Fertilization method and system based on regional land parcel image

Abstract

The invention discloses a fertilization method and system based on a regional land parcel image, and belongs to the technical field of agricultural fertilization, and the method specifically comprises the steps: collecting a target regional land parcel image through an unmanned aerial vehicle, and after radiation correction and geometric correction, extracting fertilization information containing soil color, vegetation coverage, vegetation index and landform characteristics, integrating to form a fertilization feature data set; according to the fertilization feature data set, dividing regions through a decision tree classification algorithm, identifying and visually displaying by means of a geographic information system; collecting crop variety, planting density and regional division result data, making a fertilization scheme of each region in combination with a fertilization model, and applying fertilization by using automatic equipment; images are collected again after fertilization, the crop growth response evaluation effect is monitored, and the fertilization scheme is adjusted as required, so that the agricultural production benefits are effectively improved, the fertilizer waste is reduced, and the scientificity and rationality of fertilization are optimized.

Description

Technical Field

[0001] The present invention belongs to the technical field of agricultural fertilization, and in particular relates to a fertilization method and system based on regional land plot images. Background Art

[0002] In the agricultural production process, rational fertilization is a key link in improving crop yield and quality, reducing fertilizer waste and environmental pollution. Traditional fertilization methods are often based on experience or general soil test data, and lack accurate consideration of local regional differences within the plot. For example, in large areas of farmland, soil fertility and crop growth conditions may vary significantly in different areas, but existing fertilization methods are difficult to achieve precise zoning fertilization, which can easily lead to excessive fertilizer in some areas and insufficient fertilizer in others, affecting the overall agricultural production efficiency and potentially causing waste of fertilizer resources and pollution of soil and water bodies.

[0003] For example, a Chinese patent with authorization announcement number CN117035471B discloses a fertilization decision-making method and system based on multiple vegetation index images, including: acquiring multiple vegetation index images of crops based on remote sensing sensors carried by drones, and fusing them to obtain an element distribution map of each element in the farmland based on the content of multiple elements at each of multiple locations, thereby obtaining a multi-channel image; setting a set of trainable parameters for each of the preset elements, and using multi-channel images to train the decision model and multiple sets of trainable parameters; determining the corresponding trainable parameters based on the target element, and using the decision model to determine the distribution of the target element based on the trainable parameters and the multi-channel image of the target farmland. This technical solution integrates soil fertility and vegetation index to improve decision-making accuracy, and introduces multiple trainable parameters into the decoder input, which can be flexibly adjusted to calculate different nutrient element requirements.

[0004] The above existing technologies all have the following problems: there are limitations in obtaining element content; decision-making flexibility is limited; and hardware and technology costs are high. Summary of the Invention

[0005] In response to the shortcomings of the existing technology, the present invention proposes a fertilization method and system based on regional plot images. Unmanned aerial vehicles are used to collect plot images of target areas. After radiation correction and geometric correction, fertilization information including soil color, vegetation coverage, vegetation index and topographic features is extracted and integrated to form a fertilization feature data set; based on the fertilization feature data set, regions are divided through a decision tree classification algorithm, marked and visualized with the help of a geographic information system; crop variety, planting density and regional division result data are collected, and fertilization plans for each region are formulated in combination with a fertilization model, and fertilization is implemented using automated equipment; after fertilization, images are collected again to monitor crop growth response and evaluate the effect, and the fertilization plan is adjusted as needed, effectively improving agricultural production efficiency, reducing fertilizer waste, and optimizing the scientificity and rationality of fertilization.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] The fertilization method based on regional plot images includes:

[0008] Step S1: Use a UAV to collect images of the target area and perform radiation correction and geometric correction;

[0009] Step S2: extracting fertilization feature information based on the radiation-corrected target area plot image and integrating it to form a fertilization feature dataset, wherein the fertilization feature information includes soil color features, vegetation coverage features, vegetation index features, and topographic features;

[0010] Step S3: Based on the fertilization feature dataset, a decision tree classification algorithm is used to partition the land parcels and mark the divided areas. At the same time, a geographic information system is used to visualize the marked area division results.

[0011] Step S4: Collect crop variety information, planting density information and regional division result data of the plot, and formulate a fertilization plan for each area based on the pre-established fertilization model;

[0012] Step S5: Fertilizing the regional plots according to the formulated fertilization plan;

[0013] Step S6: After fertilization, use image acquisition equipment to capture images of the regional plots, monitor the growth changes of crops, evaluate the fertilization effect based on the crop growth response, and adjust the fertilization plan based on the evaluation effect.

[0014] Specifically, the specific steps of step S2 include:

[0015] S2.1: Use the UAV-mounted LiDAR to obtain topographic data of the target area, and simultaneously obtain a calibrated image and multispectral image of the target area;

[0016] S2.2: Obtain a calibrated multispectral image of the target area and use the pixel dichotomy model method to combine the reflectance information of the near-infrared and red bands in the multispectral image to calculate the vegetation coverage characteristics;

[0017] S2.3: Generate a digital elevation model based on the acquired terrain data, extract topographic features from the digital elevation model data, and analyze the impact of terrain on soil moisture and nutrient distribution;

[0018] S2.4: Based on the corrected target area plot image, soil color features are extracted according to the spectral reflectance differences of different soil colors and the band reflectance threshold.

[0019] S2.5: Extract corresponding vegetation index features based on the requirements of different fertilization models for vegetation index;

[0020] S2.6: Integrate the extracted soil color features, vegetation cover features, vegetation index features, and topography features to form a fertilization feature dataset.

[0021] Specifically, the specific steps of step S1 include:

[0022] S1.1: Obtain the boundary coordinates of the target area through the geographic information system. Based on the determined target area, use the route planning software Pix4Dcapture to plan the flight route. At the same time, set the image acquisition time interval and shooting mode according to the camera type and shooting purpose.

[0023] S1.2: Start the UAV and fly it over the target area according to the planned flight route. During the flight, the onboard image acquisition device collects images of the target area according to the set time interval and the shooting mode triggered by the flight position. At the same time, the collected images of the target area are transmitted to the ground control station via a wireless communication link.

[0024] S1.3: Obtain camera radiometric calibration data from the manufacturer, and use a radiometric correction formula to perform preliminary radiometric value correction on the target area plot image to obtain a radiometric brightness image of the target area plot. The camera radiometric calibration data includes the camera's gain value and offset value.

[0025] Specifically, the specific steps of step S1 also include:

[0026] S1.4: Use the environmental sensor onboard the drone to obtain atmospheric parameter data, and correct the radiation brightness image of the target area based on the obtained atmospheric parameter data to obtain a radiation-corrected image of the target area. The correction formula is: ,in, Indicated by wavelength The reflectivity of the ground objects under the sun, d represents the distance between the sun and the earth, Indicated by wavelength The solar irradiance under represents the cosine of the solar zenith angle, represents the solar zenith angle, Indicates the target area plot image at wavelength The radiance value after down-correction;

[0027] S1.5: Collect the coordinate information and georeference data of the ground control points and calculate the coordinates of each pixel in the radiometrically corrected target area plot image. Perform coordinate transformation to obtain the corrected coordinate position , and according to the corrected coordinate position , obtain the target area plot image after coordinate correction;

[0028] S1.6: Use a resampling method to determine the pixel grayscale value corresponding to the corrected coordinate position, and obtain a target area plot image after radiation correction and geometric correction.

[0029] Specifically, the specific steps of S1.6 include:

[0030] S1.61: Obtain a radiation-corrected target area plot image, corrected coordinate position information, and a coordinate-corrected target area plot image;

[0031] S1.62: Traverse the coordinate positions of the target area plot image after coordinate correction For each pixel coordinate, determine the corresponding four neighboring pixels in the target area block image coordinates after radiation correction 、 、 、 , and the corresponding grayscale values ​​of the four pixels are 、 、 、 ;

[0032] S1.63: The coordinate position of the target area plot image after coordinate correction Plot each pixel coordinate in the image with the radiometrically corrected target area Calculate the difference of the corresponding elements and get The weights of the pixels on the horizontal and vertical directions respectively and ;

[0033] S1.64: Calculate the grayscale value of the pixels in the target area plot image after coordinate correction ;

[0034] S1.65: After the grayscale values ​​of all pixels in the coordinate-corrected target area plot image are determined, the determined pixels are combined into a complete image according to the correct spatial position relationship and saved as an image file to obtain the target area plot image after radiation correction and geometric correction.

[0035] Specifically, the specific steps of step S3 include:

[0036] S3.1: Extract feature vectors from the fertilization feature dataset, where each feature vector contains soil color feature, vegetation cover feature, vegetation index feature, and topographic feature data;

[0037] S3.2: Load the pre-trained decision tree model, input the extracted feature vector into the pre-trained decision tree model, and determine the region category to which the pixel corresponding to the extracted feature vector belongs based on the output of the pre-trained decision tree model;

[0038] S3.3: Assign a unique identification number to each divided area and add attribute labels to the areas based on their characteristics;

[0039] S3.4: Import the division result data with the region identifier into a geographic information system, create a new layer in the geographic information system to represent the region division, and associate the division result data with the geographic coordinates;

[0040] S3.5: Set different colors and patterns for different regions based on region identification or attribute labels for visualization, and output the visualization results.

[0041] Specifically, the fertilization plan in step S4 includes fertilizer types, fertilizer amounts, fertilization times, and fertilization methods for different areas.

[0042] The fertilization system based on regional plot images includes: image processing module, feature extraction module, regional division module, fertilization plan formulation module, fertilization execution module, and fertilization effect monitoring module;

[0043] The image processing module is used to obtain the target area plot image and perform correction processing on the target area plot image;

[0044] The feature extraction module is used to extract fertilization feature information from the corrected target area plot image and integrate it to form a fertilization feature data set;

[0045] The region division module is used to partition the plots using a decision tree classification algorithm based on the fertilization feature dataset and to visualize the partition results;

[0046] The fertilization plan formulation module is used to comprehensively consider multiple factors such as crop variety, planting density and regional division results of the plot, combine with the pre-built fertilization model, and generate a fertilization plan for each divided area;

[0047] The fertilization execution module is used to control the automated fertilization equipment to fertilize the corresponding area according to the established fertilization plan;

[0048] The fertilization effect monitoring module is used to monitor the growth of crops in the plot after fertilization, evaluate the fertilization effect based on the monitoring results, and adjust the fertilization plan.

[0049] Specifically, the region division module includes: a decision tree classification unit and a region identification unit;

[0050] The decision tree classification unit is used to use the fertilization feature data set as input data and divide the land into regions according to pre-set rules based on the decision tree classification algorithm;

[0051] The area identification unit is used to assign unique identification information to each divided area.

[0052] Specifically, the fertilization plan formulation module includes: a data collection unit, a fertilization model calling unit, and a plan generation unit;

[0053] The data collection unit is used to collect crop variety information, planting density information and regional division result data of the plot;

[0054] The fertilization model calling unit is used to load a pre-established fertilization model;

[0055] The scheme generating unit is used to determine the type, amount, time and method of fertilizer for each area according to the collected data and the preloaded fertilization model, and generate a fertilization scheme based on the situation of each area.

[0056] Compared with the prior art, the present invention has the following beneficial effects:

[0057] 1. The present invention proposes a fertilization system based on regional plot images, and optimizes and improves the architecture, operation steps and processes. The system has the advantages of simple process, low investment and operation costs, and low production work costs.

[0058] 2. This invention proposes a fertilization method based on regional plot images. By using drones to collect and calibrate plot images, the method extracts fertilization feature information, and uses a decision tree classification algorithm and a geographic information system to achieve precise zoning and visualization of plots, thereby improving the targetedness and efficiency of fertilization. Secondly, a personalized fertilization plan is formulated based on crop variety, planting density, and fertilization model. Automated fertilization equipment is used for precise fertilization, while crop growth changes are monitored to evaluate fertilization effectiveness, thus achieving intelligent management and optimized adjustment of the fertilization process.

[0059] 3. The present invention proposes a fertilization method based on regional plot images. This method uses technical means such as drone image acquisition and geographic information systems to achieve accurate extraction of plot fertilization characteristics and effective plot zoning. Combined with automated fertilization equipment and growth monitoring, it realizes the intelligent and refined fertilization process, improves fertilization effects, and can make timely adjustments based on crop growth responses, thereby optimizing fertilization strategies. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 Schematic diagram of the fertilization method based on regional land image of the present invention;

[0061] Figure 2 This is a flow chart of the principle of the fertilization method based on regional land image of the present invention;

[0062] Figure 3 This is a flow chart of radiation correction and geometric correction for the fertilization method based on regional land image of the present invention;

[0063] Figure 4 This is an architecture diagram of the fertilization system based on regional land plot images of the present invention. DETAILED DESCRIPTION

[0064] Example 1:

[0065] See also Figure 1-Figure 3 The present invention provides an embodiment of a fertilization method based on a regional land image, comprising the following steps:

[0066] Step S1: Use a UAV to collect images of the target area and perform radiation correction and geometric correction;

[0067] Among them, when collecting images of target area plots, a drone with ultra-high-definition image acquisition capabilities and intelligent adaptive functions is used. The drone is equipped with a high-resolution multispectral camera and a high-sensitivity infrared thermal imager. When collecting images of target area plots, it can automatically adjust the camera parameters according to different lighting conditions, atmospheric environment and crop growth stage to ensure that the acquired images are optimal in terms of spectral resolution, spatial resolution and image clarity.

[0068] At the same time, before the flight, the drone should be fully inspected, including the battery power, camera lens cleanliness, and flight control system stability. The flight parameters of the drone should also be set according to the size, shape, and geographical location of the target area. For example, the flight altitude is generally determined by the required image resolution and plot area. For smaller plots, the flight altitude can be set at around 50-100 meters to obtain centimeter-level resolution images; for large plots, the flight altitude can be appropriately increased to 100-300 meters, but it is necessary to ensure that the image can still meet the analysis requirements; at the same time, the flight speed and route planning parameters should be set. Among them, when planning the route, it is necessary to ensure that the entire target area is covered and there is a certain overlap rate. Usually, the heading overlap rate is 70-80%, and the lateral overlap rate is 65-75%.

[0069] Step S2: extracting fertilization feature information based on the radiation-corrected target area plot image and integrating it to form a fertilization feature dataset, wherein the fertilization feature information includes soil color features, vegetation coverage features, vegetation index features, and topographic features;

[0070] Among them, each pixel point in the obtained fertilization feature data set has a corresponding set of feature values, such as each pixel point has corresponding feature data such as soil color type, vegetation coverage, NDVI value, slope value, etc.

[0071] Step S3: Based on the fertilization feature dataset, a decision tree classification algorithm is used to partition the land parcels and mark the divided areas. At the same time, a geographic information system is used to visualize the marked area division results.

[0072] Step S4: Collect crop variety information, planting density information, and regional division result data of the plot, and formulate a fertilization plan for each area based on the pre-established fertilization model. The fertilization plan includes the type of fertilizer, amount of fertilizer, fertilization time, and fertilization method for different areas;

[0073] Furthermore, the specific steps of step S4 include:

[0074] (1) Collect crop variety information, planting density information and regional division results of the plot to ensure the accuracy and completeness of the data;

[0075] Among them, the methods for obtaining crop variety information include field surveys, agricultural archive inquiries or exchanges with farmers; multiple representative areas are selected within the plot, and the number of crop plants or holes per unit area is counted using sampling measurement methods to determine the planting density; the crop variety in the present invention refers to rice.

[0076] (2) Loading a pre-trained fertilization model, using the pre-trained fertilization model, calculating the amount of fertilizer to be applied based on the characteristic data of each region, determining the fertilization time and method, and forming a fertilization plan for each region, wherein the amount of fertilizer applied is equal to the target rice yield multiplied by the nutrient requirement per unit yield, and then subtracting the soil fertilizer supply from the product, and finally dividing the difference by the fertilizer utilization rate. In the present invention, the fertilization model adopts a crop growth simulation model, which is the existing technical content in this field and is not the inventive solution of this application, and will not be described in detail here.

[0077] Step S5: Fertilizing the regional plots according to the established fertilization plan, wherein the fertilization is performed on the regional plots by using automated fertilization equipment;

[0078] Step S6: After fertilization, use image acquisition equipment to capture images of the regional plots, monitor the growth changes of crops, evaluate the fertilization effect based on the crop growth response, and adjust the fertilization plan based on the evaluation effect.

[0079] Furthermore, the specific steps of step S6 include:

[0080] (1) Prepare image acquisition equipment and set parameters, collect images of the area after fertilization according to the preset route, and store the image data;

[0081] (2) Calculate the Normalized Difference Vegetation Index (NDVI) and the Ratio Vegetation Index (RVI) from the collected image data;

[0082] (3) Using image analysis technology to extract crop morphological characteristics, such as plant height, crown width, and leaf area index. Image analysis technology is a prior art in this field and is not an inventive solution of this application, so it will not be described in detail here;

[0083] (4) Establish an evaluation index system and evaluate the fertilization effect through fuzzy comprehensive evaluation method;

[0084] (5) Based on the results of the fertilization effect evaluation, analyze the reasons for the poor fertilization effect. If insufficient fertilizer application in a certain area leads to slow crop growth, which is manifested as low NDVI values ​​and short plant heights, consider increasing the amount of fertilizer applied in that area. If the fertilizer ratio is unreasonable, such as excessive nitrogen fertilizer causing excessive crop growth and affecting yield and quality, it is necessary to adjust the ratio of nitrogen, phosphorus, potassium and other elements in the fertilizer. Develop corresponding adjustment strategies for different reasons, such as adjusting the amount of fertilizer, changing the fertilization time, changing the type of fertilizer, or optimizing the fertilization method.

[0085] (6) Apply the adjusted fertilization plan to the corresponding area and record the content, time, basis, and other information of the adjustment. For example, after increasing the application amount of a specific fertilizer in a certain area, record the type, quantity, application date of the added fertilizer, and the evaluation of the previous fertilization effect in the area, so as to track and summarize the experience and continuously optimize the fertilization plan.

[0086] For example, in a rice farmland area of ​​A mu, first, at different growth stages such as the first week, second week, and third week after rice sowing, a drone equipped with a multispectral camera is used to collect images of the area; then, the collected images are preprocessed, and the images are cropped using image editing software to remove invalid areas at the edges of the images due to shooting angles and other reasons, and ground control points are used for geometric correction to make the image accurately match the coordinates of the actual plot, and then radiation correction is performed according to the atmospheric correction model; then, feature extraction is performed, and the NDVI vegetation index is calculated by analyzing different bands of the multispectral image, and the soil color features are extracted using color space conversion, and the topographic features such as the slope and slope direction of the plot are obtained through digital elevation model data; then, a clustering algorithm is used to divide the plot into regions, and the entire farmland is divided into three fertilization areas, such as an area with high fertility and vigorous crop growth, an area with medium fertility, and an area with high fertility. Areas with normal crop growth and areas with low fertility and weak crop growth are marked as Area 1, Area 2, and Area 3 respectively; for different areas, fertilization plans are formulated based on the needs of the cultivated rice for nutrients such as nitrogen, phosphorus, and potassium in the current growth stage, and a fertilization model established based on many years of local agricultural experimental data. For example, Area 1 adopts a fertilization plan with a small amount of nitrogen fertilizer supplemented and an appropriate amount of phosphorus fertilizer and potassium fertilizer, and the fertilization method is strip application, and the fertilization time is selected after 3 days; Area 2 determines the corresponding fertilizer type and amount according to its soil fertility and crop growth conditions; finally, use intelligent fertilizer agricultural machinery to carry out fertilization operations according to the fertilization plan. The sensors on the fertilizer agricultural machinery monitor the fertilization situation in real time. After fertilization, images of the plots are collected regularly to monitor changes in crop growth. If it is found that the crop growth in Area 3 does not meet the expected results, the fertilization plan is adjusted in time, the amount of nitrogen fertilizer applied is increased, and the fertilization method is changed to hole application, and fertilization is carried out again to improve crop yield and quality.

[0087] In summary, by utilizing regional plot images, the present invention can accurately analyze the characteristic differences between different areas within a plot, thereby formulating personalized fertilization plans, effectively improving the accuracy and scientificity of fertilization, reducing fertilizer waste, improving agricultural production efficiency and reducing environmental pollution risks.

[0088] The specific steps of step S1 include:

[0089] S1.1: Obtain the boundary coordinates of the target area through a geographic information system. Based on the determined target area, use the route planning software Pix4Dcapture to plan the flight route. Simultaneously, set the image acquisition time interval and shooting mode based on the camera type and shooting purpose. The route planning software Pix4Dcapture is prior art in this field and does not constitute an inventive solution of this application, so its detailed description is omitted here.

[0090] S1.2: Start the UAV and fly it over the target area according to the planned flight route. During the flight, the onboard image acquisition device collects images of the target area according to the set time interval and the shooting mode triggered by the flight position. At the same time, the collected images of the target area are transmitted to the ground control station via a wireless communication link.

[0091] S1.3: Obtain camera radiometric calibration data from the manufacturer and use the radiometric correction formula Perform preliminary radiation value correction on the target area plot image to obtain the target area plot radiation brightness image, the camera radiation calibration data includes the camera gain value and offset value, wherein, Indicates the target area plot image at wavelength The corrected radiance value is Indicated by wavelength The gain value of the camera is Indicates the target area plot image at wavelength The digital quantization value below, Indicated by wavelength The offset value of the lower camera, Indicated by wavelength Atmospheric correction coefficient under Indicated by wavelength Sensor correction coefficient under ;

[0092] S1.4: Use the environmental sensor onboard the drone to obtain atmospheric parameter data, and correct the radiation brightness image of the target area based on the obtained atmospheric parameter data to obtain a radiation-corrected image of the target area. The correction formula is: ,in, Indicated by wavelength The reflectivity of the ground objects under the sun, d represents the distance between the sun and the earth, Indicated by wavelength The solar irradiance under represents the cosine of the solar zenith angle, represents the solar zenith angle, Indicates the target area plot image at wavelength The radiance value after down-correction;

[0093] S1.5: Collect coordinate information and georeferenced data for ground control points and use the formula The coordinates of each pixel in the radiometrically corrected target area plot image Perform coordinate transformation to obtain the corrected coordinate position , and according to the corrected coordinate position , obtain the target area plot image after coordinate correction, where, and represents the geometric model coefficient;

[0094] S1.6: Use a resampling method to determine the pixel grayscale value corresponding to the corrected coordinate position, and obtain a target area plot image after radiation correction and geometric correction.

[0095] The specific steps of S1.6 include:

[0096] S1.61: Obtain a radiation-corrected target area plot image, corrected coordinate position information, and a coordinate-corrected target area plot image;

[0097] S1.62: Traverse the coordinate positions of the target area plot image after coordinate correction For each pixel coordinate, determine the corresponding four neighboring pixels in the target area block image coordinates after radiation correction 、 、 、 , and the corresponding grayscale values ​​of the four pixels are 、 、 、 ;

[0098] S1.63: The coordinate position of the target area plot image after coordinate correction Plot each pixel coordinate in the image with the radiometrically corrected target area Calculate the difference of the corresponding elements and get The weights of the pixels on the horizontal and vertical directions respectively and , the specific formula is: ;

[0099] S1.64: Use formula Calculate the grayscale value of the pixels in the target area plot image after coordinate correction ;

[0100] S1.65: After the grayscale values ​​of all pixels in the coordinate-corrected target area plot image are determined, the determined pixels are combined into a complete image according to the correct spatial position relationship and saved as an image file to obtain the target area plot image after radiation correction and geometric correction.

[0101] Furthermore, the specific steps of S1.65 include:

[0102] (1) Image initialization stage:

[0103] Determine the image size according to the coordinate range after geometric correction, create a blank image data structure of the corresponding color mode, and prepare for subsequent pixel filling;

[0104] (2) Pixel assignment and construction stage:

[0105] Traverse the coordinates and grayscale values ​​of all corrected pixels, fill them into the blank image data structure in order according to the correct spatial position, and gradually build the complete image data content so that the image meets the requirements after correction in terms of spatial layout and pixel value;

[0106] (3) Image saving stage:

[0107] Select the image file format and save the constructed image data as an image file with the help of image processing library tools to complete the entire process and obtain a corrected target area plot image that can be used for subsequent analysis and use.

[0108] The specific steps of step S2 include:

[0109] S2.1: Use the UAV-mounted LiDAR to obtain topographic data of the target area, and simultaneously obtain a calibrated image and multispectral image of the target area;

[0110] S2.2: Obtain a calibrated multispectral image of the target area and use the pixel dichotomy model method to combine the reflectance information of the near-infrared and red bands in the multispectral image to calculate the vegetation coverage characteristics;

[0111] Among them, vegetation coverage reflects the degree of vegetation coverage on the ground, which is closely related to the soil's water and fertilizer retention capacity and crop growth conditions. By accurately calculating vegetation coverage, we can better understand the ecological conditions of the land.

[0112] Furthermore, the specific steps of S2.2 include:

[0113] (1) Obtaining the corrected multispectral image of the target area;

[0114] (2) According to the calculation formula of Normalized Difference Vegetation Index NDVI , using the near-infrared band reflectance data NIR and the red band reflectance data R in the multispectral image, calculate each pixel in the image to obtain the NDVI value of the entire image. Among them, the NDVI image is obtained by performing band calculations on the multispectral data of the image. In the NDVI image, the NDVI value of the vegetation area is higher, and the NDVI value of the non-vegetation area, such as soil and water bodies, is lower; then a threshold h is set, and pixels with NDVI values ​​greater than the threshold h are determined as vegetation pixels. The proportion of the number of vegetation pixels to the total number of pixels is calculated, which is the vegetation coverage;

[0115] (3) Through field measurements, the NDVI values ​​of pure soil areas and pure vegetation areas are obtained and recorded as and ,It should be noted that if field measurement is not possible, ,it can also be determined based on empirical values ​​or historical ,data of the study area;

[0116] (4) The calculated NDVI, and Substitute into the pixel binary model formula , the vegetation coverage FVC of each pixel is obtained, where the calculation result of the vegetation coverage FVC of each pixel is a value between 0 and 1, where 0 represents a pure soil pixel, 1 represents a pure vegetation pixel, and the intermediate values ​​represent different degrees of vegetation coverage;

[0117] (5) After calculating the vegetation coverage of each pixel, it is necessary to verify it. This can be done by comparing it with the vegetation coverage of the field survey or by referring to other existing high-precision vegetation coverage data. If the calculated result deviates from the actual situation, it is necessary to check whether the data processing process and parameter determination are reasonable, and correct any unreasonable points and recalculate.

[0118] S2.3: Generate a digital elevation model based on the acquired terrain data, extract topographic features from the digital elevation model data, and analyze the impact of terrain on soil moisture and nutrient distribution;

[0119] For example, areas with larger slopes may have soil erosion problems, leading to loss of soil nutrients; different slope directions will affect light and water conditions, and thus affect the distribution of soil moisture and nutrients. These analysis results will provide important reference for subsequent fertilization decisions.

[0120] Furthermore, the specific steps of S2.3 include:

[0121] (1) Use GIS software to import pre-processed terrain data and select the interpolation method in the software to construct a digital elevation model;

[0122] (2) Using spatial analysis tools in GIS software, extract topographic features such as slope, aspect, and terrain relief from the generated digital elevation model data;

[0123] (3) In combination with the principles of soil science and hydrology, the effects of the extracted topographic features on soil moisture and nutrient distribution were analyzed. Generally speaking, in areas with large slopes, soil moisture is easily lost, resulting in soil nutrient leaching and relatively low fertility; in areas with gentle slopes, soil moisture and nutrients are more likely to accumulate. Different slope directions will affect the distribution of light and heat. Sunny slopes receive more solar radiation, the soil temperature is higher, water evaporates quickly, and the soil is relatively dry; on the contrary, shady slopes have better soil moisture conditions, but may have poor heat conditions. In areas with large terrain undulations, the distribution of water and nutrients is significantly different. Water is easily accumulated in low-lying areas, and nutrients are easily enriched, while in high areas, there may be a lack of water and nutrients.

[0124] S2.4: Based on the corrected target area plot image, soil color features are extracted according to the spectral reflectance differences of different soil colors and the band reflectance threshold.

[0125] For example, in the visible light band, the reflectivity combination of red, green and blue light bands can be used to distinguish different types of soil colors, and then preliminarily judge the soil fertility.

[0126] Furthermore, the specific steps of S2.4 include:

[0127] (1) Obtain the corrected image of the target area;

[0128] (2) Obtain reflectance data of different types of soil, such as black soil, loess, and red soil, in the visible and near-infrared bands by consulting literature, conducting field spectral measurements, or referring to existing soil spectral databases. Generally speaking, there are significant differences in the reflectance of different soil colors in the red, green, and blue light bands. For example, the reflectance of black soil in each band is relatively low, while the reflectance of red soil in the red light band is high. These differences are important bases for extracting soil color characteristics;

[0129] (3) According to the determined soil color spectral reflectance difference and combined with the actual soil type and characteristics of the study area, set the appropriate band reflectance threshold. The setting of the threshold needs to take into account a variety of factors, such as the diversity of soil colors, the noise level of the image, and the accuracy requirements of the subsequent analysis. Take any two soil colors as an example. Assume that in the red light band, the reflectance range of one soil is between 0.3-0.4, and the reflectance range of the other soil is between 0.5-0.6. In order to accurately distinguish the two soils, the reflectance threshold of the red light band can be set to 0.45. In this way, when the reflectance of a pixel point in the target area plot image in the red light band is less than 0.45, it can be preliminarily judged that the soil corresponding to the pixel point may belong to the former soil; otherwise, it may belong to the latter soil.

[0130] (4) Using the set band reflectance threshold, the corrected target area plot image is processed, including:

[0131] The image processing software ENVI is used to judge each pixel in the corrected target area plot image;

[0132] If the band reflectance of any pixel meets the threshold condition of a specific soil color, the pixel is classified into the corresponding soil color category. For example, in ENVI software, the band operation tool is used to binarize the image according to the set threshold. The pixel points that meet the soil color threshold are assigned a value of 1, and those that do not meet the soil color threshold are assigned a value of 0, thereby obtaining a soil color feature image.

[0133] (5) Through field sampling surveys, compare the image analysis results with the actual soil color. If deviations are found in the extraction results, it is necessary to readjust the band reflectance threshold or check whether there are problems in the image correction process. For example, if the field survey finds that the soil color of certain areas does not match the image analysis results, it is necessary to fine-tune the reflectance threshold based on the field survey results, and then re-extract features until the extraction results are consistent with the actual survey results and meet the accuracy requirements.

[0134] S2.5: According to the requirements of different fertilization models for vegetation index, extract the corresponding vegetation index characteristics, where the vegetation index characteristics include the ratio vegetation index RVI, which is calculated by the formula: Among them, RVI is more sensitive to vegetation, especially in the early stage of vegetation growth, and can more obviously reflect the changes in vegetation. By calculating the RVI value and conducting statistical analysis on it, the difference vegetation index can also be calculated. , Enhanced Vegetation Index (EVI), which evaluates vegetation growth status from different perspectives and combines these vegetation index characteristics to determine crop nutrient requirements. For example, when both NDVI and RVI show weak vegetation growth, it is necessary to increase the supply of fertilizers such as nitrogen fertilizer to promote leaf growth and photosynthesis;

[0135] For example, in the variable fertilization decision model of rice, it is necessary to extract the normalized difference vegetation index NDVI and the standard planting ratio index RISP. These vegetation indices can reflect information such as the growth status, health, and nitrogen content of vegetation, and are important parameters in the fertilization model.

[0136] S2.6: Integrate the extracted soil color features, vegetation cover features, vegetation index features, and topography features to form a fertilization feature dataset.

[0137] The specific steps of step S3 include:

[0138] S3.1: Extract feature vectors from the fertilization feature dataset, where each feature vector includes soil color features, vegetation cover features, vegetation index features, and topographic and geomorphic feature data. The feature extraction utilizes a principal component analysis method, which is known in the art and does not constitute an inventive solution of the present application and is not described in detail herein.

[0139] S3.2: Load a pre-trained decision tree model, input the extracted feature vector into the pre-trained decision tree model, and determine the region category to which the pixel corresponding to the extracted feature vector belongs based on the output of the pre-trained decision tree model. The decision tree model is prior art in this field and does not constitute an inventive solution of the present application, and is not described in detail here.

[0140] S3.3: Assign a unique identification number to each divided area and add attribute tags to the areas based on their characteristics. The unique identification number uses an RFID identification code. The identification process of the RFID identification code is a prior art in this field and does not constitute an inventive solution of this application, so it will not be described in detail here.

[0141] S3.4: Importing the resulting data with the region identifiers into a geographic information system, creating a new layer in the geographic information system to represent the region divisions, and associating the resulting data with geographic coordinates. Creation of the geographic information system layer is prior art in this field and does not constitute the inventive solution of this application, and is not described in detail here.

[0142] S3.5: Set different colors and patterns for different regions based on region identification or attribute labels for visualization, and output the visualization results.

[0143] Example 2:

[0144] See also Figure 4 Another embodiment of the present invention provides a fertilization system based on regional land image, comprising:

[0145] Image processing module, feature extraction module, area division module, fertilization plan formulation module, fertilization execution module, and fertilization effect monitoring module;

[0146] An image processing module is used to obtain the image of the target area and perform correction processing on the image of the target area;

[0147] The feature extraction module is used to extract fertilization feature information from the calibrated target area plot image and integrate it to form a comprehensive fertilization feature data set to provide data support for regional division;

[0148] The regional division module is used to reasonably divide the plots according to the fertilization feature data set using the decision tree classification algorithm, and the division results are displayed in an intuitive and visual form to facilitate users to understand the differences between different areas within the plot;

[0149] The fertilization plan formulation module is used to comprehensively consider multiple factors such as crop varieties, planting density, and regional division results of the plot, and combine with the pre-built fertilization model to generate a fertilization plan for each divided area;

[0150] The fertilization execution module is used to control the automated fertilization equipment to accurately fertilize the corresponding area according to the established fertilization plan, ensuring the accuracy and efficiency of the fertilization process;

[0151] The fertilization effect monitoring module is used to monitor the growth of crops in the field after fertilization, evaluate the fertilization effect based on the monitoring results, and adjust the fertilization plan in time if the effect is not ideal, so as to ensure that the crops can obtain appropriate nutrient supply and achieve good yield and quality goals.

[0152] The image processing module includes: image acquisition unit, radiation correction unit, and geometric correction unit;

[0153] The image acquisition unit is used to collect image data of the target area through multispectral cameras, visible light cameras and other equipment mounted on the drone, obtaining the appearance information of the land in different bands and different perspectives. These original images are the basis for subsequent analysis;

[0154] The radiation correction unit is used to perform radiation correction on the collected images based on the environmental parameters collected during the flight of the UAV and the camera's own radiation calibration data. This correction algorithm eliminates the deviation of the image radiation value caused by environmental factors such as atmosphere and lighting and the camera's own characteristics, so that the image can accurately reflect the true spectral characteristics of the ground objects. Among them, the environmental parameters include atmospheric temperature and humidity, and light intensity.

[0155] The geometric correction unit is used to correct the geometric deformation of the image through a geometric transformation algorithm based on the coordinates of the ground control points and satellite navigation positioning data, so that the shape and position of the objects in the image accurately match the geographic spatial position of the actual plot, ensuring the spatial accuracy of subsequent image-based analysis.

[0156] The feature extraction module includes: soil color feature extraction unit, vegetation coverage feature extraction unit, vegetation index feature extraction unit, topography feature extraction unit, and data integration unit;

[0157] A soil color feature extraction unit is used to identify the soil area in the image and analyze its color features using an image processing algorithm;

[0158] For example, after converting the RGB image to the HSV color space, the pixel proportions corresponding to different hue, saturation, and other ranges are counted to determine the soil color type, such as black soil, loess, etc., and further associate it with soil fertility and other properties to provide a basis for subsequent fertilization judgment, because soils of different colors often have different fertility conditions.

[0159] The vegetation coverage feature extraction unit is used to calculate the normalized difference vegetation index NDVI, combined with the image segmentation algorithm, to count the proportion of vegetation pixels to the total pixels, thereby determining the vegetation coverage;

[0160] The vegetation index feature extraction unit is used to make multi-dimensional judgments on crops by calculating the ratio vegetation index (RVI) and enhanced vegetation index (EVI) in combination with the normalized difference vegetation index (NDVI);

[0161] The terrain feature extraction unit is used to indirectly estimate the terrain conditions by using information such as shadows and textures in the image.

[0162] The data integration unit is used to summarize and organize the extracted characteristic information such as soil color, vegetation coverage, vegetation index and topography, and integrate it into a complete fertilization characteristic data set according to a unified data format and structure.

[0163] The region division module includes: decision tree classification unit, region identification unit, and visualization display unit;

[0164] The decision tree classification unit is used to take the fertilization feature dataset as input data and, based on the decision tree classification algorithm, divide the land into regions according to pre-set rules. Regions with similar features are grouped together to ensure that the internal features of each divided region are relatively uniform. The similar features refer to similar soil fertility, similar crop growth conditions, and the same terrain conditions.

[0165] The regional identification unit is used to give each divided region unique identification information, such as number and name, to facilitate the distinction and reference of different regions in subsequent data processing, fertilization plan formulation and actual operation, to avoid confusion and ensure that the operation of each link for a specific region is accurate;

[0166] The visualization display unit is used to import the results of regional division into the geographic information system software platform and render and display each area using visualization methods such as different colors and patterns;

[0167] For example, green is used to represent areas with high fertility and good crop growth, red is used to represent areas with low fertility and poor crop growth, etc., which intuitively presents the regional distribution within the plot and the differences between the areas, helping agricultural personnel to quickly grasp the overall condition of the plot and better formulate fertilization plans.

[0168] The fertilization plan formulation module includes: data collection unit, fertilization model calling unit, and plan generation unit;

[0169] A data collection unit, used to collect crop variety information, planting density information and regional division result data of the plot;

[0170] A fertilization model calling unit is used to load a pre-established fertilization model;

[0171] The plan generation unit is used to determine the type of fertilizer, amount of fertilizer, time of fertilization and method of fertilization suitable for each area through calculation and analysis based on the collected data and pre-loaded fertilization models, and generate a complete and personalized fertilization plan to ensure that fertilization can not only meet the needs of crop growth, but also maximize fertilizer utilization and reduce waste and environmental pollution.

[0172] The fertilization effect monitoring module includes: image acquisition unit, crop growth monitoring unit, fertilization effect evaluation unit, and program adjustment unit;

[0173] An image acquisition unit is used to use an image acquisition device to collect images of the plot again at a specific time point after fertilization, such as one week after fertilization, to obtain image data of the crop growth status after fertilization. These images can reflect changes in crop appearance, vegetation coverage, etc.;

[0174] The crop growth monitoring unit is used to collect relevant data on crop growth changes from multiple angles through various crop physiological sensors installed in the plot, including changes in crop physiological indicators and morphological characteristics, to fully understand the crop growth response;

[0175] The fertilization effect evaluation unit is used to comprehensively analyze the data obtained by the image acquisition unit and the crop growth monitoring unit, and use the corresponding evaluation algorithm to determine whether the fertilization effect has achieved the expected goal, whether the crop growth conditions in each area have been effectively improved due to fertilization, and whether there are problems such as nutrient excess or deficiency;

[0176] The scheme adjustment unit is used to analyze the causes of problems in areas with unsatisfactory fertilization effects based on the results of the fertilization effect evaluation unit, such as insufficient fertilizer amount, inappropriate fertilizer type, inappropriate fertilization time, etc., and automatically generate adjustment strategies, such as increasing the fertilizer amount, changing the fertilizer type, changing the fertilization time or method, etc., and feed back the adjusted fertilization scheme to the fertilization scheme formulation module and the fertilization execution module to achieve dynamic optimization and adaptive adjustment of the fertilization scheme to ensure the continued healthy growth of crops.

[0177] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific embodiments. The above-mentioned specific embodiments are merely illustrative and not restrictive. Under the guidance of the present invention, ordinary technicians in this field can also change, modify, replace and modify the above-mentioned embodiments without departing from the purpose and scope of protection of the present invention. These are all protected by the present invention.

Claims

1. A fertilization method based on regional land image, characterized in that: include: Step S1: Use a UAV to collect images of the target area and perform radiation correction and geometric correction; Step S2: extracting fertilization feature information based on the corrected target area plot image and integrating it to form a fertilization feature dataset, wherein the fertilization feature information includes soil color features, vegetation coverage features, vegetation index features, and topographic features; Step S3: Based on the fertilization feature dataset, a decision tree classification algorithm is used to partition the land parcels and mark the divided areas. At the same time, a geographic information system is used to visualize the marked area division results. Step S4: Collect crop variety information, planting density information and regional division result data of the plot, and formulate a fertilization plan for each area based on the pre-established fertilization model; Step S5: Fertilizing the regional plots according to the formulated fertilization plan; Step S6: After fertilization, use image acquisition equipment to capture images of the regional plots, monitor the growth changes of crops, evaluate the fertilization effect based on the crop growth response, and adjust the fertilization plan based on the evaluation effect.

2. The method for fertilizing based on regional land image according to claim 1, characterized in that: The specific steps of step S2 include: S2.1: Use the UAV-mounted LiDAR to obtain topographic data of the target area, and simultaneously obtain a calibrated image and multispectral image of the target area; S2.2: Obtain a calibrated multispectral image of the target area and use the pixel dichotomy model method to combine the reflectance information of the near-infrared and red bands in the multispectral image to calculate the vegetation coverage characteristics; S2.3: Generate a digital elevation model based on the acquired terrain data, extract topographic features from the digital elevation model data, and analyze the impact of terrain on soil moisture and nutrient distribution; S2.4: Based on the corrected target area plot image, soil color features are extracted according to the spectral reflectance differences of different soil colors and the band reflectance threshold. S2.5: Extract corresponding vegetation index features based on the requirements of different fertilization models for vegetation index; S2.6: Integrate the extracted soil color features, vegetation cover features, vegetation index features, and topography features to form a fertilization feature dataset.

3. The method for fertilizing based on regional land image according to claim 2, characterized in that: The specific steps of step S1 include: S1.1: Obtain the boundary coordinates of the target area through the geographic information system. Based on the determined target area, use the route planning software Pix4Dcapture to plan the flight route. At the same time, set the image acquisition time interval and shooting mode according to the camera type and shooting purpose. S1.2: Start the UAV and fly it over the target area according to the planned flight route. During the flight, the onboard image acquisition device collects images of the target area according to the set time interval and the shooting mode triggered by the flight position. At the same time, the collected images of the target area are transmitted to the ground control station via a wireless communication link. S1.3: Obtain camera radiometric calibration data from the manufacturer, and use a radiometric correction formula to perform preliminary radiometric value correction on the target area plot image to obtain a radiometric brightness image of the target area plot. The camera radiometric calibration data includes the camera's gain value and offset value.

4. The method for fertilizing based on regional land image according to claim 3, characterized in that: The specific steps of step S1 also include: S1.4: Use the environmental sensor onboard the drone to obtain atmospheric parameter data, and correct the radiation brightness image of the target area based on the obtained atmospheric parameter data to obtain a radiation-corrected image of the target area. The correction formula is: ,in, Indicated by wavelength The reflectivity of the ground objects under the sun, d represents the distance between the sun and the earth, Indicated by wavelength The solar irradiance under represents the cosine of the solar zenith angle, represents the solar zenith angle, Indicates the target area plot image at wavelength The radiance value after down-correction; S1.5: Collect the coordinate information and georeference data of the ground control points and calculate the coordinates of each pixel in the radiometrically corrected target area plot image. Perform coordinate transformation to obtain the corrected coordinate position , and according to the corrected coordinate position , obtain the target area plot image after coordinate correction; S1.6: Use a resampling method to determine the pixel grayscale value corresponding to the corrected coordinate position, and obtain a target area plot image after radiation correction and geometric correction.

5. The method for fertilizing based on regional land image according to claim 3, characterized in that: The specific steps of S1.6 include: S1.61: Obtain a radiation-corrected target area plot image, corrected coordinate position information, and a coordinate-corrected target area plot image; S1.62: Traverse the coordinate positions of the target area plot image after coordinate correction For each pixel coordinate, determine the corresponding four neighboring pixels in the target area block image coordinates after radiation correction 、 、 、 , and the corresponding grayscale values ​​of the four pixels are 、 、 、 ; S1.63: The coordinate position of the target area plot image after coordinate correction Plot each pixel coordinate in the image with the radiometrically corrected target area Calculate the difference of the corresponding elements and get The weights of the pixels on the horizontal and vertical directions respectively and ; S1.64: Calculate the grayscale value of the pixels in the target area plot image after coordinate correction ; S1.65: After the grayscale values ​​of all pixels in the coordinate-corrected target area plot image are determined, the determined pixels are combined into a complete image according to the correct spatial position relationship and saved as an image file to obtain the target area plot image after radiation correction and geometric correction.

6. The method for fertilizing based on regional land image according to claim 4, characterized in that: The specific steps of step S3 include: S3.1: Extract feature vectors from the fertilization feature dataset, where each feature vector contains soil color feature, vegetation cover feature, vegetation index feature, and topographic feature data; S3.2: Load the pre-trained decision tree model, input the extracted feature vector into the pre-trained decision tree model, and determine the region category to which the pixel corresponding to the extracted feature vector belongs based on the output of the pre-trained decision tree model; S3.3: Assign a unique identification number to each divided area and add attribute labels to the areas based on their characteristics; S3.4: Import the division result data with the region identifier into a geographic information system, create a new layer in the geographic information system to represent the region division, and associate the division result data with the geographic coordinates; S3.5: Set different colors and patterns for different regions based on region identification or attribute labels for visualization, and output the visualization results.

7. The method for fertilizing based on regional land image according to claim 1, characterized in that: The fertilization plan in step S4 includes the type of fertilizer, amount of fertilizer, time of fertilizer application and method of fertilizer application for different areas.

8. A fertilization system based on regional land image, which is used to implement the fertilization method based on regional land image according to any one of claims 1 to 7, characterized in that: include: Image processing module, feature extraction module, area division module, fertilization plan formulation module, fertilization execution module, and fertilization effect monitoring module; The image processing module is used to obtain the target area plot image and perform correction processing on the target area plot image; The feature extraction module is used to extract fertilization feature information from the corrected target area plot image and integrate it to form a fertilization feature data set; The region division module is used to partition the plots using a decision tree classification algorithm based on the fertilization feature dataset and to visualize the partition results; The fertilization plan formulation module is used to comprehensively consider multiple factors such as crop variety, planting density and regional division results of the plot, combine with the pre-built fertilization model, and generate a fertilization plan for each divided area; The fertilization execution module is used to control the automated fertilization equipment to fertilize the corresponding area according to the established fertilization plan; The fertilization effect monitoring module is used to monitor the growth of crops in the plot after fertilization, evaluate the fertilization effect based on the monitoring results, and adjust the fertilization plan.

9. The fertilization system based on regional land image according to claim 8, characterized in that: The region division module includes: a decision tree classification unit and a region identification unit; The decision tree classification unit is used to use the fertilization feature data set as input data and divide the land into regions according to pre-set rules based on the decision tree classification algorithm; The area identification unit is used to assign unique identification information to each divided area.

10. The fertilization system based on regional land image according to claim 9, characterized in that: The fertilization plan formulation module includes: a data collection unit, a fertilization model calling unit, and a plan generation unit; The data collection unit is used to collect crop variety information, planting density information and regional division result data of the plot; The fertilization model calling unit is used to load a pre-established fertilization model; The scheme generating unit is used to determine the type, amount, time and method of fertilizer for each area according to the collected data and the preloaded fertilization model, and generate a fertilization scheme based on the situation of each area.

Images