Soil fertility intelligent evaluation method and system in combination with weed growth characteristics

Abstract

The invention relates to the technical field of soil fertility detection, in particular to a soil fertility intelligent evaluation method and system combined with weed growth characteristics, and the method comprises the steps: collecting and preprocessing image data of a target planting area, and obtaining standard image data; inputting pixel blocks in the standard image data into a pre-constructed classification result prediction model, and outputting a classification result of the pixel blocks; dividing the target planting area into regular grids according to a fixed scale; the weed density in each grid area is sequentially obtained, a weed density difference threshold value is preset, when the weed density difference value of the adjacent grids is smaller than or equal to the preset weed density difference threshold value, the adjacent grids are combined into an independent fertilization management area, and a plurality of independent fertilization management areas are obtained. According to the method, the target planting area is divided into the regular grids, and the adjacent grids are combined in combination with the weed density difference threshold value, so that a plurality of independent fertilization management areas are formed, and the extensive management problem caused by traditional whole land fertilization is effectively avoided.

Description

Technical Field

[0001] This invention relates to the field of soil fertility testing technology, specifically to an intelligent soil fertility assessment method and system that incorporates weed growth characteristics. Background Technology

[0002] Chinese patent application CN116562134A discloses a method for predicting soil fertility. This method includes: determining initial sampling points and a central sampling point for a target plot; constructing a sampling point sequence for each initial sampling point; determining preliminary average soil fertility data and preliminary soil fertility change data for each sampling point sequence based on the fertility index of the soil sample corresponding to the top-ranked sampling point; predicting the fertility of soil samples corresponding to sampling points not ranked in the top position; updating the preliminary average soil fertility data and preliminary soil fertility change data based on the prediction results to obtain target average soil fertility data and target soil fertility change data; and obtaining predicted soil fertility data for the target plot based on the target average soil fertility data from all sampling point sequences.

[0003] In current agricultural production practices, soil fertility assessment mainly relies on manual sampling and laboratory testing. For example, in the actual scenario of corn planting, farmers often formulate fertilization plans based on laboratory test results. Taking a certain farmland as an example, laboratory tests showed that the nitrogen content of the plot was low, so the farmer increased nitrogen fertilizer input during the seedling stage. However, in some areas of this farmland, due to the high density of weeds such as foxtail grass and barnyard grass, these weeds rapidly consumed a large amount of nitrogen in the early stages, leading to a further shortage of nitrogen actually available to the crop. At the same time, remote sensing image analysis only showed weak crop growth but could not identify that this was due to competition from weeds. The system misjudged it as "insufficient soil nitrogen" rather than "excessive consumption by weeds," resulting in the farmer continuing to increase the amount of fertilizer. Although the soil test parameters improved, fertilization not only failed to significantly improve corn growth but also further promoted weed reproduction, intensifying the competition between crops and weeds. Ultimately, fertilizer utilization was low, and yield did not increase. Summary of the Invention

[0004] To address the aforementioned problems, this invention provides a method and system for intelligent soil fertility assessment that incorporates weed growth characteristics.

[0005] This invention employs the following technical solution: a method for intelligent soil fertility assessment that incorporates weed growth characteristics, comprising:

[0006] Collect and preprocess image data of the target planting area to obtain standard image data;

[0007] pixel blocks in standard image data The input is fed into a pre-built classification result prediction model, and the output is the classification result of the pixel block. The classification results include crops, weeds, or soil.

[0008] The target planting area is divided into regular grids according to a fixed scale to obtain the preliminary basic grid units;

[0009] The weed density in each grid area is obtained sequentially. A weed density difference threshold is preset. When the weed density difference between adjacent grids is less than or equal to the preset weed density difference threshold, the adjacent grids are merged into an independent fertilization management area. When the weed density difference between adjacent grids is greater than the preset weed density difference threshold, it is also retained as an independent fertilization management area, thus obtaining multiple independent fertilization management areas.

[0010] The field IoT sensor array is used to collect data in real time on crop type, soil nutrient parameters, and soil environmental factors in independent fertilization management areas;

[0011] Extract the average weed density of the independently fertilized management area, and preset the weed density gradient thresholds MDmin and MDax;

[0012] When the average weed density is greater than MDax, a weeding instruction is generated;

[0013] When the average weed density is less than MDmin, a normal fertilization strategy is generated;

[0014] When MDmin ≤ average weed density ≤ MDax, a weed suppression and fertilization strategy is generated.

[0015] As a further description of the above technical solution: the method for obtaining weed density includes:

[0016] Obtain the total number of pixel blocks in the grid area and the total number of pixel blocks classified as weeds;

[0017] Weed density is obtained by dividing the total number of pixels classified as weeds by the total number of pixels in the grid area.

[0018] As a further description of the above technical solution: the soil nutrient parameters include nitrogen content, phosphorus content, potassium content, organic matter content, and electrical conductivity;

[0019] The soil environmental factors include soil temperature, soil moisture, and soil osmotic pressure.

[0020] As a further description of the above technical solution: the method for generating a normal fertilization strategy includes:

[0021] The obtained crop type, soil nutrient parameters, soil environmental factors, and area of ​​independent fertilization management zone are input into the pre-constructed normal fertilization strategy recommendation model to obtain the normal fertilization strategy recommendation set label, and then obtain the normal fertilization strategy corresponding to the normal fertilization strategy set label to generate the normal fertilization strategy.

[0022] As a further description of the above technical solution: the set of normal fertilization strategies is as follows:

[0023] ;

[0024] in, This refers to the nitrogen, phosphorus, and potassium fertilizer content ratio for set label number n. This represents the amount of fertilizer applied when the set label number is n. This refers to the fertilization method corresponding to the set label number n, where n is the sequence number of the normal fertilization strategy set label; the fertilization methods include: broadcasting, strip application, hole application, and foliar spraying;

[0025] The training method for the normal fertilization strategy recommendation model includes:

[0026] Pre-assign corresponding numbers to the set of normal fertilization strategies;

[0027] The crop type, soil nutrient parameters, soil environmental factors, and area of ​​the independently fertilized management zone are converted into a set of corresponding feature vectors. Each set of feature vectors is used as input to the normal fertilization strategy recommendation model. The normal fertilization strategy recommendation model outputs a set of normal fertilization strategy numbers corresponding to each set of crop type, soil nutrient parameters, soil environmental factors, and area of ​​the independently fertilized management zone. The actual set of normal fertilization strategy numbers corresponding to each set of crop type, soil nutrient parameters, soil environmental factors, and area of ​​the independently fertilized management zone is used as the prediction target. The training objective is to minimize the loss function value of the normal fertilization strategy recommendation model. Training stops when the loss function value of the normal fertilization strategy recommendation model is less than or equal to the preset target loss value.

[0028] As a further description of the above technical solution: the method for generating a weed suppression and fertilization strategy includes:

[0029] Weed characteristic parameters are obtained, and based on the weed characteristic parameters, the competition pressure factor is calculated. The obtained competition pressure factor, crop type, soil nutrient parameters, soil environmental factors, and area of ​​independent fertilization management area are input into the pre-constructed weed suppression and fertilization strategy recommendation model to obtain the weed suppression and fertilization strategy recommendation set label. Then, the weed suppression and fertilization strategy corresponding to the weed suppression and fertilization strategy set label is obtained, and the weed suppression and fertilization strategy is generated.

[0030] As a further description of the above technical solution: the weed characteristic parameters include weed species, the coverage density of each weed species in an independent fertilization management area, spectral index NDVI, and canopy height.

[0031] As a further description of the above technical solution: the method for calculating the competition pressure factor based on weed characteristic parameters includes;

[0032] Weed species were determined by a weighted summation method based on the spectral index NDVI and canopy height. Growth index in independently fertilized management areas ;

[0033] For each type of weed, classify the weed species. The weed species are obtained by multiplying the weight, the coverage density in the independently fertilized management area, and the growth index. The competitive contribution value of each weed species in the independently fertilized management area is obtained by summing the competitive contribution values ​​of all weed species in the independently fertilized management area to obtain the competitive pressure factor.

[0034] As a further description of the above technical solution: the method for obtaining the coverage density of each weed species in an independently fertilized management area includes:

[0035] Obtain weed species Extract the pixel blocks corresponding to the independent fertilization management area based on the corresponding pixel count. To obtain weed species The corresponding number of pixel blocks indicates the type of weed. The number of pixels divided by the total number of pixels in the independently fertilized management area yields the weed species. Coverage density within independently fertilized management areas.

[0036] A soil fertility intelligent assessment system incorporating weed growth characteristics is used to implement the aforementioned soil fertility intelligent assessment method incorporating weed growth characteristics. The system includes:

[0037] The image acquisition module acquires and preprocesses image data of the target planting area to obtain standard image data;

[0038] The image analysis module separates pixel blocks from standard image data. The input is fed into a pre-built classification result prediction model, and the output is the classification result of the pixel block. The classification results include crops, weeds, or soil.

[0039] The grid division module divides the target planting area into regular grids according to a fixed scale, obtaining the preliminary basic grid units;

[0040] The grid merging module sequentially obtains the weed density in each grid area, presets a weed density difference threshold, and merges adjacent grids into an independent fertilization management area when the weed density difference between adjacent grids is less than or equal to the preset weed density difference threshold, thus obtaining multiple independent fertilization management areas.

[0041] The parameter acquisition module uses a field IoT sensor array to collect real-time data on crop type, soil nutrient parameters, and soil environmental factors in independent fertilization management areas.

[0042] The instruction generation module extracts the average weed density of the independent fertilization management area, presets weed density gradient thresholds MDmin and MDax, generates a weeding instruction when the average weed density is greater than MDax, generates a normal fertilization strategy when the average weed density is less than MDmin, and generates a weed suppression and fertilization strategy when MDmin ≤ average weed density ≤ MDax.

[0043] The beneficial effects of this invention are as follows:

[0044] This invention provides an intelligent soil fertility assessment method that incorporates weed growth characteristics. By introducing weed identification and competitive pressure analysis into the soil fertility assessment process, it effectively overcomes the shortcomings of existing technologies that rely solely on manual sampling, laboratory testing, and simple remote sensing image analysis. This application can accurately distinguish between crops, weeds, and bare soil, avoiding the misjudgment of weak crop growth due to weed nutrient consumption as insufficient soil fertility. Furthermore, by incorporating weed competitive pressure factors into fertilization strategy decisions, this application can dynamically identify the consumption effects of weeds on key nutrients such as nitrogen, phosphorus, and potassium, thereby achieving a more realistic fertility assessment in complex field environments.

[0045] The dual-path decision-making mechanism proposed in this application, consisting of a "normal fertilization strategy" and a "weed suppression and fertilization strategy," not only provides refined fertilization ratios and methods based on soil nutrient parameters, environmental factors, and crop type, but also intelligently switches between coupled management modes of fertilization and weed control when weed density falls within different ranges. When weed competition pressure is high, this application can simultaneously generate herbicide application and quantitative fertilization plans, alleviating nutrient competition between weeds and crops at the source. Therefore, this application can significantly improve fertilizer utilization, avoid the vicious cycle of fertilizer resource waste and excessive weed growth, and achieve a synergistic improvement in crop yield and fertilizer efficiency. Attached Figure Description

[0046] The present invention will be further explained below with reference to the accompanying drawings and embodiments:

[0047] Figure 1 A flowchart of an intelligent soil fertility assessment method that incorporates weed growth characteristics, provided in Embodiment 1 of the present invention;

[0048] Figure 2 This is a flowchart of the method for generating a normal fertilization strategy provided in Embodiment 1 of the present invention;

[0049] Figure 3 This is a flowchart of the method for generating a weed-suppressing and fertilizer-supplementing strategy provided in Embodiment 1 of the present invention;

[0050] Figure 4 This is a module connection diagram of an intelligent soil fertility assessment system that incorporates weed growth characteristics, as provided in Embodiment 2 of the present invention. Detailed Implementation

[0051] To make the technical means, creative features, objectives, and effects of this invention readily understandable, the invention is further described below with reference to specific illustrations. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0052] Example 1

[0053] Please see Figures 1-3 This invention provides a technical solution: a method for intelligent soil fertility assessment that combines weed growth characteristics, comprising:

[0054] Image data of the target planting area is collected and preprocessed to obtain standard image data; pixel blocks Indicates standard image in Pixel value at; These are the pixel coordinates in the image data;

[0055] It should be noted that the image data was acquired by drone. The preprocessing includes noise reduction, geometric correction, and radiometric calibration to obtain standard image data. The above preprocessing techniques are all existing technologies, and the preprocessing process will not be described in detail here.

[0056] pixel blocks in standard image data The input is fed into a pre-built classification result prediction model, and the output is the classification result of the pixel block. The classification results include crops, weeds, or soil.

[0057] in, ;

[0058] coordinates Pixel block classification results The pixel block belongs to the category The predicted probability, The parameter maximization operation means finding the category that maximizes the predicted probability. ;

[0059] Training methods for classification result prediction models include:

[0060] Use fully convolutional deep neural networks, such as U-Net, SegNet, or DeepLab series networks. The network structure includes an encoder (extracting image features), a decoder (restoring spatial resolution), and a pixel classification layer.

[0061] Images were collected from actual farmland, and pixel categories of crops, weeds, and soil were labeled manually or using semi-automatic annotation software to construct a training dataset. in To classify real pixel blocks, the dataset is divided into training, validation, and test sets.

[0062] Define cross-entropy as the loss function The expression is:

[0063] ;

[0064] In the formula, The true pixel category label is represented in one-hot encoded form. Predict the category to which a pixel belongs for the network. The probability, log(˙) is a logarithmic function;

[0065] Iterative training is performed using stochastic gradient descent (SGD) until the accuracy on the validation set reaches a preset threshold, for example, IoU > 85%, at which point the model training is complete.

[0066] The target planting area is divided into regular grids according to a fixed scale, resulting in the preliminary basic grid cells, represented as follows: ,in, For the first One basic grid cell;

[0067] It should be noted that the size of each grid can be set according to the resolution of the drone image or the precision of agricultural machinery operation, for example, 5m×5m or 10m×10m;

[0068] The weed density in each grid area is obtained sequentially. The method for obtaining the weed density is to obtain the total number of pixel blocks in the grid area and the total number of pixel blocks classified as weeds. The weed density is obtained by dividing the total number of pixel blocks classified as weeds by the total number of pixel blocks in the grid area.

[0069] A preset weed density difference threshold is set. When the weed density difference between adjacent grids is less than or equal to the preset weed density difference threshold, they are merged into an independent fertilization management area. When the weed density difference between adjacent grids is greater than the preset weed density difference threshold, they are also retained as an independent fertilization management area, resulting in multiple independent fertilization management areas.

[0070] In this embodiment, the target planting area is divided into regular grids at a fixed scale, and adjacent grids are merged based on a weed density difference threshold to form multiple independent fertilization management areas. This avoids the extensive management problem of "treating the entire plot of land as a unified fertilization unit" in existing technologies, allowing fertilization decisions to be made at a smaller spatial scale. This accurately matches the weed competition situation and soil nutrient differences in different areas, thereby reducing over-fertilization or under-fertilization. On the other hand, by introducing a weed density difference threshold, discrete noise at the area boundaries can be effectively eliminated, and grids with similar ecological environments and weed distributions can be automatically merged into one management area, ensuring that nutrient requirements and weed disturbances within the area are relatively consistent.

[0071] Therefore, this regional division method improves the spatial accuracy of fertilization management while avoiding the management complexity caused by excessive fragmentation. It achieves a balance between precision and operability, thus providing a more reasonable and efficient foundation for subsequent fertility assessment and fertilization strategy recommendations.

[0072] The field IoT sensor array is used to collect data in real time on crop type, soil nutrient parameters, and soil environmental factors in independent fertilization management areas;

[0073] The soil nutrient parameters include nitrogen content, phosphorus content, potassium content, organic matter content, and electrical conductivity;

[0074] The soil environmental factors include soil temperature, soil moisture, and soil osmotic pressure;

[0075] Extract the average weed density of the independently fertilized management area, and preset the weed density gradient thresholds MDmin and MDax;

[0076] The method for obtaining the average weed density of the independently fertilized management area includes: obtaining the sum of the weed densities of all grid areas in the independently fertilized management area, and then dividing by the number of grid areas in the independently fertilized management area to obtain the average weed density.

[0077] When the average weed density is greater than MDax, a weeding instruction is generated;

[0078] When the average weed density is less than MDmin, a normal fertilization strategy is generated;

[0079] Methods for generating normal fertilization strategies include:

[0080] The obtained crop type, soil nutrient parameters, soil environmental factors, and area of ​​independent fertilization management area are input into the pre-constructed normal fertilization strategy recommendation model to obtain the normal fertilization strategy recommendation set label, and then obtain the normal fertilization strategy corresponding to the normal fertilization strategy set label to generate the normal fertilization strategy.

[0081] The training method for the normal fertilization strategy prediction model includes:

[0082] The set of normal fertilization strategies is as follows:

[0083] ;

[0084] in, This refers to the nitrogen, phosphorus, and potassium fertilizer content ratio for set label number n. This represents the amount of fertilizer applied when the set label number is n. This refers to the fertilization method corresponding to the set label number n, where n is the sequence number of the normal fertilization strategy set label; the fertilization methods include: broadcasting, strip application, hole application, and foliar spraying.

[0085] Training methods for the normal fertilization strategy recommendation model include:

[0086] Pre-assign corresponding numbers to the set of normal fertilization strategies;

[0087] The crop type, soil nutrient parameters, soil environmental factors, and area of ​​the independently fertilized management zone are converted into a set of corresponding feature vectors. Each set of feature vectors is used as input to the normal fertilization strategy recommendation model. The normal fertilization strategy recommendation model outputs a set of normal fertilization strategy numbers corresponding to each set of crop type, soil nutrient parameters, soil environmental factors, and area of ​​the independently fertilized management zone. The actual set of normal fertilization strategy numbers corresponding to each set of crop type, soil nutrient parameters, soil environmental factors, and area of ​​the independently fertilized management zone is used as the prediction target. The training objective is to minimize the loss function value of the normal fertilization strategy recommendation model. Training stops when the loss function value of the normal fertilization strategy recommendation model is less than or equal to the preset target loss value.

[0088] The recommended model for normal fertilization strategies can be one of the following: support vector machine regression, random forest regression, or neural network regression.

[0089] The loss function value of the recommended model for the normal fertilization strategy is the mean squared error.

[0090] Mean squared error is one of the commonly used loss functions. By training the model with the goal of minimizing the loss function, the normal fertilization strategy recommendation model can better fit the data, thereby improving the model's performance and accuracy.

[0091] loss function ,in, The recommended model loss function value is given for the normal fertilization strategy, where x is the feature vector group number and p is the number of feature vector groups. Let x be the set number of the normal fertilization strategies corresponding to the x-th feature vector. The set number of the normal fertilization strategy corresponding to the x-th feature vector in real time.

[0092] Other model parameters of the recommended model for the normal fertilization strategy, such as the target loss value, optimization algorithm, ratio of training set to test set to validation set, and optimization of the loss function, were all obtained through actual engineering implementation and continuous experimental tuning.

[0093] When MDmin ≤ average weed density ≤ MDax, a weed suppression and fertilization strategy is generated.

[0094] Methods for generating weed suppression and fertilization strategies include:

[0095] Weed characteristic parameters are obtained, and based on the weed characteristic parameters, the competition pressure factor is calculated. The obtained competition pressure factor, crop type, soil nutrient parameters, soil environmental factors and independent fertilization management area are input into the pre-constructed weed suppression and fertilization strategy recommendation model to obtain the weed suppression and fertilization strategy recommendation set label. Then, the weed suppression and fertilization strategy corresponding to the weed suppression and fertilization strategy set label is obtained, and the weed suppression and fertilization strategy is generated.

[0096] Methods for obtaining weed characteristic parameters and calculating competition pressure factors based on these parameters include:

[0097] The weed characteristic parameters include weed species, coverage density of each weed species in an independently fertilized management area, spectral index NDVI, and canopy height;

[0098] The methods for obtaining the cover density of each weed species in an independently fertilized management area include:

[0099] Obtain weed species Extract the pixel blocks corresponding to the independent fertilization management area based on the corresponding pixel count. To obtain weed species The corresponding number of pixel blocks indicates the type of weed. The number of pixels divided by the total number of pixels in the independently fertilized management area yields the weed species. Coverage density within independently fertilized management areas;

[0100] The method for obtaining the canopy height includes:

[0101] The calculation is labeled as weed species. The average height of the pixel block is calculated using the following formula: In the formula, Weed species canopy height, Weed species The corresponding number of pixel blocks Weed species A collection of pixels within an independently managed fertilization area; This indicates the height of the vegetation canopy in a pixel within an independently fertilized management area.

[0102] It should be noted that the spectral index NDVI is obtained using existing technology, calculated by acquiring the reflectance of pixel blocks in the NIR and RED bands.

[0103] Methods for calculating competition pressure factors based on weed characteristic parameters include:

[0104] Weed species were determined by a weighted summation method based on the spectral index NDVI and canopy height. Growth index in independently fertilized management areas ;

[0105] Optionally, the growth index The calculation formula is: ;

[0106] It should be noted that, and These are weighting coefficients, and all are greater than 0. + =1. The formulas mentioned above are all dimensionless calculations, and are obtained by software simulation based on a large amount of data to get the closest formula to the real situation. The weight coefficients in the formulas and the preset thresholds in the analysis process are set by those skilled in the art according to the actual situation or obtained by simulation of a large amount of data. The size of the weight coefficient is a specific value obtained by quantifying each parameter to facilitate subsequent comparison. The size of the weight coefficient depends on the amount of sample data and the processing coefficients initially set by those skilled in the art for each set of sample data. As long as it does not affect the ratio between the parameter and the quantified value, it is acceptable.

[0107] For each type of weed, classify the weed species. The weed species are obtained by multiplying the weight, the coverage density in the independently fertilized management area, and the growth index. The competitive contribution value of each weed species in the independently fertilized management area is obtained, and the competitive pressure factor is obtained by summing them up.

[0108] It should be noted that the formula for calculating the competitive pressure factor is as follows:

[0109] In the formula, As a competitive pressure factor, Weed species The weight, For the coverage density in an independently fertilized management area, This is a growth index;

[0110] It should be noted that, The growth index is used to indicate the differences in the competitive ability of different weeds to crops, and it indicates the degree of vigorous growth of the weed.

[0111] Training methods for recommendation models of weed suppression and fertilization strategies include:

[0112] The set of weed suppression and fertilizer supplementation strategies is as follows: ;in, This refers to the nitrogen, phosphorus, and potassium fertilizer content ratio when the set label is m. This represents the amount of fertilizer applied when the set label is m. This refers to the fertilization method corresponding to set label m. This refers to the herbicide type corresponding to set label m. The set label is m, which corresponds to the herbicide suppression dosage. m is the sequence number of the herbicide suppression and fertilization strategy set label. The fertilization methods include broadcasting, strip application, hole application, and foliar spraying.

[0113] Training methods for recommendation models of weed suppression and fertilization strategies include:

[0114] The recommended model set for weed suppression and fertilizer supplementation strategies is assigned corresponding numbers in advance;

[0115] The competitive pressure factors, crop type, soil nutrient parameters, soil environmental factors, and area of ​​independently fertilized management zones are converted into a set of corresponding feature vectors. Each set of feature vectors is used as input to the weed suppression and fertilization strategy recommendation model. The weed suppression and fertilization strategy recommendation model outputs a set of weed suppression and fertilization strategy numbers corresponding to each set of competitive pressure factors, crop type, soil nutrient parameters, soil environmental factors, and area of ​​independently fertilized management zones. The actual set of weed suppression and fertilization strategy numbers corresponding to each set of competitive pressure factors, crop type, soil nutrient parameters, soil environmental factors, and area of ​​independently fertilized management zones is used as the prediction target. The training objective is to minimize the loss function value of the weed suppression and fertilization strategy recommendation model. Training stops when the loss function value of the weed suppression and fertilization strategy recommendation model is less than or equal to the preset target loss value.

[0116] The recommended model for the weed suppression and fertilization strategy can be one of the following: support vector machine regression, random forest regression, or neural network regression.

[0117] The loss function value of the recommended model for the weed suppression and fertilization strategy is also the mean squared error.

[0118] Other model parameters of the recommended model for the weed suppression and fertilization strategy, such as the target loss value, optimization algorithm, ratio of training set to test set to validation set, and optimization of the loss function, were all obtained through actual engineering implementation and continuous experimental tuning.

[0119] In this embodiment, by introducing weed identification and competitive pressure analysis into the soil fertility assessment process, this application effectively overcomes the limitations of existing technologies that rely on manual sampling, laboratory testing, and rough judgment based on remote sensing images. The system can distinguish between crops, weeds, and bare soil at the image level, and quantify weed density, coverage, and growth parameters, converting them into calculable competitive pressure factors. This avoids the problem of weak crop growth due to weed consumption of nutrients being misjudged as insufficient soil fertility. Simultaneously, the dual-path decision-making mechanism of "normal fertilization strategy" and "weed suppression and fertilization strategy" proposed in this application can intelligently switch between fertilization and weed control modes according to different weed density ranges, achieving dynamic regulation of "weed consumption + crop demand," thereby improving fertilizer utilization and avoiding the vicious cycle of "fertilization promoting weed growth." Ultimately, this achieves synergistic optimization of crop nutrient supply, yield increase, and field ecological balance.

[0120] Furthermore, this application divides the target planting area into regular grids and merges adjacent grids based on weed density difference thresholds to form multiple independent fertilization management zones. This effectively avoids the extensive management problems caused by traditional whole-plot fertilization. This method improves the spatial accuracy of fertilization management, enabling decisions to match nutrient and weed differences within the area, while avoiding the operational complexity caused by excessive fragmentation, ensuring consistency in nutrient requirements and weed distribution within the area. Thus, zone division not only provides more reasonable spatial support for fertility assessment and supplemental fertilization strategies but also promotes the transformation of fertilization management from "average" to "differentiated + intelligent," further highlighting the advantages of this application in terms of creativity and application value.

[0121] Example 2

[0122] Please see Figure 4 This invention provides a technical solution: an intelligent soil fertility assessment system that incorporates weed growth characteristics, used to implement the aforementioned intelligent soil fertility assessment method that incorporates weed growth characteristics. The system includes:

[0123] The image acquisition module acquires and preprocesses image data of the target planting area to obtain standard image data;

[0124] The image analysis module separates pixel blocks from standard image data. The input is fed into a pre-built classification result prediction model, and the output is the classification result of the pixel block. The classification results include crops, weeds, or soil.

[0125] The grid division module divides the target planting area into regular grids according to a fixed scale, obtaining the preliminary basic grid units;

[0126] The grid merging module sequentially obtains the weed density in each grid area, presets a weed density difference threshold, and merges adjacent grids into an independent fertilization management area when the weed density difference between adjacent grids is less than or equal to the preset weed density difference threshold, thus obtaining multiple independent fertilization management areas.

[0127] The parameter acquisition module uses a field IoT sensor array to collect real-time data on crop type, soil nutrient parameters, and soil environmental factors in independent fertilization management areas.

[0128] The instruction generation module extracts the average weed density of the independent fertilization management area, presets weed density gradient thresholds MDmin and MDax, generates a weeding instruction when the average weed density is greater than MDax, generates a normal fertilization strategy when the average weed density is less than MDmin, and generates a weed suppression and fertilization strategy when MDmin ≤ average weed density ≤ MDax.

[0129] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A method for intelligently assessing soil fertility by incorporating weed growth characteristics, characterized in that, include: Collect and preprocess image data of the target planting area to obtain standard image data; pixel blocks in standard image data The input is fed into a pre-built classification result prediction model, and the output is the classification result of the pixel block. The classification results include crops, weeds, or soil. The target planting area is divided into regular grids according to a fixed scale to obtain the preliminary basic grid units; The weed density in each grid area is obtained sequentially. A weed density difference threshold is preset. When the weed density difference between adjacent grids is less than or equal to the preset weed density difference threshold, the adjacent grids are merged into an independent fertilization management area, thus obtaining multiple independent fertilization management areas. The field IoT sensor array is used to collect data in real time on crop type, soil nutrient parameters, and soil environmental factors in independent fertilization management areas; Extract the average weed density of the independently fertilized management area, and preset the weed density gradient thresholds MDmin and MDax; When the average weed density is greater than MDax, a weeding instruction is generated; When the average weed density is less than MDmin, a normal fertilization strategy is generated; When MDmin ≤ average weed density ≤ MDax, a weed suppression and fertilization strategy is generated.

2. The intelligent soil fertility assessment method combining weed growth characteristics according to claim 1, characterized in that, The method for obtaining the weed density includes: Obtain the total number of pixel blocks in the grid area and the total number of pixel blocks classified as weeds; Weed density is obtained by dividing the total number of pixels classified as weeds by the total number of pixels in the grid area.

3. The intelligent soil fertility assessment method combining weed growth characteristics according to claim 1, characterized in that, The soil nutrient parameters include nitrogen content, phosphorus content, potassium content, organic matter content, and electrical conductivity; The soil environmental factors include soil temperature, soil moisture, and soil osmotic pressure.

4. The intelligent soil fertility assessment method combining weed growth characteristics according to claim 1, characterized in that, Methods for generating normal fertilization strategies include: The obtained crop type, soil nutrient parameters, soil environmental factors, and area of ​​independent fertilization management zone are input into the pre-constructed normal fertilization strategy recommendation model to obtain the normal fertilization strategy recommendation set label, and then obtain the normal fertilization strategy corresponding to the normal fertilization strategy set label to generate the normal fertilization strategy.

5. The intelligent soil fertility assessment method combining weed growth characteristics according to claim 4, characterized in that, The set of normal fertilization strategies is as follows: ； in, This refers to the nitrogen, phosphorus, and potassium fertilizer content ratio for set label number n. This represents the amount of fertilizer applied when the set label number is n. This refers to the fertilization method corresponding to the set label number n, where n is the sequence number of the normal fertilization strategy set label; the fertilization methods include: broadcasting, strip application, hole application, and foliar spraying; The training method for the normal fertilization strategy recommendation model includes: Pre-assign corresponding numbers to the set of normal fertilization strategies; The crop type, soil nutrient parameters, soil environmental factors, and area of ​​the independently fertilized management zone are converted into a set of corresponding feature vectors. Each set of feature vectors is used as input to the normal fertilization strategy recommendation model. The normal fertilization strategy recommendation model outputs a set of normal fertilization strategy numbers corresponding to each set of crop type, soil nutrient parameters, soil environmental factors, and area of ​​the independently fertilized management zone. The actual set of normal fertilization strategy numbers corresponding to each set of crop type, soil nutrient parameters, soil environmental factors, and area of ​​the independently fertilized management zone is used as the prediction target. The training objective is to minimize the loss function value of the normal fertilization strategy recommendation model. Training stops when the loss function value of the normal fertilization strategy recommendation model is less than or equal to the preset target loss value.

6. The intelligent soil fertility assessment method combining weed growth characteristics according to claim 1, characterized in that, Methods for generating weed suppression and fertilization strategies include: Weed characteristic parameters are obtained, and based on the weed characteristic parameters, the competition pressure factor is calculated. The obtained competition pressure factor, crop type, soil nutrient parameters, soil environmental factors, and area of ​​independent fertilization management area are input into the pre-constructed weed suppression and fertilization strategy recommendation model to obtain the weed suppression and fertilization strategy recommendation set label. Then, the weed suppression and fertilization strategy corresponding to the weed suppression and fertilization strategy set label is obtained, and the weed suppression and fertilization strategy is generated.

7. The intelligent soil fertility assessment method combining weed growth characteristics according to claim 6, characterized in that, The weed characteristic parameters include weed species, the coverage density of each weed species in an independently fertilized management area, the spectral index NDVI, and the canopy height.

8. The intelligent soil fertility assessment method combining weed growth characteristics according to claim 7, characterized in that, The method for calculating the competition pressure factor based on weed characteristic parameters includes: Weed species were determined by a weighted summation method based on the spectral index NDVI and canopy height. Growth index in independently fertilized management areas ; For each type of weed, classify the weed species. The weed species are obtained by multiplying the weight, the coverage density in the independently fertilized management area, and the growth index. The competitive contribution value of each weed species in the independently fertilized management area is obtained by summing the competitive contribution values ​​of all weed species in the independently fertilized management area to obtain the competitive pressure factor.

9. The intelligent soil fertility assessment method combining weed growth characteristics according to claim 7, characterized in that, The method for obtaining the coverage density of each weed species in an independently fertilized management area includes: Obtain weed species Extract the pixel blocks corresponding to the independent fertilization management area based on the corresponding pixel count. To obtain weed species The corresponding number of pixel blocks indicates the type of weed. The number of pixels divided by the total number of pixels in the independently fertilized management area yields the weed species. Coverage density within independently fertilized management areas.

10. A soil fertility intelligent assessment system incorporating weed growth characteristics, used to implement the soil fertility intelligent assessment method incorporating weed growth characteristics as described in any one of claims 1-9, characterized in that, The system includes: The image acquisition module acquires and preprocesses image data of the target planting area to obtain standard image data; The image analysis module separates pixel blocks from standard image data. The input is fed into a pre-built classification result prediction model, and the output is the classification result of the pixel block. The classification results include crops, weeds, or soil. The grid division module divides the target planting area into regular grids according to a fixed scale, obtaining the preliminary basic grid units; The grid merging module sequentially obtains the weed density in each grid area, presets a weed density difference threshold, and merges adjacent grids into an independent fertilization management area when the weed density difference between adjacent grids is less than or equal to the preset weed density difference threshold, thus obtaining multiple independent fertilization management areas. The parameter acquisition module uses a field IoT sensor array to collect real-time data on crop type, soil nutrient parameters, and soil environmental factors in independent fertilization management areas. The instruction generation module extracts the average weed density of the independent fertilization management area, presets weed density gradient thresholds MDmin and MDax, generates a weeding instruction when the average weed density is greater than MDax, generates a normal fertilization strategy when the average weed density is less than MDmin, and generates a weed suppression and fertilization strategy when MDmin ≤ average weed density ≤ MDax.

Images