Detection Method and Image Acquisition Device for Submerged Wetland Plants Using Hyperspectral Measurement

Abstract

This invention discloses a method and image acquisition device for detecting submerged plants in wetlands using hyperspectral imaging. The detection method includes: S1 acquiring hyperspectral images of submerged plants in the area to be monitored, the water depth above the canopy of the submerged plants, and the transparency of the area where the submerged plants are located; S2 preprocessing and data conversion of the hyperspectral images, and then calculating multiple spectral indices and multiple vegetation indices of the submerged plants based on the spectral information in the hyperspectral images; S3 calculating the transparency and the water depth above the canopy of the submerged plants based on the water depth H. i S4 Calculate the hyperspectral characteristic vegetation index of submerged plants; S5 Select the characteristic spectral bands of the hyperspectral information of submerged plants using the stepwise discriminant analysis method SDA-R based on information entropy, based on multiple spectral indices, multiple vegetation indices and hyperspectral characteristic vegetation index; S6 Input the selected characteristic spectral bands into the trained classification decision tree to identify the category of submerged plants.

Description

Technical Field

[0001] This invention relates to the field of wetland submerged plant survey and measurement technology, and in particular to a detection method and image acquisition device for wetland submerged plants using hyperspectral imaging. Background Technology

[0002] Submerged wetland plants are important primary producers in aquatic ecosystems, significantly mitigating eutrophication by effectively inhibiting algal growth and enhancing the self-purification capacity of water bodies. They have become a crucial biological measure for controlling and managing lake eutrophication. With the increasing severity of wetland degradation, aquatic ecosystem damage, and water pollution, the restoration of submerged plants has become a key focus of ecological protection, restoration, and repair of degraded wetlands.

[0003] In the investigation and measurement of submerged plants in wetlands, traditional methods mainly employ quadrat surveys, remote sensing satellite interpretation, and underwater acoustic methods. However, these traditional survey and monitoring methods have limitations such as being time-consuming, labor-intensive, and easily affected by weather. Specifically: for quadrat surveys of submerged plants, obtaining information on submerged plants in the surveyed area requires the use of sampling devices and aquatic plant clips for underwater sampling. This is not only labor-intensive and time-consuming, but the sampling work is also often limited by adverse weather conditions. Furthermore, quantitative quadrat surveys typically involve uprooting submerged plants within the quadrat, which can disrupt the plant community structure and function, and interfere with the diversity of submerged plants in the same water area, making scientific comparative observation difficult. In the interpretation of remote sensing satellite data for submerged plants, due to objective limitations, the spatiotemporal resolution of remote sensing imagery data currently used for thematic interpretation is relatively low, and it is often affected by weather conditions, making it difficult to reflect the current distribution status of submerged plants in wetlands in a timely and accurate manner. Especially during the peak growth season for submerged plants, the surface of wetland water bodies is often largely covered by cyanobacteria. When the spectral characteristics of cyanobacteria and submerged plants are similar, it can severely affect the accuracy of remote sensing interpretation of plant species results. While underwater acoustic methods can avoid destructive monitoring of submerged plants in wetlands, they can only monitor the height and coverage of submerged plants and cannot accurately determine characteristics such as plant species. Summary of the Invention

[0004] To address the aforementioned problems in existing technologies, the present invention provides a method and image acquisition device for detecting submerged plants in wetlands using hyperspectral imaging, which solves the problem that the investigation and measurement of submerged plants in existing technologies is greatly affected by weather conditions.

[0005] To achieve the above-mentioned objectives, the technical solution adopted by this invention is as follows:

[0006] Firstly, a method for detecting submerged plants in wetlands using hyperspectral imaging is provided, comprising the following steps:

[0007] S1. Collect hyperspectral images of submerged plants in the area to be monitored, the water depth above the canopy of submerged plants, and the transparency of the area where submerged plants are located.

[0008] S2. Perform preprocessing and data conversion on the hyperspectral image, and then calculate multiple spectral indices and multiple vegetation indices of wetland submerged plants based on the spectral information in the hyperspectral image.

[0009] S3. Based on transparency and water depth H above the submerged plant canopy. i Calculate the hyperspectral vegetation index of submerged plants:

[0010]

[0011]

[0012] Among them, HVIs are hyperspectral characteristic vegetation indices; R 798 R 670 R 550 Spectral reflectance at wavelengths of 798 nm, 670 nm, and 550 nm, respectively; ΔT is the illumination correction factor; k is the transparency attenuation factor; e is the natural logarithm; T i For transparency;

[0013] S4. Based on multiple spectral indices, multiple vegetation indices, and hyperspectral characteristic vegetation indices, the characteristic spectral bands of submerged plant hyperspectral information are selected using the information entropy-based step discriminant analysis method SDA-R.

[0014] S5. Input the selected feature spectral bands into the trained classification decision tree to identify the category of submerged plants.

[0015] Secondly, an image acquisition device is provided for a detection method for measuring submerged plants in wetlands using hyperspectral imaging. The device includes a hull and a control module that communicates with a ground control console via a wireless transmission module. The control module is connected to the hull's power mechanism, navigation control system, and lifting mechanism. A hyperspectral imager and a transparency sensor are installed on the lifting rod of the lifting mechanism.

[0016] Multiple ultrasonic rangefinders positioned at different locations on the boom constitute the boom obstacle avoidance system; data collected by the hyperspectral imager and transparency sensor are transmitted to the ground control console via the control module and wireless transmission module; data collected by the ultrasonic rangefinders are also transmitted to the control module.

[0017] The beneficial effects of this invention are as follows: When identifying submerged plant types, the light correction coefficient and transparency attenuation coefficient introduced in this solution can dynamically correct the hyperspectral vegetation index of submerged plants in real time, thereby obtaining the effective spectral reflectance characteristic value of each submerged plant; the vegetation index realizes the light-corrected chlorophyll absorption ratio index, which can effectively eliminate the influence of background environmental interference factors such as water depth, light and chlorophyll of underwater submerged plant canopy, thereby ensuring the accuracy of subsequent submerged plant identification.

[0018] The image acquisition device, through an underwater hyperspectral imager and a transparency sensor, enables simultaneous imaging of wetland submerged plants and monitoring of water environment indicators, providing technical support for the study of the ecological response mechanism of wetland submerged plants under water environment changes. The transparency sensor and the underwater hyperspectral imager are fixed on a lifting rod, and by controlling the lifting of the rod, measurement points can be selected at a fixed depth and distance, making the measurement more flexible and the information obtained more accurate, without causing damage to the wetland submerged plants and their surrounding ecological environment, thus improving environmental protection.

[0019] The measured images of submerged wetland plants and their surrounding transparency information can be transmitted to the ground control station in real time via a wireless transmission system, making the measurement process more efficient. This provides an effective way to quickly identify the species, community composition and distribution of submerged wetland plants, as well as their diversity dynamics. Furthermore, it enables real-time monitoring, facilitating the real-time examination of underwater submerged wetland plant images and water environment monitoring data. It also provides conditions for intuitive, rapid, and real-time understanding of the monitoring work. Attached Figure Description

[0020] Figure 1 This is a flowchart of a detection method for determining submerged plants in wetlands using hyperspectral imaging.

[0021] Figure 2 This is a top view of an image acquisition device used for measuring submerged plants in wetlands using hyperspectral imaging.

[0022] Figure 3 A side view of an image acquisition device for measuring submerged plants in wetlands using hyperspectral imaging.

[0023] The components include: 1. First hull; 2. Second hull; 3. Battery; 5. Ducted propeller; 6. Navigation controller; 7. Lifting mechanism; 71. Control module; 72. Gear motor; 73. Gear; 74. Rack; 8. Expansion interface; 9. Remote control signal receiver; 10. Lifting mast; 11. Satellite signal receiver; 14. Transparency sensor; 15. Hyperspectral imager; 16. Mounting platform; and 17. Wireless transmission module. Detailed Implementation

[0024] The specific embodiments of the present invention are described below to enable those skilled in the art to understand the present invention. However, it should be understood that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, various changes are obvious as long as they are within the spirit and scope of the present invention as defined and determined by the appended claims. All inventions utilizing the concept of the present invention are protected.

[0025] refer to Figure 1 , Figure 1 A method for detecting submerged plants in wetlands using hyperspectral imaging is shown, the method comprising steps S1 to S5.

[0026] In step S1, hyperspectral images of submerged plants in the area to be monitored, the water depth above the canopy of submerged plants, and the transparency of the area where submerged plants are located are acquired.

[0027] In step S2, the hyperspectral image undergoes preprocessing and data conversion. In practice, the preferred method for preprocessing the hyperspectral image includes:

[0028] Radiometric correction was performed on the spectra of wetland submerged plants collected in the field using standard white reference and black measurement values, and noise was removed using the Savitzky-Golay filtering algorithm. The original hyperspectral reflectance data were then smoothed.

[0029] Methods for data conversion of preprocessed hyperspectral images include:

[0030] The smoothed spectral reflectance information of submerged plants is subjected to first-derivative transformation d(R) and logarithmic transformation ln(R):

[0031]

[0032] ln(R)=ln(r1),ln(r2),…,ln(r m )]

[0033] Where R is the spectral reflectance value of submerged plants; r i λ is the wavelength of band i in the nth band; m is the total number of bands; Δλ is twice the band spacing.

[0034] Then, based on the spectral information in the hyperspectral image, multiple spectral indices and multiple vegetation indices of submerged wetland plants were calculated; the multiple spectral characteristic indices included 15, namely:

[0035] Average spectral reflectance (BMV) in the blue light band (360–500 nm), average spectral reflectance (GMV) in the green light band (510–570 nm), average spectral reflectance (RMV) in the red light band (650–670 nm), average spectral reflectance (NMV) in the near-infrared band (680–760 nm), maximum green-edge spectral reflectance (GPV), maximum red-edge spectral reflectance (RPV), maximum near-infrared spectral reflectance (NPV), maximum green-edge value of the first derivative (DGEV), maximum red-edge value of the first derivative (DREV), maximum green-edge value of the natural logarithm (LGEV), maximum red-edge value of the natural logarithm (LREV), ratio of near-infrared to maximum green-edge spectral reflectance (NGP), position of maximum red-edge value of the first derivative (FDRP), position of maximum red-edge value of the spectrum (REP), and position of maximum green-edge value of the spectrum (GEP).

[0036] The vegetation indices include three: Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), and Ratio Vegetation Index (RVI). The formulas for calculating these three vegetation indices are as follows:

[0037] NDVI=(NMV-RMV) / (NMV+RMV), RVI=NMV / RMV

[0038] EVI=2.5(NMV-RMV) / (NMV+6RMV-7BMV+1)

[0039] In step S3, based on the transparency and the water depth H above the submerged plant canopy... i Calculate the hyperspectral vegetation index of submerged plants:

[0040]

[0041]

[0042] Among them, HVIs are hyperspectral characteristic vegetation indices; R 798 R 670 R 550 Spectral reflectance at wavelengths of 798 nm, 670 nm, and 550 nm, respectively; ΔT is the illumination correction factor; k is the transparency attenuation factor; e is the natural logarithm; T i For transparency;

[0043] The illumination correction coefficient and transparency attenuation coefficient constructed in this scheme can eliminate the influence of interference factors related to underwater illumination and background environment.

[0044] In step S4, based on multiple spectral indices, multiple vegetation indices, and hyperspectral characteristic vegetation indices, the characteristic spectral bands of submerged plant hyperspectral information are selected using the information entropy-based step discriminant analysis method SDA-R.

[0045] The method employed in this study utilizes the SDA-R algorithm, which is based on information entropy. This algorithm can remove irrelevant and redundant information, thereby improving the efficiency of the underwater submerged plant identification and analysis model. SDA-R is a multivariate statistical analysis method. Based on discriminant analysis, each step introduces the variable with the best discriminant analysis capability into the discriminant function and removes the variable with the worst discriminant capability. In this way, the optimal band is selected using the SDA calculation results, and finally, the spectral band information of submerged plants is formed.

[0046] In step S5, the selected characteristic spectral bands are input into the trained classification decision tree to identify the category of submerged plants.

[0047] The training method for the classification decision tree includes: acquiring hyperspectral images of more than ten submerged plants (such as Vallisneria natans, Potamogeton malaianus, Hydrilla verticillata, and Potamogeton crispus, the specific number of species depending on the submerged plant species in the monitored waters) using a U285 UW underwater high-speed hyperspectral snapshot sensor. For each hyperspectral image, ten sample spectra are selected. Considering that illumination and water background environment are direct factors related to the hyperspectral variability of plant leaves, CUBE-Pilot software is used to process and randomly select 60 sample patches from the corresponding hyperspectral images. Each patch corresponds to 20*20 pixels and is used as the region of interest. The spectral reflectance of the sample patch is calculated as the average value of the selected three-dimensional pixels. 500 sample spectral values ​​are randomly selected. Using these sample spectral values, a 10x cross-validation method is used to train and validate the classification decision tree.

[0048] In implementation, this scheme preferably uses a classification decision tree to classify submerged plants, with the split point that maximizes information gain as the optimal split point for wetland submerged plants.

[0049]

[0050]

[0051] Where D is the dataset constructed using selected characteristic spectral bands; Ent(D) is the information entropy of dataset D; P k The proportion of the k-th type of feature spectral band in dataset D, k = 1, 2, ..., |y|, where |y| is the total number of types of feature spectral bands; Gain(D,a) is the information gain obtained by partitioning the sample set D using discrete attribute a; T a Let t be the partition point of dataset D on discrete attribute α; Gain(D,a,t) is the information gain of dataset D after bisecting based on partition point t. Let be the datasets ≤t and >t in dataset D, respectively; n is the number of possible values ​​for the discrete attribute.

[0052] In one embodiment of the present invention, a method for acquiring hyperspectral images of submerged plants in a monitored area includes:

[0053] S11. Divide the area to be monitored into several sub-areas, randomly select a preset number of sub-areas as the collection area, and randomly select several sampling points within the collection area.

[0054] S12. Control the image acquisition device to move to the initial sampling point in the preset trajectory;

[0055] S13, The lifting mechanism 7 of the image acquisition device moves downward with the hyperspectral imager 15 on it, and simultaneously acquires the distance between the lifting mechanism 7 and the surrounding obstacles.

[0056] S14. When the distance L is less than the safe distance, retract the hyperspectral imager 15, and then the image acquisition device moves in the opposite direction by 0.5 to 1 meter, returning to step S13.

[0057] S14. When the distance L is greater than or equal to the safe distance, when the hyperspectral imager 15 moves to the preset distance from the submerged water, a hyperspectral image of the submerged plant is captured, and then the process proceeds to step S15.

[0058] S15. After retracting the hyperspectral imager 15, determine whether all sampling points in the preset trajectory have completed hyperspectral image acquisition. If so, complete image acquisition; otherwise, return to step S13.

[0059] When using the above method for image acquisition, this scheme can select measurement points at a fixed depth and distance by controlling the lifting mechanism 7, making the measurement more flexible and the information obtained more accurate. It will not cause damage to wetland submerged plants and their surrounding ecological environment, thus improving environmental protection.

[0060] During data acquisition, obstacles around the lifting mechanism 7 are also assessed simultaneously to prevent them from damaging the hyperspectral imager 15, thus ensuring the safety of image acquisition.

[0061] like Figure 2 and Figure 3 As shown, this solution also provides an image acquisition device for detecting submerged plants in wetlands using hyperspectral imaging. The device includes a hull, and preferably, the hull includes a first hull 1 and a second hull 2 ​​arranged side by side, and a mounting platform 16 that can be detachably connected to the first hull 1 and the second hull 2. A lifting hole for the lifting rod 10 to pass through is provided through the middle of the mounting platform 16.

[0062] The catamaran design, consisting of the first hull 1 and the second hull 2 ​​arranged side by side, increases the ability to resist water flow impact. The mounting platform 16 provides a fixed condition for various instruments and equipment. The mounting platform 16 is detachably connected to the first hull 1 and the second hull 2, making it easy to assemble and disassemble, expandable, and highly flexible and functionally expandable.

[0063] The ship is equipped with a control module 71 that communicates with the ground control console via a wireless transmission module 17. The control module 71 is connected to the ship's power mechanism, navigation control system and lifting mechanism 7 on the ship. The lifting mechanism 7 includes a rack 74 fixed on the lifting rod 10. The rack 74 meshes with a gear 73. The gear 73 is driven to rotate by a reduction motor 72. The reduction motor 72 is electrically connected to the control module 71.

[0064] The lifting mechanism 7's lifting boom 10 is mounted on the hull and extends underwater. A hyperspectral imager 15 and a transparency sensor 14 are mounted on the lifting boom 10. To avoid poor underwater lighting conditions and unclear information acquisition, supplementary lighting equipment can also be installed on the lifting boom 10. The main component of the hyperspectral imager 15 is the German U285 UW underwater high-speed hyperspectral snapshot sensor. The U285 UW hyperspectral image consists of 50-50 pixel hyperspectral three-dimensional data and a 1000-1000 pixel resolution panchromatic image. This sensor can capture 138 spectral bands in the 450-998nm spectral range at 4nm intervals.

[0065] Multiple ultrasonic rangefinders positioned at different locations on the lifting mast 10 constitute its obstacle avoidance system. These rangefinders are positioned in six directions: front, rear, left, right, up, and down. The rangefinders monitor the distance to obstacles in each of these six directions to ensure the normal operation of the underwater hyperspectral imager 15 and transparency sensor 14 at the lower end of the lifting mast 10, preventing damage to the equipment. Data collected by the hyperspectral imager 15 and transparency sensor 14 is transmitted to the ground control console via the control module 71 and wireless transmission module 17; data collected by the ultrasonic rangefinders is also transmitted to the control module 71.

[0066] To facilitate direct reading of data from the image acquisition device or writing of data to it, an expansion interface 8 connected to the control module 71 can also be installed on the hull.

[0067] like Figure 3 As shown, the ship's power mechanism includes a ducted propeller 5 that is recessed at the bottom of the hull. The ducted propeller 5 is driven by a brushless motor, which is electrically connected to an electronic speed controller.

[0068] The ducted propeller 5 is installed in a recessed manner at the bottom of the catamaran hull, which can effectively avoid the problem of the propeller getting tangled in shallow water or when there is a lot of aquatic plants. The ducted propeller 5 has a duct on the outside of the propeller. The duct is wider at the front and narrower at the rear, which helps to increase the propulsion effect of the water flow and enhance the hull's sailing power. The electronic speed governor can control the power of the ducted propeller 5 according to the signal commands of the navigation control system, thereby achieving the purpose of adjusting the hull's sailing speed.

[0069] The preferred electronic speed controller is the EASYCO FS-70 brushless motor electronic speed controller; the preferred brushless motor model is CNS-BLD-2S-NC from Xingsong Technology. Batteries 3 are installed at the bottom of the first hull 1 and the second hull 2 ​​to power the entire unmanned surface vessel.

[0070] The navigation control system includes a remote control signal receiver 9, a navigation controller 6, and a satellite navigation system. The remote control signal receiver 9 receives control signals sent by the ground control console and transmits them to the navigation control command converter for signal conversion before inputting them into the navigation controller 6.

[0071] The navigation controller 6 is pluggable and detachable and has a route memory card for storing preset tracks. The navigation controller 6 is electrically connected to the electronic speed controller. The satellite navigation system includes a satellite signal receiver 11, which is electrically connected to the navigation controller 6.

[0072] The navigation controller 6 controls the ship's start, stop, speed, and direction of travel, enabling functions such as constant speed cruise and remote control cruise. By pre-storing route files on the route memory card, the ship can cruise according to a predetermined trajectory, achieving automatic monitoring of high-precision areas.

[0073] The ship is equipped with inertial navigation sensors that monitor its attitude, and these sensors are electrically connected to the navigation controller 6. The inertial navigation sensors detect and measure the ship's speed, tilt, impact, vibration, rotation, and other motion states in real time, and feed this information back to the navigation controller 6. The navigation controller 6 then performs real-time attitude control to ensure the ship's stable navigation.

Claims

1. A method for detecting submerged plants in wetlands using hyperspectral imaging, characterized by: Including the following steps: S1. Collect hyperspectral images of submerged plants in the area to be monitored, the water depth above the canopy of submerged plants, and the transparency of the area where submerged plants are located. S2. Perform preprocessing and data conversion on the hyperspectral image, and then calculate multiple spectral indices and multiple vegetation indices of wetland submerged plants based on the spectral information in the hyperspectral image. S3. Based on transparency and water depth above the submerged plant canopy. H i Calculate the hyperspectral vegetation index of submerged plants: ， in, HVIs Hyperspectral vegetation index; R 798 、R 670 、R 550 Spectral reflectance at wavelengths of 798nm, 670nm, and 550nm, respectively; This is the illumination correction factor; k This is the transparency attenuation coefficient; e It is the natural logarithm; T i For transparency; S4. Based on multiple spectral indices, multiple vegetation indices, and hyperspectral characteristic vegetation indices, the characteristic spectral bands of submerged plant hyperspectral information are selected using the information entropy-based step discriminant analysis method SDA-R. S5. Input the selected feature spectral bands into the trained classification decision tree to identify the category of submerged plants; when classifying submerged plants using the classification decision tree, the split point with the maximum information gain is taken as the optimal split point for wetland submerged plants: ， in, D This is a dataset constructed using selected characteristic spectral bands; For dataset D Information entropy; For dataset D The Middle k The proportion of characteristic spectral bands, , This represents the total number of types of characteristic spectral bands. Discrete properties a For sample set D The information gain obtained by partitioning; For dataset D In discrete properties α The dividing point on; For dataset D Based on the dividing point t Information gain after binary splitting; , Each is a dataset D Data sets containing ≤t and >t; n The number of possible values ​​for the discrete attribute.

2. The method for detecting submerged wetland plants using hyperspectral imaging according to claim 1, characterized in that: The multiple spectral characteristic indices include 15, namely: average spectral reflectance (BMV) in the blue light band (360~500nm), average spectral reflectance (GMV) in the green light band (510~570nm), average spectral reflectance (RMV) in the red light band (650~670nm), average spectral reflectance (NMV) in the near-infrared band (680~760nm), maximum green-edge spectral reflectance (GPV), maximum red-edge spectral reflectance (RPV), maximum near-infrared spectral reflectance (NPV), maximum green-edge value of the first derivative (DGEV), maximum red-edge value of the first derivative (DREV), maximum green-edge value of the natural logarithm (LGEV), maximum red-edge value of the natural logarithm (LREV), ratio of near-infrared to maximum green-edge spectral reflectance (NGP), position of maximum red-edge value of the first derivative (FDRP), position of maximum red-edge value of the spectrum (REP), and position of maximum green-edge value of the spectrum (GEP). The vegetation indices include three: Normalized Difference Vegetation Index (NDVI), Enhanced Vegetation Index (EVI), and Ratio Vegetation Index (RVI). The formulas for calculating these three vegetation indices are as follows: NDVI=(NMV-RMV) / (NMV+RMV),RVI= NMV / RMV EVI=2.5(NMV-RMV) / (NMV+6 RMV-7BMV+1).

3. The method for detecting submerged wetland plants using hyperspectral imaging according to claim 1, characterized in that: Methods for acquiring hyperspectral images of submerged plants in the monitored area include: S11. Divide the area to be monitored into several sub-areas, randomly select a preset number of sub-areas as the collection area, and randomly select several sampling points within the collection area. S12. Control the image acquisition device to move to the initial sampling point in the preset trajectory; S13. The lifting mechanism of the image acquisition device moves downward with the hyperspectral imager on it, and simultaneously acquires the distance between the lifting mechanism and the obstacles around it. S14. When the distance L is less than the safe distance, retract the hyperspectral imager, and then the image acquisition device moves in the opposite direction by 0.5 to 1 meter, returning to step S13. S14. When the distance L is greater than or equal to the safe distance, when the hyperspectral imager moves to the preset distance from the submerged water, a hyperspectral image of the submerged plant is captured, and then proceed to step S15. S15. After retrieving the hyperspectral imager, determine whether all sampling points in the preset trajectory have completed hyperspectral image acquisition. If so, complete image acquisition; otherwise, return to step S13.

4. The method for detecting submerged wetland plants using hyperspectral imaging according to claim 1, characterized in that: Methods for preprocessing hyperspectral images include: Radiometric correction was performed on the spectra of wetland submerged plants collected in the field using standard white reference and black measurement values, and noise was removed using the Savitzky-Golay filtering algorithm. The original hyperspectral reflectance data were then smoothed. Methods for data conversion of preprocessed hyperspectral images include: First derivative transformation of the smoothed spectral reflectance information of submerged plants. And logarithmic transformation : Where R is the spectral reflectance value of submerged plants; r i For the first n band i wavelength; m Δλ represents the total number of bands; Δλ is twice the band spacing.

5. An image acquisition device applied to the detection method for determining submerged wetland plants using hyperspectral imaging as described in any one of claims 1-4, characterized in that, The system includes a hull, on which a control module is mounted to communicate with a ground control console via a wireless transmission module. The control module is connected to the hull's power mechanism, navigation control system, and lifting mechanism. The lifting mechanism's lifting rod is equipped with a hyperspectral imager and a transparency sensor. The multiple ultrasonic rangefinders installed at different locations on the lifting pole constitute the lifting pole obstacle avoidance system; the data collected by the hyperspectral imager and transparency sensor are sent to the ground control console through the control module and wireless transmission module; the data collected by the ultrasonic rangefinders are sent to the control module.

6. The image acquisition device according to claim 5, characterized in that, The hull power mechanism includes a ducted propeller recessed at the bottom of the hull, which is driven by a brushless motor and is electrically connected to an electronic speed controller. The navigation control system includes a remote control signal receiver, a navigation controller, and a satellite navigation system. The remote control signal receiver receives control signals sent by the ground control console and transmits them to the navigation control command converter for signal conversion before inputting them into the navigation controller. The navigation controller is pluggable and detachable and has a route memory card for storing preset trajectories. The navigation controller is electrically connected to the electronic speed controller. The satellite navigation system includes a satellite signal receiver, which is electrically connected to the navigation controller.

7. The image acquisition device according to claim 5, characterized in that, The lifting mechanism includes a rack fixed to the lifting rod, the rack meshing with a gear, the gear being driven to rotate by a reduction motor, and the reduction motor being electrically connected to the control module.

8. The image acquisition device according to claim 5, characterized in that, The hull includes a first hull and a second hull arranged side by side, and a mounting platform that can be detachably connected to the first hull and the second hull. A lifting hole is provided through the middle of the mounting platform for the lifting rod to pass through.

Images