Method and device for precise restoration of garden vegetation based on unmanned aerial vehicle intelligent monitoring

Abstract

The present application relates to the garden vegetation precision recovery method and device based on unmanned aerial vehicle intelligent monitoring, including the following: step 110, obtain the multispectral image of target monitoring garden through unmanned aerial vehicle, multispectral image includes a plurality of NIR band images and a plurality of Red band images which are continuously shot at the same position with extremely short interval; Step 120, obtain NIR fusion image and Red fusion image by processing a plurality of NIR band images and a plurality of Red band images respectively through image fusion, then pretreat NIR fusion image and Red fusion image respectively; Step 130, construct NDVI model based on pretreated NIR fusion image and Red fusion image, and determine the vegetation recovery area of target detection garden according to NDVI model; Step 140, carry out vegetation recovery on vegetation recovery area. The present application carries out precision recovery on vegetation recovery area in target detection garden through unmanned aerial vehicle, improves the accuracy and efficiency of recovery, and can greatly overcome the influence of environment.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of image processing and vegetation restoration, and particularly relates to a method and device for precise restoration of garden vegetation based on intelligent monitoring of unmanned aerial vehicles. BACKGROUND

[0002] Traditional forestry is restricted by many environmental factors, insufficient technical support, and other factors. Some work is relatively extensive and inefficient. For example, traditional forestry vegetation resource investigation is labor-intensive, time-consuming, and costly, and is usually limited to sample plots, making it difficult to obtain large-scale forestry vegetation parameter spatial distribution information. For areas with high mountains, steep slopes, and poor transportation, traditional manual afforestation methods are labor-intensive, inefficient, and have certain operation risks. Existing unmanned aerial vehicle seeding vegetation restoration methods also have limitations such as low planting accuracy and low survival rate of seeded vegetation seeds in such operating environments.

[0003] With the continuous progress of unmanned aerial vehicle technology, the multi-spectral or hyperspectral cameras carried by the unmanned aerial vehicle can collect positioning information data and build ecological models, and through related software, the growth state of the forestry vegetation can be monitored and understood, and the intelligent monitoring of the unmanned aerial vehicle and the rapid diagnosis of the vegetation state can be realized. This provides new possibilities for solving the problems existing in the traditional operation mode, such as low efficiency, low accuracy, and high operation cost. By applying unmanned aerial vehicle technology, the efficiency, accuracy, and labor intensity of forestry work can be improved, and the informatization and intelligentization level of forestry can be improved, providing technical support for the sustainable development of forestry. However, when the unmanned aerial vehicle collects images, random noise may interfere with the pictures, and details cannot be guaranteed, which may affect the subsequent judgment of the garden vegetation area to be restored. SUMMARY

[0004] The purpose of the present application is to at least solve one of the problems of the prior art, and to provide a method and device for precise restoration of garden vegetation based on intelligent monitoring of unmanned aerial vehicles.

[0005] In order to achieve the above-mentioned purpose, the present application adopts the following technical solutions:

[0006] Specifically, a method for precise restoration of garden vegetation based on intelligent monitoring of unmanned aerial vehicles is proposed, which includes the following steps:

[0007] Step 110: acquiring a multi-spectral image of the target monitoring garden by the unmanned aerial vehicle, the multi-spectral image including a plurality of NIR band images and a plurality of Red band images taken continuously at the same position with extremely short intervals;

[0008] Step 120, obtain an NIR fusion image and a Red fusion image by respectively processing a plurality of NIR band images and a plurality of Red band images through image fusion, and then respectively pre-process the NIR fusion image and the Red fusion image;

[0009] Step 130, construct an NDVI model based on the pre-processed NIR fusion image and Red fusion image, and determine a vegetation to be recovered area of the target detection garden according to the NDVI model;

[0010] Step 140, perform vegetation recovery on the vegetation to be recovered area.

[0011] Further, specifically, the way of performing vegetation recovery on the vegetation to be recovered area is to spray planting capsules on the vegetation to be recovered area by a drone, wherein the planting capsule is in the shape of a bullet, and the capsule shell is integrally formed, and the conical head cavity and the column cavity are respectively used to fill planting growth required fertilizers, seeds and medicines.

[0012] Further, specifically, obtaining the NIR fusion image and the Red fusion image by respectively processing the plurality of NIR band images and the plurality of Red band images through image fusion includes:

[0013] For the NIR band image or the Red band image, if it has N images, and the images are sorted in the order of collection to obtain an image set {TP_1,..., TP_t,..., TP_N}, t is a variable, t∈[1, N];

[0014] First, the N images are fused by averaging the pixel values of the same pixel points to obtain a classical fusion image;

[0015] Then, in the N images, find the detail performance area of each image;

[0016] Then execute the following loop program,

[0017] Step 210, initialize the variable t=1, and load the classical fusion image;

[0018] Step 220, replace the detail performance area in TP_t with the corresponding pixel point position in the classical fusion image to update the classical fusion image;

[0019] Step 230, increase the value of t by 1, and then determine whether t is greater than N, if yes, go to step 240, if not, go to step 220;

[0020] Step 240, output the classical fusion image at this time and mark it as the NIR fusion image or the Red fusion image.

[0021] Further, specifically, in N images, find out the detail performance area of each image, including:

[0022] First, for any image TP_t, let photo_t(t,mt,nt) be the pixel value of the mth row and nth column pixel point in its image matrix photo_t(t);

[0023] Next, calculate the detail presentation coefficient coefficient(t,mt,nt) corresponding to photo_t(t,mt,nt), and the calculation method is,

[0024] ;

[0025] Wherein, , Indicates the average, maximum and minimum of the pixel values of the neighborhood pixel points of photo_t(t,mt,nt);

[0026] Finally, based on the detail presentation coefficient, find out the detail performance area of all images in N images, including:

[0027] For TP_1, traverse its image matrix photo_1(1), and find out the two pixel points with the top two detail presentation coefficients from it, denoted as Max_point_1 and Max_point_2, connect Max_point_1 and Max_point_2 to get a straight line line_1,2, then find out the pixel point Max_point_3 with the maximum detail presentation coefficient from the pixel points in photo_1(1) except the straight line line_1,2, then the area formed by Max_point_1, Max_point_2 and Max_point_3 as the vertex is the detail performance area of TP_1;

[0028] For TP_t, calculate the average of the pixel values of the pixel points in the detail performance area of photo_t-1(t-1) , traverse its image matrix photo_t(t), find out two random pixel points with detail presentation coefficient greater than Denoted as point_1 and point_2, connect point_1 and point_2 to get a straight line line_point_1,2, then find out the pixel point Max_point_4 with the maximum detail presentation coefficient from the pixel points in photo_t(t) except the straight line line_point_1,2, then the area formed by point_1, point_2 and Max_point_4 as the vertex is the detail performance area of TP_t.

[0029] Further, specifically, the preprocessing comprises:

[0030] The NIR fusion image and the Red fusion image are registered, the registration comprises geometric correction and radiation correction, and then pixel values are scaled to the range of [0, 1] through normalization processing.

[0031] Further, specifically, an NDVI model is constructed based on the preprocessed NIR fusion image and the Red fusion image, and a vegetation to-be-restored region of the target detection garden is determined according to the NDVI model.

[0032] The NDVI value is calculated through a traditional formula to construct the NDVI model, and a region with an NDVI value less than a preset threshold is determined as the vegetation to-be-restored region.

[0033] Further, the method further comprises constructing an NDVI model based on a pre-trained U-net network and determining a vegetation to-be-restored region of the target detection garden.

[0034] The application further provides a device for accurate restoration of garden vegetation based on intelligent monitoring of a UAV.

[0035] The data acquisition module is configured to acquire a multispectral image of the target detection garden through the UAV, and the multispectral image comprises a plurality of NIR band images and a plurality of Red band images which are continuously captured at the same position with extremely short intervals.

[0036] The image fusion module is configured to process the plurality of NIR band images and the plurality of Red band images through image fusion to obtain a NIR fusion image and a Red fusion image, and then preprocess the NIR fusion image and the Red fusion image.

[0037] The vegetation to-be-restored region determination module is configured to construct an NDVI model based on the preprocessed NIR fusion image and the Red fusion image, and determine a vegetation to-be-restored region of the target detection garden according to the NDVI model.

[0038] The vegetation restoration execution module is configured to perform vegetation restoration on the vegetation to-be-restored region.

[0039] The application has the following beneficial effects.

[0040] The present application proposes a method and device for precise restoration of garden vegetation based on unmanned aerial vehicle intelligent monitoring, which on the one hand uses unmanned aerial vehicle technology to collect multispectral images, then analyzes the images according to the collected images to construct an NDVI model, and determines the vegetation area to be restored in the target detection garden according to the NDVI model, and then restores the vegetation area to be restored by the unmanned aerial vehicle, so as to overcome the influence of environmental factors in the prior art; on the other hand, considering the problem of random noise that may exist when the unmanned aerial vehicle collects images, which may interfere with the pictures, and the details cannot be guaranteed, which may affect the subsequent judgment of the garden vegetation area to be restored, the above problems are overcome by collecting multiple image fusion, so as to ensure the accuracy of subsequent analysis. BRIEF DESCRIPTION OF DRAWINGS

[0041] The above and other features of the present disclosure will become more apparent from the following detailed description taken in conjunction with the accompanying drawings, in which like reference numerals denote like elements or similar elements throughout the accompanying drawings, and in which:

[0042] Figure 1 A flowchart of the method for precise restoration of garden vegetation based on unmanned aerial vehicle intelligent monitoring is shown;

[0043] Figure 2 A flowchart of the method for precise restoration of garden vegetation based on unmanned aerial vehicle intelligent monitoring is shown;

[0044] Figure 3 An example of unmanned aerial vehicle route selection in specific application is shown;

[0045] Figure 4 An example of the planting capsule involved in the present application is shown. DETAILED DESCRIPTION

[0046] The concept, specific structure and resulting technical effects of the present application will be described clearly and completely in conjunction with the embodiments and the drawings below, so as to fully understand the purpose, scheme and effects of the present application. It should be noted that the embodiments and features in the embodiments in the present application can be combined with each other without conflict. The same reference numerals in the drawings indicate the same or similar parts.

[0047] Embodiment 1, refer to Figure 1 and Figure 2 The present application proposes a method and device for precise restoration of garden vegetation based on unmanned aerial vehicle intelligent monitoring, which includes the following:

[0048] Step 110, acquiring a multi-spectral image of the target monitoring garden by the unmanned aerial vehicle, wherein the multi-spectral image comprises a plurality of NIR band images and a plurality of Red band images which are continuously captured at the same position with extremely short intervals;

[0049] Step 120, processing the plurality of NIR band images and the plurality of Red band images respectively by image fusion to obtain an NIR fusion image and a Red fusion image, and then pre-processing the NIR fusion image and the Red fusion image respectively;

[0050] Step 130, constructing an NDVI model based on the pre-processed NIR fusion image and Red fusion image, and determining a vegetation recovery area of the target monitoring garden according to the NDVI model;

[0051] Step 140, performing vegetation recovery on the vegetation recovery area.

[0052] In the embodiment 1, on the one hand, multi-spectral image acquisition is performed by using the unmanned aerial vehicle technology, and then image analysis is performed according to the acquired images to construct an NDVI model, and a vegetation recovery area of the target monitoring garden is determined according to the NDVI model, and then the vegetation recovery area is accurately recovered by the unmanned aerial vehicle, so as to overcome the influence of environmental factors in the prior art; on the other hand, considering that random noise may exist when the unmanned aerial vehicle acquires images, which may interfere with the pictures, and details cannot be guaranteed, thereby affecting the subsequent determination of the vegetation recovery area of the garden, the above problems are overcome by acquiring a plurality of images and fusing them, so as to ensure the accuracy of subsequent analysis.

[0053] In addition, when performing specific operations,

[0054] Construction preparation 1, before construction, on-site investigation needs to be performed to understand the topography, landform, vegetation condition and other information of the construction area, so as to provide basis for subsequent unmanned aerial vehicle flight path planning and data acquisition.

[0055] 2, before unmanned aerial vehicle aerial survey, first, the coordinate system, elevation system and control points and other related data of the area to be measured need to be determined, the flight scheme is formulated according to the data, and then the flight frequency is determined by using DJI ZhiTu software, the flight range is delimited, and the ground resolution, heading and lateral overlap rate, flight line angle and other parameters are clarified.

[0056] 3, meteorological data is consulted, and it is determined that the weather condition on the day of flight meets the condition that the visibility is greater than 3000m and the wind force is less than five levels.

[0057] 4, the unmanned aerial vehicle operator is well briefed, and the matters needing attention and operation points during operation are clarified.

[0058] Unmanned aerial vehicle equipment debugging

[0059] The data acquisition equipment such as hyperspectral camera is installed on the unmanned aerial vehicle, and the connection line of the equipment such as hyperspectral camera is connected with the interface of the unmanned aerial vehicle, the power supply of the unmanned aerial vehicle is connected, the mobile phone APP is opened, the flight parameters of the unmanned aerial vehicle and the parameters of the data acquisition equipment are adjusted through the mobile phone APP interface, and various equipment is adjusted to ensure that the performance is normal during flight.

[0060] Data acquisition and processing

[0061] 1, after arriving at the scene, judge the flight height of the unmanned aerial vehicle, carry out horizontal calibration and compass calibration, and prepare for take-off. Connect the remote controller of the unmanned aerial vehicle, control the take-off of the unmanned aerial vehicle through the DJI ZhiTu software to carry out the aerial survey task. The flight height is set to 100m, the flight speed is 13m / s, the heading overlap rate is 75%, and the lateral overlap rate is 75%. The unmanned aerial vehicle navigation and the whole process of the gimbal pitch angle are automatically controlled by the software parameters, and the aerial photography is carried out along the flight line in turn. Figure 3 The specific example of the route selection of the unmanned aerial vehicle is shown in the figure.

[0062] 2, a large amount of image data collected by the unmanned aerial vehicle can be processed and analyzed to obtain relevant information of forestry resources. The image processing technology is used to extract the characteristics of the image, such as vegetation index and vegetation coverage, to analyze the growth status of the forest and the distribution of vegetation types. According to the processing and analysis, a geographic information system (GIS) can be established to realize the combination of the data obtained by the unmanned aerial vehicle and the geographic spatial data, so that the distribution and change trend of the forestry resources can be more accurately analyzed.

[0063] As a preferred embodiment of the present application, specifically, the method for restoring vegetation in the vegetation restoration area is to spray planting capsules in the vegetation restoration area by the unmanned aerial vehicle, wherein the planting capsule is in the shape of a bullet, the capsule shell is integrally formed, and the conical head cavity and the column body cavity are respectively used to fill the fertilizer, seeds and medicines required for planting and growth.

[0064] Reference Figure 4 In the preferred embodiment, the termination capsule is launched by the unmanned aerial vehicle, which can overcome the problem that manual planting cannot be performed due to bad environment, and the specific process is as follows,

[0065] 1, a 3D modeling software is used to establish a 3D model and optimize it, and the model is sent to a printer for printing. The printing process usually adopts layering processing and stacking forming to complete the 3D entity printing. After printing, the model needs to be post-processed, including removing support structures, surface treatment and other operations, to obtain the final product.

[0066] 2. The planting capsule is in the shape of a bullet, and the capsule shell is integrally formed. The conical head cavity and the column cavity are used to fill the fertilizers required for planting and growth, seeds, and medicines, etc. When making the planting capsule, the fertilizers, seeds, rodent and insect repellent medicines, and water retention agents are sequentially filled into the capsule in order, and finally the end of the capsule is closed by hot pressing.

[0067] 3. The planting capsule is made of degradable material (PLU polylactic acid, a substance extracted from corn starch), with a thickness of 2 mm, a length of 5-7 cm, and a diameter of 2-3 cm. It has a certain thickness and hardness, and does not break after colliding with the ground, maintaining the integrity of the appearance and protecting the seeds from being eaten by birds, rodents and other animals. The ratio of the length of the head cavity to the length of the seed cavity is 1:2, and the ratio of the length of the head cavity to the length of the seed cavity is moderate, which helps the seed capsule to maintain vertical falling after being ejected.

[0068] High-pressure injection planting device assembly and debugging

[0069] 1. The high-pressure injection device is used to launch the planting capsule. The high-pressure injection device has a circular planting capsule storage bin at the top. The storage bin can store 50 planting capsules at a time. During construction, the capsules are filled into the bin with the conical head facing down. Then the sensing line of the high-pressure injection device is passed through the soft hose at the bottom of the unmanned aerial vehicle, and the sensing line is connected to the "BS" interface inside the aircraft. After the sensing line is connected, the sowing box is installed at the bottom of the unmanned aerial vehicle.

[0070] 2. The unmanned aerial vehicle positioning system is calibrated before flight. A new "flow calibration" module is created using 5G technology to debug the high-pressure injection device to ensure that the injection device can be injected by the unmanned aerial vehicle operation. Set the pressure intensity of the high-pressure injection device (0.1-0.3 MPa), the flight speed (10-13 m / s), etc., and then select the flow calibration to ensure timely feedback on the APP interface during the injection process.

[0071] Installation of intelligent pouring device

[0072] Using the model obtained by unmanned aerial vehicle monitoring, the coordinates to be restored are extracted by professional software and input into the unmanned aerial vehicle controller. After system calibration, the unmanned aerial vehicle is started and flies to the upper space to be operated. After descending to the specified height, the high-pressure injection device is started to launch the planting capsule at a high speed to penetrate the ground cover of weeds and bury the planting capsule in the ground 3-5 cm.

[0073] Unmanned aerial vehicle monitoring and maintenance

[0074] 1. After the construction of the area to be repaired is completed, the construction area is monitored regularly at a certain time, and if necessary, an unmanned aerial vehicle is used to match a small water tank to water the planted tree area, so as to promote the germination and growth of the tree species. Water enters the inside from the small hole in the side wall of the seed cavity, the water retaining agent in the cavity absorbs water and expands, providing water for the seed and generating pressure on the capsule shell, and the capsule shell is made of degradable biomass material, which degrades faster when it comes into contact with water. The biomass capsule coating can also provide part of the nutrients needed for seed growth during the degradation process, which is beneficial to the germination and hatching of the seed when it encounters water again.

[0075] 2. The plants in the area are monitored by using an unmanned aerial vehicle, and the monitoring includes the growth status of the vegetation and diseases and pests, which are used to guide the maintenance personnel to construct.

[0076] As a preferred embodiment of the present application, specifically, the NIR fusion image and the Red fusion image are obtained by processing a plurality of NIR band images and a plurality of Red band images respectively through image fusion, comprising:

[0077] For the NIR band image or the Red band image, if it has N images, and the images are sorted in the order of collection to obtain an image set {TP_1,..., TP_t,..., TP_N}, t is a variable, t∈[1, N];

[0078] First, the N images are fused by averaging the pixel values of the same pixel points to obtain a classical fusion image;

[0079] Next, in the N images, find the detail expression area of each image;

[0080] Then execute the following loop program,

[0081] Step 210, initialize the variable t=1, and load the classical fusion image;

[0082] Step 220, replace the corresponding pixel point position in the classical fusion image with the detail expression area in TP_t to update the classical fusion image;

[0083] Step 230, increase the value of t by 1, and then determine whether t is greater than N, if yes, go to step 240, if not, go to step 220;

[0084] Step 240, output the classical fusion image at this time and mark it as NIR fusion image or Red fusion image.

[0085] In the preferred embodiment, considering that the existing method is to perform pixel averaging when performing image fusion, which cannot effectively suppress random noise and may lose key details, the improved image fusion method is proposed to replace the corresponding content in the classical fusion image obtained by the existing method with the calculated detail performance area in each image, so that the classical fusion image is always updated in the direction of more details (here, the estimated area higher than the average of the previous image is found each time), and the final fusion image is more detailed and suppresses random noise to a certain extent.

[0086] As a preferred embodiment of the present application, specifically, in N images, the detail performance area of each image is found, including:

[0087] First, for any image TP_t, let photo_t(t,mt,nt) be the pixel value of the mth row and nth column pixel point in the image matrix photo_t(t) thereof;

[0088] Next, the detail performance coefficient coefficient(t,mt,nt) corresponding to photo_t(t,mt,nt) is calculated, and the calculation method is,

[0089] ;

[0090] Wherein, , photo_t(t,mt,nt) represents the average value, maximum value, and minimum value of the pixel values of the neighborhood pixel points of photo_t(t,mt,nt);

[0091] Finally, based on the detail performance coefficient, the detail performance area of all images in N images is found, specifically including:

[0092] For TP_1, traverse the image matrix photo_1(1) thereof, find two pixel points with the top two detail performance coefficients from photo_1(1), and mark them as Max_point_1 and Max_point_2, connect Max_point_1 and Max_point_2 to obtain a straight line line_1,2, then find the pixel point Max_point_3 with the maximum detail performance coefficient from the pixel points in photo_1(1) except the straight line line_1,2, and the area formed by Max_point_1, Max_point_2, and Max_point_3 as vertices is the detail performance area of TP_1;

[0093] For TP_t, the average value of the pixel values of the pixel points in the detail performance area of photo_t-1(t-1) is calculated Then, the image matrix photo_t(t) is traversed again to find two random pixel points point_1 and point_2 with a detail presentation coefficient greater than 0.5, a straight line line_point_1,2 is obtained by connecting point_1 and point_2, and then a pixel point Max_point_4 with the largest detail presentation coefficient is found from the pixel points in photo_t(t) except for the pixel points on the straight line line_point_1,2, and a region formed by point_1, point_2 and Max_point_4 as vertices is the detail presentation region of TP_t.

[0094] As a preferred embodiment of the present application, the preprocessing specifically includes:

[0095] The NIR fusion image and the Red fusion image are registered, the registration includes geometric correction and radiation correction, and then normalization processing is performed to scale the pixel value to the range of [0, 1].

[0096] As a preferred embodiment of the present application, the NDVI model is constructed based on the preprocessed NIR fusion image and Red fusion image, and the vegetation to be restored region of the target detection garden is determined according to the NDVI model, which specifically includes:

[0097] The NDVI value is calculated by a traditional formula to construct the NDVI model, and the region with an NDVI value less than a preset threshold is determined as the vegetation to be restored region.

[0098] As a preferred embodiment of the present application, the method further includes constructing the NDVI model based on the pre-trained U-net network and determining the vegetation to be restored region of the target detection garden.

[0099] In the present preferred embodiment, considering that the U-net network is suitable for image-to-image translation and can generate a higher quality NDVI model, the determination of the vegetation to be restored region of the target detection garden is based on the pre-trained U-net network, and the specific implementation process is as follows:

[0100] The model design of the U-net network is as follows:

[0101] Input: NIR image (single channel) and Red image (single channel), combined into 2-channel input.

[0102] Output: NDVI image (single channel, range [-1, 1] or [0, 1]).

[0103] Network structure:

[0104] Encoder-decoder architecture (such as U-Net):​

[0105] Encoder: Extract multi-scale features (reuse pre-trained model like ResNet).

[0106] Decoder: Restore image resolution, fuse high and low-level features.

[0107] Skip connection: Preserve detail information, improve edge and texture restoration ability.

[0108] Loss function:

[0109] Mean Squared Error (MSE) or Perceptual Loss, match real NDVI distribution.

[0110] Data processing,

[0111] Data alignment: Ensure strict registration of NIR and Red images (geometric correction + radiometric correction).

[0112] Normalization: Scale pixel values to the range [0, 1].

[0113] Data augmentation: Random cropping, rotation, flipping, increase training sample diversity.

[0114] Training strategy

[0115] Label generation: Calculate NDVI using traditional formula as supervision label:

[0116] ,

[0117] Optimizer: Adam (learning rate = 1e-4), batch size = 8.

[0118] Evaluation indicators: Peak Signal-to-Noise Ratio (PSNR), Structural Similarity (SSIM).

[0119] According to the NDVI model, determine the target detection garden vegetation area to be restored as follows:

[0120] Threshold segmentation:

[0121] According to the NDVI value, divide the level (example threshold needs domain knowledge adjustment):

[0122] Healthy vegetation: NDVI ≥ 0.3,

[0123] Mild degradation: 0.1 ≤ NDVI < 0.3,

[0124] Severe degradation: NDVI < 0.1 (area to be restored),

[0125] Output binary mask or classification map (such as using color to mark different levels).

[0126] Morphological post-processing:

[0127] Dilation / erosion operation eliminates isolated noise points, merges contiguous regions.

[0128] Dynamic threshold optimization (as a preferred solution):

[0129] Adaptive thresholding in conjunction with historical data or statistical distribution (such as Otsu algorithm).

[0130] The present application also proposes a device for precise restoration of garden vegetation based on unmanned aerial vehicle intelligent monitoring, comprising the following:

[0131] A data acquisition module is configured to acquire multispectral images of a target monitoring garden through an unmanned aerial vehicle, wherein the multispectral images include a plurality of NIR band images and a plurality of Red band images taken at the same position with extremely short intervals.

[0132] An image fusion module is configured to process the plurality of NIR band images and the plurality of Red band images respectively to obtain NIR fusion images and Red fusion images through image fusion, and then pre-process the NIR fusion images and the Red fusion images respectively.

[0133] A vegetation area to be restored determination module is configured to construct an NDVI model based on the pre-processed NIR fusion images and Red fusion images, and determine a vegetation area to be restored of the target monitoring garden according to the NDVI model.

[0134] A vegetation restoration execution module is configured to perform vegetation restoration on the vegetation area to be restored.

[0135] In addition, each functional module in each embodiment of the present application can be integrated in one processing module, or each module can exist physically independently, or two or more modules can be integrated in one module. The integrated module can be realized in the form of hardware or in the form of a software functional module.

[0136] The integrated module, if implemented in the form of a software function module and sold or used as an independent product, can be stored in a computer readable storage medium. Based on such understanding, all or part of the processes in the above-mentioned embodiment methods can also be completed by a computer program instructing related hardware, and the computer program can be stored in a computer readable storage medium. When the computer program is executed by a processor, the steps of the above-mentioned various method embodiments can be implemented. The computer program includes computer program code, which can be in the form of source code, object code, executable files or some intermediate forms. The computer readable medium can include any entity or system capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal and software distribution medium, etc.

[0137] Although the description of the present application has been quite detailed and particularly described with respect to several described embodiments, it is not intended to be limited to any of these details or embodiments or any particular embodiment, but should be considered to be a broad interpretation of the claims in view of the prior art, so as to effectively encompass the intended scope of the present application. In addition, the present application is described above in embodiments that the inventors can foresee, and the purpose is to provide a useful description, and those non-essential modifications to the present application that have not yet been foreseen can still represent equivalent modifications of the present application.

[0138] The above is only the preferred embodiment of the present application, and the present application is not limited to the above-mentioned embodiments, as long as the same means achieve the technical effect of the present application, it should belong to the protection scope of the present application. The technical solutions and / or embodiments within the protection scope of the present application can have various modifications and changes.

Claims

1. A method for precise restoration of garden vegetation based on unmanned aerial vehicle intelligent monitoring, characterized in that, The method comprises the following steps: Step 110, acquiring, by a UAV, a multi-spectral image of a target monitoring garden, the multi-spectral image comprising a plurality of NIR band images and a plurality of Red band images continuously captured at the same position; Step 120, processing the plurality of NIR band images and the plurality of Red band images respectively by image fusion to obtain an NIR fusion image and a Red fusion image, and then pre-processing the NIR fusion image and the Red fusion image respectively; Step 130, constructing an NDVI model based on the pre-processed NIR fusion image and Red fusion image, and determining a vegetation recovery area of the target monitoring garden according to the NDVI model; Step 140, performing vegetation recovery on the vegetation recovery area; Specifically, the plurality of NIR band images and the plurality of Red band images are processed respectively by image fusion to obtain the NIR fusion image and the Red fusion image, comprising: For the NIR band images or the Red band images, there are N images in total, and the images are sorted in the order of acquisition to obtain an image set {TP_1,..., TP_t,..., TP_N}, t is a variable, t ∈ [1, N]; First, the N images are fused by averaging the pixel values of the same pixel points to obtain a classical fusion image; Next, the detail expression area of each image is found in the N images; Then, the following loop program is executed, Step 210, initializing a variable t = 1, and loading the classical fusion image; Step 220, replacing the detail expression area in TP_t with the corresponding pixel point position in the classical fusion image to update the classical fusion image; Step 230, increasing the value of t by 1, and then determining whether t is greater than N, if yes, proceeding to step 240, if not, proceeding to step 220; Step 240, outputting the classical fusion image at this time and recording it as the NIR fusion image or the Red fusion image; Specifically, the detail expression area of each image is found in the N images, comprising: First, for any image TP_t, photo_t(t, mt, nt) is recorded as the pixel value of the pixel point in the mth row and the nth column of the image matrix photo_t(t) of the image; Next, the detail expression coefficient coefficient(t, mt, nt) corresponding to photo_t(t, mt, nt) is calculated, and the calculation method is ； wherein, , denote the average, maximum, minimum of the pixel values of the neighboring pixel points of photo_t(t, mt, nt). Finally, the detail expression area of all images in the N images is found based on the detail expression coefficient, comprising: For TP_1, traverse its image matrix photo_1(1), find two pixel points with the top two detail presentation coefficients from photo_1(1), mark them as Max_point_1 and Max_point_2, connect Max_point_1 and Max_point_2 to obtain a straight line line_1_2, then find a pixel point Max_point_3 with the largest detail presentation coefficient from the pixel points in photo_1(1) except those on line_1_2, and the area formed by Max_point_1, Max_point_2 and Max_point_3 is the detail presentation area of TP_1; For TP_t, the average value of the pixel values of the pixels in the detail presentation region of photo_t-1(t-1) is calculated Again, the image matrix photo_t(t) is traversed to find out the pixels with a detail presentation coefficient greater than Random two pixel points of the photo_t(t) are recorded as point_1 and point_2, a straight line line_point_1_2 is obtained by connecting point_1 and point_2, and then the pixel point Max_point_4 with the largest detail presentation coefficient is found from the pixel points other than the straight line line_point_1_2 in the photo_t(t). The area formed by point_1, point_2 and Max_point_4 as vertices is the detail presentation area of the TP_t.

2. The method of claim 1, wherein the method further comprises: Specifically, the method for restoring vegetation in the vegetation restoration area is to spray planting capsules in the vegetation restoration area by a UAV, wherein the planting capsules are in the shape of bullets, and the capsule shell is integrally formed, and the cavity of the conical head and the cavity of the column body are respectively used to fill the fertilizers, seeds and medicines required for planting and growth. 3.The method of claim 1, wherein, Specifically, the preprocessing includes: The NIR fusion image and the Red fusion image are registered, the registration includes geometric correction and radiation correction, and then normalized processing is performed to scale the pixel value to the range of [0, 1]. 4.The method of claim 1, wherein, Specifically, the NDVI model is constructed based on the preprocessed NIR fusion image and Red fusion image, and the vegetation restoration area of the target detection garden is determined according to the NDVI model, including: The NDVI value is calculated by a traditional formula to construct the NDVI model, and the area with an NDVI value less than a preset threshold is determined as the vegetation restoration area. 5.The method of claim 4, wherein, The method further includes constructing the NDVI model based on the pre-trained U-net network and determining the vegetation restoration area of the target detection garden.

6. The device for precise restoration of garden vegetation based on unmanned aerial vehicle intelligent monitoring, characterized in that, The device includes the following steps of the method of any one of claims 1-5: The data acquisition module is configured to acquire, by a UAV, multispectral images of a target detection garden, the multispectral images including a plurality of NIR band images and a plurality of Red band images continuously captured at the same position; The image fusion module is configured to process the plurality of NIR band images and the plurality of Red band images respectively by image fusion to obtain a NIR fusion image and a Red fusion image, and then preprocess the NIR fusion image and the Red fusion image respectively; The vegetation restoration area determination module is configured to construct an NDVI model based on the preprocessed NIR fusion image and Red fusion image, and determine the vegetation restoration area of the target detection garden according to the NDVI model; The vegetation restoration execution module is configured to restore vegetation in the vegetation restoration area. The vegetation restoration execution module is configured to restore vegetation in the vegetation restoration area.

Images