the defect that two-dimensional mapping method is invalid when the water surface ratio of single image is too high; at the same time, by collecting spectral signal in a non-imaging way, the influence of perspective distortion introduced by unmanned aerial vehicle attitude change and long flight cumulative error on aerial triangulation calculation is avoided, and the wide band coverage of 200nm-1000nm and the high spectral resolution of 0.5nm-5nm make up for the limitation of the band range of the prior art, which can provide more rich spectral information for water quality detection. In addition, the method reconstructs spatial grid data of any wavelength by band index, can flexibly call specific band information according to demand, reduces data storage space and improves response speed, realizes efficient and accurate reconstruction of water body hyperspectral image in landless area, and provides a reliable technical means for large-area water environment monitoring. BRIEF DESCRIPTION OF DRAWINGS

[0029] In order to more clearly illustrate the technical solutions of the present application, the drawings needed in the embodiments will be briefly introduced below, and obviously, other drawings can also be obtained by those skilled in the art without creative labor on the premise of not paying the creative labor.

[0030] Figure 1 The flow chart of the water body hyperspectral image reconstruction method based on unmanned aerial vehicle carrying non-imaging sensors;

[0031] Figure 2 The field of view diagram of the non-imaging hyperspectral detector;

[0032] Figure 3 The reconstruction principle diagram of the water body hyperspectral image in the landless area. DETAILED DESCRIPTION

[0033] In order to make the objects, technical solutions and advantages of the present application clearer, the technical solutions of the present application will be described clearly and completely below in combination with specific embodiments of the present application and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor fall within the protection scope of the present application. The technical solutions provided by the embodiments of the present application will be described in detail below in combination with the drawings.

[0034] Please refer to Figure 1 The embodiment of the present application provides a water hyperspectral image reconstruction method based on an unmanned aerial vehicle carrying a non-imaging sensor, which comprises the following steps:

[0035] In step one, a non-imaging hyperspectral detector is carried based on an unmanned aerial vehicle platform, and after planning a flight route, spectral signals of each footprint uniformly distributed inside the water body are acquired in a non-imaging manner, wherein the wave band range of the spectral signals is 200nm-1000nm, and the spectral resolution is 0.5nm-5nm.

[0036] During the flight of the unmanned aerial vehicle, a series of radiation brightness values are recorded every certain period of time. Since there is a field of view angle, the field of view of the non-imaging hyperspectral detector is a circle, and the circle is the footprint in the present application. The diameter of the circle changes with the change of the flight height of the airplane. Please refer to Figure 2 and Figure 3In the technical path of acquiring water body data, there are significant differences between non-imaging and traditional imaging methods, which directly affect data quality and application effect. The core principle of non-imaging method is to collect circular fields of view, i.e. footprints, one by one through sensors. Each footprint can be regarded as an independent observation unit for a local area of the water body. In the data processing stage, the system will arrange a large number of circular footprints in a grid according to the spatial position relationship, and finally splice them to form a square data image covering the target water area. The advantage of this method is that the collection parameters of each footprint have high consistency, and there is no edge error caused by field splicing, so the generated image performs better in spatial continuity and parameter uniformity. In contrast, the traditional imaging method usually relies on optical cameras or remote sensing devices to take pictures of the water body in a planar manner, and the raw data obtained is continuous image frames. Due to the fact that the light and shadow conditions on the surface of the water body are easily affected by factors such as the solar elevation angle, the shooting angle, cloud cover, water surface ripples, etc., there may be obvious differences in brightness and color tone between different image frames. In the later splicing process, although the image edges are aligned through algorithms, it is difficult to completely eliminate the problem of uneven light and shadow, which ultimately leads to the appearance of patches with different color depths in the panoramic image after splicing. The present application directly collects spectral signals through non-imaging detection method, breaking through the technical bottleneck of traditional imaging relying on land texture features. The wide band coverage of 200nm-1000nm and the high spectral resolution of 0.5nm-5nm ensure the integrity and fineness of the water body spectral information, providing a high-quality data foundation for subsequent image reconstruction.

[0037] Specifically, the unmanned aerial vehicle is controlled to fly to a preset height from the water surface, and when the preset height is 60m, the diameter of a single footprint is 1.5m-2m. If the flight height is 60m and the diameter of the footprint is 2m, the recommended spatial resolution of the grid is 2m. This matching design of height and footprint diameter ensures the uniformity of the spatial distribution of the footprint on the water surface, avoiding both insufficient spatial resolution caused by too high height and reduced flight efficiency caused by too low height, while the footprint diameter of 1.5m-2m can cover a large area of water on the premise of ensuring data density. The non-imaging hyperspectral detector collects the radiance values of the 200nm-1000nm band with a spectral resolution of 0.5nm-5nm to form the spectral signal of each footprint. The non-imaging detection method directly acquires spectral data without imaging through image sensors, fundamentally avoiding the interference of water surface reflection and dynamic texture on imaging quality, and the spectral resolution of 0.5nm-5nm ensures the fine distinction of radiance values at different wavelengths, providing rich spectral dimension information for water quality parameter inversion.

[0038] The method further comprises: based on the radiation brightness value of each footprint at a certain wavelength, performing spatial resampling to obtain spatial grid data of the radiation brightness value at the wavelength; reconstructing spatial grid data of any one of 200nm-1000nm according to requirements; and arranging the spatial grid data of different wavelengths in sequence along the wavelength to form a hyperspectral data set. The spatial resampling step converts discrete footprint data into continuous grid data through an interpolation algorithm or the like, thereby solving the spatial discreteness problem of non-imaging detection data, and the ability to reconstruct grid data of any wavelength according to requirements improves the flexibility of data use, and users can select a target waveband according to specific application requirements, thereby avoiding the redundancy of full-waveband data storage and processing, and the finally formed hyperspectral data set completely retains the three-dimensional characteristics of water body spectra.

[0039] Step two: converting the spectral signal obtained in step one into hyperspectral digital matrix text information, wherein the hyperspectral digital matrix text information comprises spectral waveband data of each footprint and corresponding spatial position information.

[0040] Specifically, the spectral signal of each footprint is labeled with a spatial position, and a corresponding relationship between the spectral waveband data and the spatial coordinates of the footprint is established; and the labeled spectral waveband data is arranged according to a spatial distribution rule to generate hyperspectral digital matrix text information. The spatial position labeling binds the spectral data with the latitude and longitude coordinates, thereby ensuring the spatial locatability of each data point, and the matrix text information generated according to the spatial distribution rule realizes the structured storage of data, facilitates subsequent fast indexing and calling according to the waveband, and improves the efficiency of data processing. When the spectral signal of each footprint is labeled with a spatial position, the latitude and longitude coordinates of each footprint are recorded based on the GPS positioning data of the unmanned aerial vehicle, and the latitude and longitude coordinates are bound with the spectral waveband data of the corresponding footprint to form original data with a spatial index. The GPS positioning data provides a high-precision geographical reference for the spatial position, and the binding of the latitude and longitude coordinates with the spectral data ensures the spatial consistency of the data from the collection to the reconstruction process, thereby avoiding the accumulation of spatial errors caused by perspective distortion in traditional image stitching.

[0041] The present application converts discrete spectral signals into structured text information by establishing the association between the spectral data and the spatial position, thereby retaining the radiation brightness value of each footprint and realizing the spatial indexing of the data through the spatial position information, laying a data organization foundation for subsequent waveband reconstruction of images, and avoiding the invalidation of the spatial correspondence caused by the missing of feature points in traditional image stitching. The present application records water body features through text-based digital information, and the memory occupation of non-imaging data is much lower than that of image data. Compared with the traditional method, the non-imaging method adopted by the present application has a significant advantage in data storage efficiency.

[0042] Step three, the hyperspectral digital matrix text information obtained in step two is indexed according to the wave band information, converted into an image with a spatial data format, and given time attribute and spatial resolution information, so as to realize reconstruction of the hyperspectral image of the water body in the landless area.

[0043] The present application converts the text information into an image with a spatial data format by wave band indexing, directly skipping the complex process of traditional image splicing, and the time attribute and spatial resolution information given ensure the spatio-temporal traceability of the reconstructed image, finally realizing efficient acquisition of the hyperspectral image of the water body in the landless area, and solving the core problem that the prior art cannot splice in open water.

[0044] Specifically, the spatial data format is tif format. The tif format is selected as the output format, which not only ensures the compatibility of the image data and facilitates subsequent analysis by using mainstream remote sensing software, but also completely retains the metadata such as wave band information, spatial resolution and time attribute, and ensures the usability and interchangeability of the reconstructed image. Meanwhile, the hyperspectral digital matrix text information of specific wave bands is called according to requirements to reconstruct the hyperspectral image of the water body in the landless area, and this feature enables the user to flexibly select wave band combinations according to specific application scenarios, reduces the cost of data transmission and storage, and improves the system response speed.

[0045] In order to ensure the accuracy of the water body data and eliminate the error caused by the change of sunlight, the present application designs two data calibration methods. The first method is to integrate a downlink radiation brightness sensor on the top of the unmanned aerial vehicle. The second method is to set a standard white board with known reflectivity as a reference datum in the visible range of the monitoring area when the radiation sensor cannot be deployed, so as to further reduce the system error caused by the fluctuation of solar radiation.

[0046] The embodiment of the present application provides a water body hyperspectral image reconstruction system based on an unmanned aerial vehicle carrying a non-imaging sensor, which is characterized by comprising: an unmanned aerial vehicle, a non-imaging hyperspectral detector and a processor. The unmanned aerial vehicle is used for carrying the non-imaging hyperspectral detector based on the unmanned aerial vehicle platform. The non-imaging hyperspectral detector is used for acquiring the spectral signal of each footprint uniformly distributed in the space of the water body in a non-imaging manner after planning a flight route, the wave band range of the spectral signal is 200nm-1000nm, and the spectral resolution is 0.5nm-5nm. The processor is used for converting the spectral signal into hyperspectral digital matrix text information, the hyperspectral digital matrix text information contains spectral wave band data and corresponding spatial position information of each footprint; the hyperspectral digital matrix text information is indexed according to the wave band information, converted into an image with a spatial data format, and given time attribute and spatial resolution information, so as to realize reconstruction of the hyperspectral image of the water body in the landless area.

[0047] From the above embodiment, the application proposes a water hyperspectral image reconstruction technical route based on a non-imaging unmanned aerial vehicle platform, through a non-imaging detection, digital matrix conversion and band index reconstruction method, the dependence on land feature points of traditional imaging splicing is broken, and the technical problem that water body images cannot be effectively obtained in non-land areas is solved. The application replaces imaging shooting with non-imaging spectral collection, fundamentally avoiding the splicing failure problem caused by the lack of water surface texture and dynamic change; the application realizes the deep binding of spectral data and spatial position through hyperspectral digital matrix text information, providing a structured data basis for flexible image reconstruction according to bands; the application realizes the expansion of band coverage, and the fine sampling of 200nm-400nm ultraviolet band fills the band blank of the prior art, improving the spectral dimension of water environment monitoring.

[0048] The above-mentioned embodiments of the application do not constitute a limitation on the protection scope of the application.

Claims

1. A method for reconstructing hyperspectral images of water bodies based on a UAV equipped with a non-imaging sensor, characterized in that, include: Step 1: Based on the UAV platform equipped with a non-imaging hyperspectral detector, after planning the flight route, the spectral signals of each footprint uniformly distributed in the interior space of the water body are acquired in a non-imaging manner. The wavelength range of the spectral signals is 200nm-1000nm, and the spectral resolution is 0.5nm-5nm. Step 2: Convert the spectral signal obtained in Step 1 into hyperspectral digital matrix text information, which includes spectral band data of each footprint and corresponding spatial location information. Specifically, the spatial location of the spectral signal of each footprint is marked to establish the correspondence between the spectral band data and the spatial coordinates of the footprint; the marked spectral band data are arranged according to the spatial distribution pattern to generate hyperspectral digital matrix text information. Spatial location annotation is performed on the spectral signals of each footprint to establish the correspondence between spectral band data and footprint spatial coordinates. This includes: recording the latitude and longitude coordinates of each footprint based on the GPS positioning data of the UAV; binding the latitude and longitude coordinates with the spectral band data of the corresponding footprint to form raw data with spatial index. Step 3: The hyperspectral digital matrix text information obtained in Step 2 is indexed by band information, converted into an image with spatial data format, and given temporal attributes and spatial resolution information to realize the reconstruction of hyperspectral images of water bodies in landless areas. Among these methods, the required hyperspectral digital matrix text information for the specified bands is retrieved to reconstruct hyperspectral images of water bodies in landless areas.

2. The method for reconstructing water hyperspectral images based on a UAV equipped with a non-imaging sensor as described in claim 1, characterized in that, Step one includes: Control the drone to fly to a preset height above the water surface. When the preset height is 60m, the diameter of a single footprint is 1.5m-2m. The radiance values ​​in the 200nm-1000nm band are collected with a spectral resolution of 0.5nm-5nm using a non-imaging hyperspectral detector, forming the radiance values ​​of each footprint. At a flight altitude of 60m, the spatial distribution values ​​of the footprint radiance values ​​are formed, with a diameter of 1.5m-2m, arranged neatly in space.

3. The method for reconstructing water hyperspectral images based on a UAV equipped with a non-imaging sensor as described in claim 1, characterized in that, The method further includes: Based on the radiance values ​​of different wavelengths on each footprint, spatial resampling is performed to obtain spatial raster data of the radiance values ​​at that wavelength. Reconstruct spatial raster data for any wavelength between 200nm and 1000nm according to requirements; Spatial raster data of different wavelengths are arranged in wavelength order to form a hyperspectral dataset.

4. The method for reconstructing water hyperspectral images based on a UAV equipped with a non-imaging sensor as described in claim 1, characterized in that, In step three, the spatial data format is TIFF.

5. A water hyperspectral image reconstruction system based on a UAV equipped with a non-imaging sensor, characterized in that, include: Unmanned aerial vehicles (UAVs), non-imaging hyperspectral detectors, and processors; The drone is used to carry a non-imaging hyperspectral detector on a drone platform. The non-imaging hyperspectral detector is used to acquire the spectral signals of each footprint uniformly distributed in the interior space of the water body in a non-imaging manner after planning the flight path. The spectral signal has a wavelength range of 200nm-1000nm and a spectral resolution of 0.5nm-5nm. The processor is configured to convert the spectral signal into hyperspectral digital matrix text information, which includes spectral band data of each footprint and corresponding spatial location information. Specifically, the processor performs spatial location annotation on the spectral signal of each footprint to establish a correspondence between spectral band data and footprint spatial coordinates; the annotationated spectral band data is arranged according to spatial distribution patterns to generate hyperspectral digital matrix text information; the process of spatially annotating the spectral signal of each footprint to establish a correspondence between spectral band data and footprint spatial coordinates includes: recording the latitude and longitude coordinates of each footprint based on GPS positioning data from the UAV; and binding the latitude and longitude coordinates with the corresponding footprint's spectral band data to form raw data with spatial index. The hyperspectral digital matrix text information is indexed by band information, converted into an image with spatial data format, and given temporal attributes and spatial resolution information to reconstruct hyperspectral images of water bodies in landless areas; wherein, according to the requirements, the required band hyperspectral digital matrix text information is called to reconstruct hyperspectral images of water bodies in landless areas.

Images