Remote sensing monitoring system for typical grassland degradation indicator species based on airborne multi-source spectral data

Abstract

The application provides a kind of based on airborne multi-source spectral data typical grassland degradation indicator remote sensing monitoring system, including remote sensing information acquisition device and operation control module.Remote sensing monitoring device includes: rotor unmanned aerial vehicle, drive component and two acquisition modules.Operation control module includes: image acquisition unit, semantic recognition neural network, auxiliary image acquisition unit, auxiliary semantic recognition neural network, feature introduction unit and control unit.The application provides the based on airborne multi-source spectral data typical grassland degradation indicator remote sensing monitoring system, can improve the monitoring precision and efficiency of grassland degradation indicator, to facilitate monitoring and evaluating the health status of grassland, promptly find and respond to the problem of grassland degradation.

Description

Technical Field

[0001] This invention belongs to the field of grassland degradation monitoring technology, and in particular relates to a remote sensing monitoring system for typical grassland degradation indicator species based on airborne multi-source spectral data. Background Technology

[0002] Grasslands are important ecosystems that play a vital role in maintaining the Earth's ecological balance and ensuring sustainable human development. Identifying degraded indicator species (such as wolfsbane) growing on grasslands can help monitor and assess grassland health, promptly identify and address grassland degradation issues, and facilitate appropriate protection and restoration measures.

[0003] Currently, drone remote sensing is commonly used to monitor degraded indicator species. However, this method has the following problems: First, when the drone flies at a high altitude, the visual recognition accuracy of degraded indicator species is low, and lowering the flight altitude will significantly increase the aerial photography time, increasing the cost of manpower and resources. Second, at a fixed drone altitude, since degraded indicator species usually grow interspersed with other vegetation, when the plant height of degraded indicator species is low, the clump diameter is small, or it is not in its flowering period, the distinction between degraded indicator species and other plants is not obvious, which will lead to a decrease in recognition accuracy and thus affect the early warning effect of the grassland degradation monitoring system. In addition, when the area to be monitored is large, if sufficient accuracy of the monitoring results is required, the flight time of the drone should be extended, thus occupying more manpower and time. Summary of the Invention

[0004] In view of this, the present invention aims to propose a remote sensing monitoring system for typical grassland degradation indicator species based on airborne multi-source spectral data, so as to improve the monitoring accuracy and efficiency of grassland degradation indicator species.

[0005] To achieve the above objectives, the technical solution created by this invention is implemented as follows:

[0006] A remote sensing monitoring system for typical grassland degradation indicator species based on airborne multi-source spectral data includes a remote sensing information acquisition device and a computing and control module; the remote sensing monitoring device includes: a rotary-wing UAV, a drive assembly, and two acquisition modules;

[0007] The operation control module includes:

[0008] An image acquisition unit is used to acquire at least three consecutive images from an acquired image sequence and mark the corresponding acquisition parameters;

[0009] A semantic recognition neural network is used to process at least three types of input images respectively and output the recognition result of typical grassland degradation indicator species. The semantic recognition neural network includes no more than a first preset number of convolutional layers.

[0010] An auxiliary image acquisition unit is used to select a corresponding low-altitude image and / or a normally acquired image according to the acquisition parameters.

[0011] An auxiliary semantic recognition neural network is used to process input low-altitude images and / or normally acquired images, and includes no less than a second preset number of convolutional layers, wherein the second preset number is much greater than the first preset number.

[0012] The feature introduction unit is used to compare the output of the last convolutional layer of the semantic recognition neural network with a preset threshold. When the output is less than the preset threshold, the unit obtains the output of the last convolutional layer of the auxiliary semantic recognition neural network, compares it with the preset threshold, determines the pixels that are greater than the preset threshold, expands it to the surrounding pixels, obtains the convolutional features of the pixels and the surrounding pixels from the previous convolutional layer, and inputs the convolutional features into the second-to-last convolutional layer of the semantic recognition neural network.

[0013] The control unit is used to determine whether a degenerate indicator species exists based on the output of the semantic recognition neural network; when a degenerate indicator species is determined to exist, the distance between the acquisition module and the body is adjusted by the drive component, so that the adjustment component adjusts the orientation of the acquisition component to achieve accurate scanning and monitoring.

[0014] Furthermore, the two acquisition modules are arranged in a mirror image on both sides of the rotorcraft drone's fuselage. A drive assembly is used to adjust the distance between the acquisition modules and the fuselage, and both ends of the drive assembly are connected to the two acquisition modules respectively. Each acquisition module includes: a carrier box, an adjustment assembly, and an acquisition assembly. The carrier box is located between two adjacent arms of the rotorcraft drone, with a mounting cutout on its bottom surface and mounting holes on its side wall near the arm. The adjustment assembly is used to adjust the orientation of the acquisition assembly according to the distance between the acquisition module and the fuselage. The adjustment assembly includes an adjustment shaft and two connecting... The adjustment shaft is rotatably mounted inside the carrier box. The adjustment shaft has a channel hole communicating with the mounting hole. Two connecting rods are located on opposite sides of the carrier box. The first end of each connecting rod is slidably connected to the arm of the rotary-wing UAV via a connecting seat, and the second end is slidably mounted inside the channel hole. The sliding of the second end of the connecting rod inside the channel hole drives the adjustment shaft to rotate inside the carrier box. A mounting strip extending out of the carrier box along the mounting cut is also provided on the side wall of the adjustment shaft. The acquisition component is detachably mounted on the mounting strip and includes an image sensor and a multispectral sensor.

[0015] Furthermore, a guide protrusion is provided on the side wall of the second end of the connecting rod, and a spiral guide groove for accommodating the guide protrusion is provided on the inner side wall of the channel hole.

[0016] Furthermore, the drive assembly includes a drive box and two drive rods. The drive box is detachably mounted on the body of the rotary-wing UAV, and drive holes are provided at both ends of the drive box. The two drive rods are located on both sides of the drive box, with the end of the drive rod closer to the drive box slidably mounted inside the drive hole, and the end of the drive rod away from the drive box connected to the connecting post at the top of the carrier box.

[0017] Furthermore, the drive assembly also includes: a drive motor, a drive gear, and two drive racks; the drive motor is mounted on the drive box, and the output shaft of the drive motor is inserted into the drive box; the drive gear and the two drive racks are both mounted inside the drive box, the drive gear is connected to the output shaft of the drive motor, and the two drive racks are located on both sides of the drive gear; any one of the drive racks meshes with the drive gear, and the two drive racks are respectively connected to two drive rods.

[0018] Furthermore, a limiting baffle is provided at the end of the drive rack away from the drive rod.

[0019] Furthermore, the acquisition component is provided with a mounting base, and the mounting base is provided with mounting bolts; the mounting strip is provided with a plurality of bolt holes for accommodating the mounting bolts, and the plurality of bolt holes are arranged along the length direction of the mounting strip.

[0020] Furthermore, the connecting seat includes two parallel connecting plates, and two rotatable limiting rollers are provided between the two connecting plates. The axes of the two limiting rollers are perpendicular to the connecting plates, and a receiving channel for accommodating the machine arm is formed between the two limiting rollers. The first end of the connecting rod is provided with a connecting shaft, the axis of the connecting shaft is parallel to the axis of the limiting rollers, and the two ends of the connecting shaft are rotatably connected to the two connecting plates respectively.

[0021] Furthermore, the remote sensing monitoring system for typical grassland degradation indicator species based on airborne multi-source spectral data also includes a multispectral processing module, which includes:

[0022] The image preprocessing unit is used to smooth the multispectral images acquired by the multispectral sensor;

[0023] The spectral segment classification unit is used to extract spectral segments from the smoothed multispectral image, identify the spectral segment angles of the spectral segments, and calculate the spectral segment similarity between every two spectral segments based on the spectral segment angles to classify the spectral segments.

[0024] The plant identification unit is used to query the spectral attributes of various spectral bands and identify the grassland plants corresponding to each spectral band based on the spectral attributes.

[0025] Compared with existing technologies, the remote sensing monitoring system for typical grassland degradation indicator species based on airborne multi-source spectral data described in this invention has the following advantages:

[0026] (1) The remote sensing monitoring system for indicator species of typical grassland degradation based on airborne multi-source spectral data created in this invention utilizes a relatively simple semantic recognition neural network to fuse the judgment results of multiple high-altitude images from different locations, thereby improving the accuracy of indicator species identification and judgment in low-precision high-altitude images and overcoming the impact of low resolution of high-altitude images on recognition accuracy. Simultaneously, an auxiliary semantic recognition neural network extracts features from low-altitude and normal images related to the images input to the semantic recognition neural network, and inputs the extracted features into the semantic recognition neural network. This allows the semantic recognition neural network to learn and utilize the extracted features, effectively compensating for the lack of high-altitude image samples and further improving the accuracy of indicator species judgment and identification.

[0027] (2) The remote sensing monitoring system for typical grassland degradation indicator species based on airborne multi-source spectral data described in this invention has a drive component in its remote sensing information acquisition device that can adjust the distance between the acquisition module and the aircraft body. Furthermore, the two acquisition modules are arranged in a mirror image on both sides of the aircraft body. Therefore, adjusting the position of the acquisition module avoids affecting the center of gravity of the rotary-wing UAV, thus enabling the UAV to achieve better flight stability. Secondly, the adjustment component in the remote sensing information acquisition device can adjust the orientation of the acquisition component according to the distance between the acquisition module and the aircraft body. Therefore, the shooting angle of the acquisition component can be flexibly adjusted according to data acquisition needs, facilitating the acquisition of more effective data and achieving data acquisition with different precision. In addition, since the first end of the connecting rod is slidably connected to the arm through a connecting seat, and the sliding of the second end of the connecting rod in the channel hole drives the adjustment shaft to rotate inside the carrier box, the structural complexity of the remote sensing information acquisition device can be reduced. This allows the position adjustment of the acquisition module and the orientation adjustment of the acquisition component to be achieved solely through the drive component, facilitating manufacturing and subsequent maintenance. Attached Figure Description

[0028] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments and descriptions of the invention are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0029] Figure 1 The principle block diagram of the remote sensing monitoring system for typical grassland degradation indicator species based on airborne multi-source spectral data described in the embodiments of the present invention;

[0030] Figure 2 A schematic diagram of the structure of the remote sensing information acquisition device described in the embodiment of the present invention;

[0031] Figure 3 An exploded view of the remote sensing information acquisition device described in the embodiments of the present invention;

[0032] Figure 4 A cross-sectional view of the driving component described in the embodiments of the present invention;

[0033] Figure 5 An exploded view of the acquisition module described in the embodiment of the present invention;

[0034] Figure 6 A schematic diagram of the structure of the carrier box described in the embodiment of the present invention;

[0035] Figure 7 This is a schematic diagram of the structure of the connecting rod and connecting seat as described in the embodiment of the present invention;

[0036] Figure 8 A cross-sectional view of the adjusting shaft as described in an embodiment of the present invention.

[0037] Explanation of reference numerals in the attached figures:

[0038] 11-Machine body; 12-Machine arm; 21-Drive box; 22-Drive rod; 23-Drive motor; 24-Drive gear; 25-Drive rack; 251-Limit baffle; 3-Carrier box; 31-Installation notch; 32-Installation hole; 33-Connecting column; 4-Adjusting shaft; 41-Channel hole; 411-Spiral guide groove; 42-Installation strip; 421-Bolt hole; 5-Connecting rod; 51-Guide protrusion; 52-Connecting shaft; 61-Connecting plate; 62-Limiting roller shaft; 7-Collection component; 71-Mounting base; 711-Installation bolt. Detailed Implementation

[0039] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0040] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation on this invention. Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, features defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0041] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0042] The invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0043] A remote sensing monitoring system based on airborne multi-source spectral data of typical grassland degradation indicator species is proposed to monitor and assess grassland health status, enabling timely detection and response to grassland degradation problems. Figure 1As shown, a remote sensing monitoring system for typical grassland degradation indicator species based on airborne multi-source spectral data includes: a remote sensing information acquisition device and a software system operation and control module. The remote sensing monitoring device includes: a rotary-wing UAV, a drive component, and two acquisition modules. The operation and control module includes: an image acquisition unit for acquiring at least three consecutive images from an acquired image sequence and marking the corresponding acquisition parameters; a semantic recognition neural network for processing the at least three input images and outputting the identification result of typical grassland degradation indicator species, wherein the semantic recognition neural network includes no more than a first preset number of convolutional layers; an auxiliary image acquisition unit for selecting corresponding low-altitude images and / or normally acquired images according to the acquisition parameters; and an auxiliary semantic recognition neural network for processing the input low-altitude images and / or normally acquired images, wherein the auxiliary semantic recognition neural network includes no less than a second preset number of convolutional layers. The system includes a set number of convolutional layers, where the second preset number is significantly greater than the first preset number. A feature acquisition unit compares the output of the last convolutional layer of the semantic recognition neural network with a preset threshold. If the output is less than the preset threshold, it acquires the output of the last convolutional layer of the auxiliary semantic recognition neural network, compares it with the preset threshold, identifies pixels greater than the preset threshold, expands this threshold to surrounding pixels, and acquires the convolutional features of the pixels and surrounding pixels from the previous convolutional layer. These convolutional features are then input into the penultimate convolutional layer of the semantic recognition neural network. A control unit determines whether a degradation indicator exists based on the output of the semantic recognition neural network. When a degradation indicator is determined to exist, the distance between the acquisition module and the machine is adjusted by a driving component, allowing the adjustment component to adjust the orientation of the acquisition component, thus achieving precise scanning and monitoring.

[0044] In this embodiment, the limited number of training images acquired from high altitudes, coupled with their limited resolution, results in low recognition accuracy using a convolutional neural network model. This embodiment addresses this issue by optimizing the model. Specifically, a semantic neural network model with a simple structure and low computational cost can be used as the primary method for identifying degradation indicators. Its simple structure, low computational cost, and short processing time make it suitable for processing on simple portable workstations.

[0045] To improve the accuracy of the semantic neural network model's judgment, in this embodiment, the judgment result can be derived by combining the judgment results of multiple images from different angles based on the relative angle of the UAV to a certain location on the ground while flying at high altitude. For example, the judgment can be made by combining the credibility of pixels after convolution processing of different images using the semantic neural network model. That is, the probability given by different images is fused by combining the matching results of pixels from different images, and the fused result is compared with a preset threshold to determine the accuracy of the recognition result.

[0046] On the other hand, due to the limited sample size of high-altitude images, the semantic neural network model is not sufficiently trained, thus affecting its accuracy. Therefore, in this embodiment, an auxiliary semantic recognition neural network is introduced. This auxiliary semantic recognition neural network has a more complex structure and higher computational load than the standard semantic neural network model. Its purpose is to utilize low-altitude images and normally acquired images that have a certain similarity to the images input to the semantic neural network model—that is, images acquired from multiple angles over land—to obtain image features of the indicator species. These extracted image features are then input into the semantic neural network model to further improve its accuracy in identification. For example, the acquisition parameters can be the acquisition height, coordinates, and the angle of the suspected indicator species calculated based on the coordinates. Correspondingly, the appropriate low-altitude image and / or normally acquired image can be determined based on the angle of the suspected indicator species.

[0047] Optionally, the probability of each element as an indicator can be determined by the output of the last convolutional layer of the semantic neural network model and the auxiliary semantic recognition neural network. Further, the output of the last convolutional layer of the auxiliary semantic recognition neural network is compared with the preset threshold to identify pixels with values ​​greater than the preset threshold, and this is then extended to surrounding pixels. This feature fully considers the interactivity between the convolutional processing results corresponding to different pixels, further improving the comprehensiveness of the features used to input the semantic neural network model, thereby improving the accuracy of the semantic neural network model's judgment and recognition. Inputting the convolutional features into the penultimate convolutional layer of the semantic recognition neural network does not affect the previous convolutional processing results of the semantic neural network model, while also facilitating its learning and integration of the input features, and avoiding bias caused by repeated convolutional processing of the input features.

[0048] Additionally, as an optional implementation of this embodiment, the remote sensing monitoring system for typical grassland degradation indicator species based on airborne multi-source spectral data may further include a multispectral processing module. Specifically, the multispectral processing module may include: an image preprocessing unit for smoothing multispectral images acquired by a multispectral sensor; a spectral classification unit for extracting spectral bands from the smoothed multispectral images, identifying the spectral band angles, calculating the spectral similarity between every two spectral bands based on the spectral band angles, and classifying the spectral bands; and a plant identification unit for querying the spectral band attributes of various spectral bands and identifying the grassland plants corresponding to each spectral band based on the spectral band attributes.

[0049] After the remote sensing information acquisition device acquires a spectral image, the image preprocessing unit performs a smoothing process to remove noise while maintaining the clarity of image edges. Correspondingly, after smoothing, differential transformations and envelope removal can be performed on the spectral image to make the waveforms in the spectral data clearer, facilitating the extraction of effective information in subsequent localization processes.

[0050] After preprocessing, the spectral classification unit classifies the spectral bands, which reduces the overall computational load of data processing, making it easier to select spectral bands that can effectively represent plant characteristics for identification and analysis in subsequent processing.

[0051] The plant identification unit can utilize existing SVM and RF models to identify grassland plants corresponding to various spectral bands. This allows for the determination of the presence of degraded indicator species in the current area through spectral images, further improving the monitoring and identification accuracy of the system. It should be noted that RF (Random Forest) is an ensemble classifier that generates multiple decision trees using randomly selected training sample subsets and variables, and then votes on the classification results of these trees to determine the final identification classification result. SVM (Support Vector Machine) is a supervised classifier that, by setting a multinomial kernel function, allows for the subjective setting of the exponent to make originally linearly inseparable data linearly separable, thereby solving the classification problem of nonlinear datasets and ultimately achieving the desired identification effect. The specific settings and operating principles of the above classifiers are well-known and are not part of the core invention content of this application, therefore, they will not be elaborated upon here.

[0052] Figure 2 This is a schematic diagram of a remote sensing information acquisition device. As shown in the figure, the device includes a rotary-wing UAV, a drive assembly, and two acquisition modules. The two acquisition modules are arranged mirror-imagely on both sides of the UAV's fuselage 11. The drive assembly is connected to each of the two acquisition modules at its two ends. During operation, the drive assembly adjusts the distance between the acquisition modules and the fuselage 11. When adjusting the distance, the drive assembly ensures that the two acquisition modules remain mirror-imagely distributed on both sides of the fuselage 11, thus preventing the UAV's center of gravity from being affected by the module position adjustments and ensuring good flight stability.

[0053] Optionally, the driving assembly in this embodiment may include a driving box 21 and two driving rods 22. The driving box 21 has driving holes at both ends, and the two driving rods 22 are located on opposite sides of the driving box 21. The end of the driving rod 22 closest to the driving box 21 is slidably disposed inside the driving hole, while the end of the driving rod 22 furthest from the driving box 21 is connected to the connecting post 33 at the top of the carrier box 3. During assembly, the operator can detachably install the driving box 21 onto the body 11 of the rotary-wing UAV using bolts, thereby using the sliding motion of the driving rods 22 to drive the acquisition module to move towards or away from the body 11.

[0054] like Figure 4 As shown, to drive the drive rod 22 and ensure that the acquisition modules on both sides of the body 11 obtain the same displacement stroke, the drive assembly may further include: a drive motor 23, a drive gear 24, and two drive racks 25. The drive motor 23 is mounted on the drive box 21, and the output shaft of the drive motor 23 is inserted into the drive box 21. The drive gear 24 and the two drive racks 25 are both mounted inside the drive box 21, and the drive gear 24 is connected to the output shaft of the drive motor 23. The two drive racks 25 are located on both sides of the drive gear 24, and either drive rack 25 meshes with the drive gear 24. The two drive racks 25 are respectively connected to the two drive rods 22. When the drive motor 23 starts, the drive gear 24 will start to rotate. Since the two drive racks 25 that mesh with the drive gear 24 are located on both sides of the drive gear 24, after the drive gear 24 starts to rotate, the two drive racks 25 will simultaneously drive the two drive rods 22 to move in the drive box 21 (or to the outside of the drive box 21) with the same stroke, thereby ensuring that the two acquisition modules are always mirror-distributed on both sides of the body 11.

[0055] In addition, to prevent the drive rod 22 from disengaging from the drive housing 21, a limiting baffle 251 can be provided at the end of the drive rack 25 away from the drive rod 22. When the drive rod 22 extends to its limit position outside the drive housing 21, the limiting baffle 251 will abut against the drive gear 24, thereby preventing the drive rod 22 from continuing to move outward.

[0056] like Figure 5 and Figure 6As shown, in this embodiment, each acquisition module includes a carrier box 3, an adjustment component, and an acquisition component 7. The carrier box 3 provides a mounting base for the adjustment component, the acquisition component 7 is used for data acquisition, and the adjustment component is used to adjust the orientation of the acquisition component 7 according to the distance between the acquisition module and the body 11, so as to acquire more effective data and achieve data acquisition with different precision. Specifically, the carrier box 3 is located between two adjacent arms 12 of the rotary-wing UAV. An installation notch 31 is provided on the bottom surface of the carrier box 3, and an installation hole 32 is provided on the side wall of the carrier box 3 near the arm 12. The adjustment component includes an adjustment shaft 4 and two connecting rods 5. The adjustment shaft 4 is rotatably disposed inside the carrier box 3, and a channel hole 41 communicating with the installation hole 32 is provided on the adjustment shaft 4. During assembly, the two connecting rods 5 should be located on both sides of the carrier box 3. The first end of the connecting rod 5 is slidably connected to the arm 12 of the rotary-wing UAV through the connecting seat. The second end of the connecting rod 5 is inserted into the carrier box 3 along the mounting hole 32 and slidably disposed inside the channel hole 41. The sliding of the second end of the connecting rod 5 inside the channel hole 41 can drive the adjusting shaft 4 to rotate inside the carrier box 3. A mounting strip 42 is also provided on the side wall of the adjusting shaft 4, extending out of the carrier box 3 along the mounting cut 31. The acquisition component 7 is detachably disposed on the mounting strip 42, and the acquisition component 7 includes an image sensor and a multispectral sensor.

[0057] When the drive assembly adjusts the distance between the acquisition module and the body 11, the carrier box 3 moves towards or away from the body 11 under the drive assembly's influence. Since the carrier box 3 is located between two adjacent arms 12 of the rotary-wing UAV, and these two adjacent arms 12 are typically arranged in a V-shape with their tips facing the body 11, when the carrier box 3 moves, the first end of the connecting rod 5 slides along the length of the arm 12, and the second end slides inside the channel hole 41. At this time, the adjusting shaft 4 rotates due to the sliding of the second end of the connecting rod 5 inside the channel hole 41, causing the mounting strip 42 to swing within the mounting cutout 31, thereby changing the orientation of the acquisition assembly.

[0058] like Figure 7 and Figure 8 As shown, to enable the connecting rod 5 to drive the adjusting shaft 4, a guide protrusion 51 can be provided on the side wall of the second end of the connecting rod 5, and a spiral guide groove 411 for accommodating the guide protrusion 51 is provided on the inner side wall of the channel hole 41. When the second end of the connecting rod 5 slides inside the channel hole 41, the guide protrusion 51 will move along the spiral guide groove 411. At this time, the cooperation between the spiral guide groove 411 and the guide protrusion 51 enables the adjusting shaft 4 to rotate with the linear movement of the connecting rod 5, thereby achieving the purpose of adjusting the orientation of the acquisition component.

[0059] Since the orientation of the acquisition components directly affects the data acquisition range and accuracy, operators can adjust the orientation of the acquisition components according to the actual situation to ensure that the data acquisition range and accuracy of the remote sensing information acquisition device meet the usage requirements. For example, when performing a coarse scan of the target area, operators can increase the distance between the acquisition module and the body 11 by using the drive component and swing the mounting strip 42 away from the body 11. This causes the acquisition components on both sides of the body 11 to be in a V-shaped, downward-opening position for data acquisition, thereby increasing the data acquisition range and facilitating the collection of more data samples during a single aerial survey, thus improving data acquisition efficiency. When performing a precise scan of the target area, operators can decrease the distance between the acquisition module and the body 11 by using the drive component and swing the mounting strip 42 closer to the body 11. At this time, the data acquisition area of ​​the acquisition components on both sides of the body 11 will converge directly below the body 11, thus enabling precise scanning of the area directly below the body 11, improving data acquisition accuracy, and consequently improving the accuracy of the degradation indicator identification results.

[0060] To achieve a detachable connection between the data acquisition component 7 and the mounting strip 42, this embodiment may also include a mounting base 71 on the data acquisition component 7, and mounting bolts 711 on the mounting base 71. Correspondingly, the mounting strip 42 may have multiple bolt holes 421 for accommodating the mounting bolts 711, and these bolt holes 421 should be arranged along the length of the mounting strip 42. During assembly of the data acquisition component 7 and the mounting strip 42, the operator can adjust the relative positions of the mounting base 71 and the mounting strip 42 according to the actual situation, and insert the mounting bolts 711 into the corresponding bolt holes 421 to ensure that the installation position of the data acquisition component 7 meets the actual usage requirements.

[0061] Furthermore, to achieve a sliding connection between the connecting rod 1 and the arm 12, the connecting seat in this embodiment may include two parallel connecting plates 61, and two rotatable limiting rollers 62 are provided between the two connecting plates 61. The first end of the connecting rod 5 is provided with a connecting shaft 52, and the two ends of the connecting shaft 52 are rotatably connected to the two connecting plates 61 respectively. Figure 7 As shown, the axes of both limiting rollers 62 should be perpendicular to the connecting plate 61, forming a receiving channel between the two limiting rollers 62 to accommodate the machine arm 12, and the axis of the connecting shaft 52 should be parallel to the axis of the limiting rollers 62. During assembly, the operator can constrain and limit the machine arm 12 using the two limiting rollers 62 to ensure that the connecting seat is reliably installed on the machine arm 12. When the connecting seat slides along the machine arm 12 due to the change in the position of the acquisition module, the rotation of the limiting rollers 62 can reduce the frictional resistance between the connecting seat and the machine arm 12, thereby improving the smoothness of the sliding of the connecting seat.

[0062] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A remote sensing monitoring system for typical grassland degradation indicator species based on airborne multi-source spectral data, characterized in that: Includes remote sensing information acquisition devices and computing control modules; The remote sensing information acquisition device includes: a rotary-wing UAV, a drive assembly, and two acquisition modules; The operation control module includes: An image acquisition unit is used to acquire at least three consecutive images from an acquired image sequence and mark the corresponding acquisition parameters; A semantic recognition neural network is used to process at least three types of input images respectively and output the recognition result of typical grassland degradation indicator species. The semantic recognition neural network includes no more than a first preset number of convolutional layers. An auxiliary image acquisition unit is used to select a corresponding low-altitude image and / or a normally acquired image according to the acquisition parameters. An auxiliary semantic recognition neural network is used to process input low-altitude images and / or normally acquired images. The auxiliary semantic recognition neural network includes no less than a second preset number of convolutional layers, where the second preset number is much larger than the first preset number. The feature introduction unit is used to compare the output of the last convolutional layer of the semantic recognition neural network with a preset threshold. When the output is less than the preset threshold, the unit obtains the output of the last convolutional layer of the auxiliary semantic recognition neural network, compares it with the preset threshold, determines the pixels that are greater than the preset threshold, expands it to the surrounding pixels, obtains the convolutional features of the pixels and the surrounding pixels from the previous convolutional layer, and inputs the convolutional features into the second-to-last convolutional layer of the semantic recognition neural network. The control unit is used to determine whether a degenerate indicator species exists based on the output of the semantic recognition neural network; when a degenerate indicator species is determined to exist, the distance between the acquisition module and the body is adjusted by the drive component, so that the adjustment component adjusts the orientation of the acquisition component to achieve accurate scanning and monitoring.

2. The remote sensing monitoring system for typical grassland degradation indicator species based on airborne multi-source spectral data according to claim 1, characterized in that: Two acquisition modules are arranged in a mirror image on both sides of the body (11) of the rotary-wing UAV. The drive assembly is used to adjust the distance between the acquisition modules and the body (11), and the two ends of the drive assembly are respectively connected to the two acquisition modules. Each acquisition module includes: a carrier box (3), an adjustment assembly, and an acquisition assembly (7). The carrier box (3) is located between two adjacent arms (12) of the rotary-wing UAV. An installation cutout (31) is provided on the bottom surface of the carrier box (3), and an installation hole (32) is provided on the side wall of the carrier box (3) near the arm (12). The adjustment assembly is used to adjust the orientation of the acquisition assembly (7) according to the distance between the acquisition module and the body (11). The adjustment assembly includes an adjustment shaft (4) and two connecting rods (5). The adjustment shaft (4) is rotatable. The adjustment shaft (4) is movably set inside the carrier box (3). A channel hole (41) communicating with the mounting hole (32) is provided on the adjustment shaft (4). The two connecting rods (5) are located on both sides of the carrier box (3). The first end of the connecting rod (5) is slidably connected to the arm (12) of the rotary-wing UAV through the connecting seat. The second end is slidably set inside the channel hole (41). The sliding of the second end of the connecting rod (5) inside the channel hole (41) can drive the adjustment shaft (4) to rotate inside the carrier box (3). An installation strip (42) extending out of the carrier box (3) along the installation cut (31) is also provided on the side wall of the adjustment shaft (4). The acquisition component (7) is detachably set on the installation strip (42). The acquisition component (7) includes an image sensor and a multispectral sensor.

3. The remote sensing monitoring system for typical grassland degradation indicator species based on airborne multi-source spectral data according to claim 2, characterized in that: The second end of the connecting rod (5) has a guide protrusion (51) on its side wall, and the inner side wall of the channel hole (41) has a spiral guide groove (411) for accommodating the guide protrusion (51).

4. The remote sensing monitoring system for typical grassland degradation indicator species based on airborne multi-source spectral data according to claim 2, characterized in that: The drive assembly includes a drive box (21) and two drive rods (22). The drive box (21) is detachably mounted on the body (11) of the rotary-wing UAV. Drive holes are provided at both ends of the drive box (21). The two drive rods (22) are located on both sides of the drive box (21). The end of the drive rod (22) closer to the drive box (21) is slidably mounted inside the drive hole, and the end of the drive rod (22) away from the drive box (21) is connected to the connecting post (33) at the top of the carrier box (3).

5. The remote sensing monitoring system for typical grassland degradation indicator species based on airborne multi-source spectral data according to claim 4, characterized in that: The drive assembly further includes: a drive motor (23), a drive gear (24), and two drive racks (25); the drive motor (23) is mounted on the drive box (21), and the output shaft of the drive motor (23) is inserted into the drive box (21); the drive gear (24) and the two drive racks (25) are both mounted inside the drive box (21), the drive gear (24) is connected to the output shaft of the drive motor (23), and the two drive racks (25) are located on both sides of the drive gear (24); any one of the drive racks (25) meshes with the drive gear (24), and the two drive racks (25) are connected to two drive rods (22) respectively.

6. The remote sensing monitoring system for typical grassland degradation indicator species based on airborne multi-source spectral data according to claim 5, characterized in that: The drive rack (25) is provided with a limiting baffle (251) at the end away from the drive rod (22).

7. The remote sensing monitoring system for typical grassland degradation indicator species based on airborne multi-source spectral data according to claim 2, characterized in that: The acquisition component (7) is provided with a mounting base (71) and a mounting bolt (711) is provided on the mounting base (71); the mounting strip (42) is provided with a plurality of bolt holes (421) for accommodating the mounting bolts (711), and the plurality of bolt holes (421) are arranged along the length direction of the mounting strip (42).

8. The remote sensing monitoring system for typical grassland degradation indicator species based on airborne multi-source spectral data according to claim 2, characterized in that: The connecting seat includes two parallel connecting plates (61), and two rotatable limiting rollers (62) are provided between the two connecting plates (61). The axes of the two limiting rollers (62) are perpendicular to the connecting plates (61), and a receiving channel for accommodating the machine arm (12) is formed between the two limiting rollers (62). The first end of the connecting rod (5) is provided with a connecting shaft (52), the axis of the connecting shaft (52) is parallel to the axis of the limiting rollers (62), and the two ends of the connecting shaft (52) are rotatably connected to the two connecting plates (61) respectively.

9. The remote sensing monitoring system for typical grassland degradation indicator species based on airborne multi-source spectral data according to claim 1, characterized in that: The remote sensing monitoring system for typical grassland degradation indicator species based on airborne multi-source spectral data also includes a multispectral processing module, which includes: The image preprocessing unit is used to smooth the multispectral images acquired by the multispectral sensor; The spectral segment classification unit is used to extract spectral segments from the smoothed multispectral image, identify the spectral segment angles of the spectral segments, and calculate the spectral segment similarity between every two spectral segments based on the spectral segment angles to classify the spectral segments. The plant identification unit is used to query the spectral attributes of various spectral bands and identify the grassland plants corresponding to each spectral band based on the spectral attributes.

Images