Corn topdressing management method and system based on image recognition

Abstract

The application is suitable for the technical field of corn topdressing, and particularly relates to a corn topdressing management method and system based on image recognition, which comprises the following steps: delimiting a monitoring area needing corn topdressing, dividing the monitoring area into a plurality of blocks, collecting real scene images in the monitoring area, selecting a target block based on a preset evaluation rule, constructing a test scene, collecting a plurality of reflection signals by using a light source group and a light intensity sensor pre-installed in the block, wherein the light source group comprises near-infrared light and red light, and each light corresponds to a group of reflection signals. The application can quantify the fertility deficiency degree of an abnormal block by calculating a reference value, further improves the accuracy of fertilization, avoids the deviation problem caused by unified fertilization, provides a reference basis for subsequent accurate fertilization of the monitoring area, finally realizes the change from experience guidance to accurate fertilization, and effectively improves the fertilizer utilization efficiency.

Description

Technical Field

[0001] This invention relates to the field of corn topdressing technology, and in particular to a corn topdressing management method and system based on image recognition. Background Technology

[0002] Topdressing management of corn refers to the management measures taken during the corn growth process to supplement nutrients such as nitrogen, phosphorus, and potassium according to the changes in nutrient requirements at different growth stages, combined with soil fertility, climate conditions, and plant growth. In the current technology, fertilization is generally carried out on a whole plot or large area as a unit, lacking fine segmentation of the monitoring area, making it difficult to identify the uneven distribution of nutrients in the farmland, resulting in local nutrient deficiency or over-fertilization.

[0003] Therefore, "how to conduct precise monitoring of corn growth and apply fertilizer accurately" is the technical problem that this invention aims to solve. Summary of the Invention

[0004] The purpose of this invention is to provide a corn topdressing management method and system based on image recognition, so as to solve the problem of "how to conduct refined monitoring of corn growth and precise topdressing" mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] A corn topdressing management method based on image recognition, the method comprising:

[0007] The monitoring area for corn topdressing is delineated, and the monitoring area is divided into several blocks. Real-world images of the monitoring area are collected. Based on the preset evaluation rules, target blocks are selected, test scenarios are constructed, and multiple sets of reflection signals are collected using light source groups and light intensity sensors pre-installed in the blocks. The light source groups include near-infrared light and red light, and each light ray corresponds to a set of reflection signals.

[0008] Blocks with enhanced near-infrared light reflection signals and weakened red light reflection signals are defined as abnormal blocks. Topdressing tasks are created and sent to preset terminals. Nutrient data is collected using sensing devices deployed in the abnormal blocks, and the topdressing tasks and abnormal blocks are corrected.

[0009] Draw a planar distribution map of the monitoring area, set the initial weights of the blocks, and mark them on the planar distribution map. Iterate through the neighborhood of each block in turn, determine whether there are abnormal blocks in the neighborhood, and if so, use the preset set value to offset the initial weight of the central block to obtain the true weight. Summarize the offset results of all blocks to generate a weight distribution map.

[0010] The processing results of the topdressing task are collected, wherein the processing results include at least: fertilizer application amount and growth changes, standard values ​​are generated, reference values ​​for each abnormal block are calculated through the true weight, and written into the corresponding block.

[0011] Furthermore, the step of delineating the monitoring area requiring corn topdressing, dividing the monitoring area into several blocks, and collecting real-scene images within the monitoring area includes:

[0012] Create a state recognition model to label risk features in real-world images, wherein the risk features include at least: stunted growth, yellowing leaves, and curled leaves;

[0013] All real-world images and risk features are integrated to generate a training set, which is then used to train the state recognition model.

[0014] Furthermore, the step of acquiring multiple sets of reflection signals using light source groups and light intensity sensors pre-installed in the blocks includes:

[0015] Record the time corresponding to the test scene, extract the light intensity parameters in the reflected signal, and establish the correspondence between time and light intensity parameters;

[0016] Based on the aforementioned correspondence, a light intensity variation curve is plotted with time as the horizontal axis and light intensity parameter as the vertical axis, wherein each ray in the light source group corresponds to a light intensity variation curve.

[0017] Furthermore, the step of collecting nutrient data using sensing devices deployed in the anomaly block and correcting the topdressing task and the anomaly block includes:

[0018] The nutrient data is divided into several individual items, and a fluctuation range corresponding to each individual item is created;

[0019] When the nutrient data exceeds the fluctuation range, the corresponding item is defined as an anomaly, and the pre-edited emergency response rules are activated.

[0020] Furthermore, the step of drawing a planar distribution map of the monitoring area and setting the initial weights of the blocks includes:

[0021] Factors influencing the setting of initial weights include, at a minimum, planting location and growth vigor.

[0022] Based on the aforementioned influencing factors, the actual weight of each block is adjusted.

[0023] Furthermore, the step of calculating the reference value of each anomalous block via the true weight and writing it into the corresponding block includes:

[0024] Collect environmental data within the monitoring area and construct several triggering rules;

[0025] When the real-time value of the environmental data meets the triggering rules, the abnormal block and the true weight are updated.

[0026] Furthermore, the system includes:

[0027] The delineation module is used to delineate the monitoring area where corn topdressing is required. The monitoring area is divided into several blocks, real-scene images of the monitoring area are collected, target blocks are selected based on preset evaluation rules, test scenarios are constructed, and multiple sets of reflection signals are collected using light source groups and light intensity sensors pre-installed in the blocks. The light source groups include near-infrared light and red light, and each light ray corresponds to a set of reflection signals.

[0028] The correction module is used to define blocks that enhance near-infrared light reflection signals and weaken red light reflection signals as abnormal blocks, create topdressing tasks, and send them to preset terminals. The module uses sensing devices deployed in the abnormal blocks to collect nutrient data and correct the topdressing tasks and abnormal blocks.

[0029] The summary module is used to draw a planar distribution map of the monitoring area, set the initial weight of the blocks and mark them on the planar distribution map, traverse the neighborhood of each block in turn, determine whether there are abnormal blocks in the neighborhood, if so, use the preset set value to offset the initial weight of the central block to obtain the true weight, summarize the offset results of all blocks, and generate a weight distribution map.

[0030] The writing module is used to collect the processing results of the topdressing task, wherein the processing results include at least: fertilizer application amount and growth changes, generate standard values, calculate the reference value of each abnormal block through the real weight, and write it into the corresponding block.

[0031] Furthermore, the delineation module includes:

[0032] A creation unit is used to create a state recognition model and label risk features in real-world images, wherein the risk features include at least: stunted growth, yellowing leaves, and curled leaves;

[0033] The training unit is used to integrate all real-world images and risk features, generate a training set, and train the state recognition model.

[0034] The recording unit is used to record the time corresponding to the test scene, extract the light intensity parameters in the reflected signal, and establish the correspondence between time and light intensity parameters.

[0035] The drawing unit is used to draw a light intensity change curve based on the correspondence, with time as the horizontal axis and light intensity parameter as the vertical axis, wherein each ray in the light source group corresponds to a light intensity change curve.

[0036] Furthermore, the correction module includes:

[0037] The segmentation unit is used to segment the nutrient data into several individual items and create a fluctuation range that corresponds one-to-one with each individual item;

[0038] The activation unit is used to define the corresponding item as an anomaly when the nutrient data exceeds the fluctuation range, and to activate the pre-edited emergency response rules.

[0039] Furthermore, the aggregation module includes:

[0040] The setting unit is used to set the influencing factors of the initial weights, wherein the influencing factors include at least: planting location and growth status;

[0041] The adjustment unit is used to adjust the actual weight of each block based on the influencing factors.

[0042] Compared with the prior art, the beneficial effects of the present invention are:

[0043] By collecting and comparing the reflected signals of near-infrared and red light from each segment, the combined effects of some environmental factors can be offset, providing a clear view of the corn growth in each segment and timely identification of abnormal areas. This provides an effective data foundation for predicting corn growth. Creating topdressing tasks enables precise fertilization, avoiding over- or under-fertilization caused by uniform fertilization, thereby improving fertilizer utilization, reducing costs, and shifting corn topdressing from extensive to intensive methods. Collecting nutrient data allows monitoring of dynamic changes in nutrients within each segment, providing a clear assessment of whether the fertilization has achieved the expected results and improving the accuracy of corn topdressing. Generating a weighted distribution map transforms discrete segment monitoring results into a spatially continuous and intuitive representation, clearly reflecting the spatial distribution characteristics of nutrient deficiency within the entire monitoring area, thus providing reliable data support for topdressing management. Calculating reference values ​​quantifies the degree of nutrient deficiency in abnormal segments, further improving fertilization accuracy, avoiding deviations caused by uniform fertilization, and providing a reference for subsequent precise fertilization in the monitored area. Ultimately, this shift from experience-driven to precision fertilization effectively improves fertilizer utilization efficiency. Attached Figure Description

[0044] Figure 1 This is an example diagram of a weight distribution plot;

[0045] Figure 2A flowchart illustrating the corn topdressing management method based on image recognition provided in an embodiment of the present invention;

[0046] Figure 3 This is a first sub-flowchart of the corn topdressing management method based on image recognition provided in an embodiment of the present invention;

[0047] Figure 4 This is a second sub-flow flowchart of the corn topdressing management method based on image recognition provided in an embodiment of the present invention;

[0048] Figure 5 This is a third sub-flow flowchart of the corn topdressing management method based on image recognition provided in an embodiment of the present invention;

[0049] Figure 6 This is a fourth sub-flowchart of the corn topdressing management method based on image recognition provided in an embodiment of the present invention;

[0050] Figure 7 A block diagram illustrating the composition of the corn topdressing management system based on image recognition provided in an embodiment of the present invention;

[0051] Figure 8 This is a block diagram of the delineation module in the corn topdressing management system based on image recognition provided in an embodiment of the present invention;

[0052] Figure 9 A block diagram showing the composition of the correction module in the corn topdressing management system based on image recognition provided in an embodiment of the present invention;

[0053] Figure 10 A block diagram showing the composition of the summary module in the corn topdressing management system based on image recognition provided in an embodiment of the present invention;

[0054] Figure 11 This is a block diagram of the writing module in the corn topdressing management system based on image recognition provided in an embodiment of the present invention. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0056] In Example 1, Figure 1 and Figure 2 The implementation flow of the corn topdressing management method based on image recognition provided in this embodiment of the invention is illustrated below in detail:

[0057] S100: Delineate the monitoring area where corn topdressing is required, divide the monitoring area into several blocks, collect real-world images within the monitoring area, select target blocks based on preset evaluation rules, construct test scenarios, and collect multiple sets of reflection signals using light source groups and light intensity sensors pre-installed in the blocks. The light source groups include near-infrared light and red light, and each light ray corresponds to a set of reflection signals.

[0058] The area requiring monitoring of corn topdressing is determined, i.e., the monitoring area, which can be simply understood as the corn planting area. The monitoring area is divided into grids, defining multiple blocks. For example, the grid can be divided according to preset row and column spacing or fixed area units (such as 10m×10m or 20m×20m), or simply by dividing each acre of land into blocks. Real-world images of the monitoring area are collected using fixed cameras, drones, or manual photography. The collected images undergo preprocessing operations, including noise reduction, illumination correction, and geometric calibration, to extract key features of the blocks. These key features include leaf density, the presence of yellow leaves, and leaf curling. Based on evaluation rules, several target blocks are selected. The evaluation rules are mainly used to determine whether there are abnormalities in the corn plants within a block. For example, one evaluation rule is that if the number of yellow leaves detected in a block reaches or exceeds 3, that block is identified as a target block.

[0059] A test scenario is constructed, which refers to a reproducible data acquisition environment artificially created. In this embodiment, the test scenario is: a closed box structure is constructed at night to cover several corn plants. A light source group and a light intensity sensor are installed within this box structure. The light source group provides stable and controllable incident light (such as near-infrared and red light), and the light intensity sensor receives the light signals reflected back from the corn leaves in real time. During the actual test, the near-infrared light source is first activated, illuminating the surface of the corn plant leaves inside the box at a fixed angle and intensity, and the light intensity sensor simultaneously collects the reflected light signals. Then, the red light source is activated, and the reflected light signals from the corn plant leaves illuminated by the red light source are collected. Finally, the collected signals are repeatedly recorded, with each light source corresponding to one reflected signal.

[0060] In actual production, red light is strongly absorbed by chlorophyll in leaves and mainly participates in photosynthesis. Therefore, when corn is growing well and has a high chlorophyll content, the red light reflection signal is weak. Once there is nutrient deficiency, premature aging, or a decrease in chlorophyll, the leaves' ability to absorb red light weakens, and the reflection signal will significantly increase. In short, by analyzing the changing trend of the red light reflection signal, we can determine the changes in the nutritional status of corn plant leaves. Near-infrared light is basically not absorbed by chlorophyll. If the size, internal structure, water content, and leaf posture of corn plant leaves are constant, and without considering light source attenuation, the near-infrared light reflection signal remains constant. Therefore, the near-infrared light reflection signal can characterize the changes in related influencing factors. By comparing the reflection signals of near-infrared and red light, we can determine whether they have the same changing trend. If they do, it means that the change in the reflection signal mainly comes from the interference of related influencing factors. Conversely, if the changing trends are inconsistent, it can be considered that the internal physiological characteristics of corn leaves (such as chlorophyll content or nutrient levels) have changed, i.e., there are phenomena such as yellowing and leaf curling.

[0061] In actual testing scenarios, multiple light source groups and multiple wavelengths of light can be set inside the enclosure structure to further reduce interference from relevant influencing factors.

[0062] S200: Define the block with enhanced near-infrared light reflection signal and weakened red light reflection signal as an abnormal block, create a topdressing task, and send it to a preset terminal. Use the sensing devices deployed in the abnormal block to collect nutrient data and correct the topdressing task and the abnormal block.

[0063] The reflection signals of near-infrared and red light are compared and judged. When it is detected that the near-infrared light reflection signal in a certain block shows an increasing trend compared to historical data, but the red light reflection signal shows a decreasing trend, the block corresponding to the test scene is defined as an abnormal block. An abnormal block indicates that the corn plants in this area may have a decrease in chlorophyll content or insufficient nutrient supply. The spatial location and reflection signal of the abnormal block are written into a preset template to generate a topdressing task, which is then sent to a preset terminal, namely the terminal of the management personnel in the monitoring area.

[0064] Nutrient data in the soil is collected using sensors installed in the abnormal blocks. These sensors include nitrogen, phosphorus, and potassium sensors. A suitable growth range for corn plants is set for each data point. The abnormal block is then identified as such if the data falls within this suitable range; otherwise, it is removed. The amount of topdressing applied during the application process should also be determined based on the nutrient data.

[0065] S300: Draw a planar distribution map of the monitoring area, set the initial weights of the blocks, and mark them on the planar distribution map. Iterate through the neighborhood of each block in turn, determine whether there are abnormal blocks in the neighborhood, and if so, use the preset set value to offset the initial weight of the central block to obtain the true weight. Summarize the offset results of all blocks and generate a weight distribution map.

[0066] Based on the terrain and corn planting conditions of the monitoring area, a planar distribution map of the monitoring area is drawn. An initial weight is assigned to each block, representing the degree of nutrient deficiency in each block. This initial weight is then labeled onto the corresponding block. Each block is traversed, and a corresponding neighborhood is constructed centered on the current block. Within this neighborhood, adjacent blocks are individually examined to identify any anomalous blocks. If an anomalous block is detected in the neighborhood, it indicates that the nutrient status of the central block may be affected by abnormal distributions in the surrounding area. A set value is then added to the initial weight of this central block; this set value is determined by the monitoring area management personnel. The initial weights of all blocks are offset to obtain the true weights. The true weights corresponding to each block are then uniformly summarized and spatially mapped to generate a weight distribution map. This weight distribution map visually reflects the nutrient differences and spatial distribution characteristics of each block within the monitoring area. In short, if an anomalous block exists among the eight adjacent blocks of a given block, the set value is added to the initial weight of that block to obtain the true weight.

[0067] Included in the instruction manual Figure 1 For example, Figure 1 This is a weight distribution map where each grid represents a block, and all blocks have an initial weight of 1 (not all are labeled). X and Y are abnormal blocks, with a value of 2. The weight values ​​of blocks in the neighborhood of X are all increased by 2. However, due to the existence of Y, the weight values ​​of blocks in the overlapping neighborhoods of X and Y are increased by 2. The final accumulated result is the true weight. The larger the true weight, the higher the demand for nutrients in that block and the more urgent the need for fertilization.

[0068] S400: Collect the processing results of the topdressing task, wherein the processing results include at least: fertilizer application amount and growth changes, generate standard values, calculate the reference value of each abnormal block through the true weight, and write it into the corresponding block.

[0069] The processing results of each block are collected and summarized. The processing results include: the actual amount of fertilizer applied to the abnormal block and the changes in maize growth before and after fertilization. The collected multi-source data are comprehensively analyzed to generate standard values. The standard values ​​are used to characterize the standard nutrient supplementation level in the target area. The standard values ​​are multiplied by the true weights to obtain reference values. The reference values ​​are used to reflect the predicted amount of fertilizer applied to the abnormal block after comprehensively considering its nutrient deficit degree and fertilization effect. The reference values ​​are written into each grid in the weight distribution map.

[0070] In Example 2, Figure 3 The diagram shows a first sub-flowchart of the corn topdressing management method based on image recognition provided in this embodiment of the invention. The following details the steps of delineating the monitoring area for corn topdressing, dividing the monitoring area into several blocks, and collecting real-scene images within the monitoring area:

[0071] S101: Create a state recognition model to label risk features in real-world images, wherein the risk features include at least: stunted growth, yellowing leaves, and curled leaves.

[0072] Using deep learning algorithms, a state recognition model is constructed. This model can identify risk features in real-world images, including stunted growth, yellowing leaves, and curled leaves.

[0073] S102: Integrate all real-world images and risk features to generate a training set, and train the state recognition model.

[0074] In the generation stage of the state recognition model, the risk features in the real-world images are manually labeled by the monitoring area management personnel. The real-world images and the corresponding risk features are integrated to generate a training set, which is then used to train the state recognition model.

[0075] In Example 3, Figure 3 The diagram shows the first sub-process flow of the corn topdressing management method based on image recognition provided in an embodiment of the present invention. The following details the step of acquiring multiple sets of reflection signals using a light source group and light intensity sensor pre-installed in the blocks:

[0076] S103: Record the time corresponding to the test scene, extract the light intensity parameters in the reflected signal, and establish the correspondence between time and light intensity parameters.

[0077] Record the time of acquiring the reflected signal in each test scenario, extract the light intensity parameter from the reflected signal of each test scenario, and establish the correspondence between time and light intensity parameter.

[0078] S104: Based on the aforementioned correspondence, plot the light intensity variation curve with time as the horizontal axis and light intensity parameter as the vertical axis, wherein each ray in the light source group corresponds to a light intensity variation curve.

[0079] Plot the light intensity variation curve with time as the horizontal axis and the light intensity parameter at the corresponding time as the vertical axis. The light intensity variation curve can characterize the trend of light intensity change over time.

[0080] In Example 4, Figure 4 The second sub-process flowchart of the corn topdressing management method based on image recognition provided in an embodiment of the present invention is shown. The following details the steps of collecting nutrient data using sensing devices deployed in the anomaly block and correcting the topdressing task and the anomaly block:

[0081] S201: Divide the nutrient data into several individual items and create a fluctuation range that corresponds one-to-one with each individual item.

[0082] Nutrient data is divided into several individual items, such as nitrogen content, phosphorus content, and potassium content. Based on the changes in nutrient data under normal planting conditions, a corresponding fluctuation range is created for each individual item.

[0083] S202: When the nutrient data exceeds the fluctuation range, the corresponding item is defined as an anomaly, and the pre-edited emergency response rule is activated.

[0084] When the nutrient data for a particular item exceeds the corresponding fluctuation range, that item is determined to be an anomaly and recorded. At the same time, the pre-edited emergency response rules are triggered, which include: promptly reminding the management personnel in the monitoring area to adjust the amount of fertilizer applied.

[0085] In Example 5, Figure 5 The diagram shows the third sub-process flowchart of the corn topdressing management method based on image recognition provided in this embodiment of the invention. The steps of drawing the planar distribution map of the monitoring area and setting the initial weights of the blocks are described in detail below:

[0086] S301: Set the influencing factors for the initial weights, wherein the influencing factors include at least: planting location and growth status.

[0087] Identify the factors that may affect the amount of fertilizer applied in each section, i.e., the influencing factors, which include: planting location and growth status, etc.

[0088] S302: Based on the aforementioned influencing factors, adjust the actual weight of each block.

[0089] The actual weights should be adjusted based on the influencing factors for each segment, and these adjustments should be implemented by management personnel. For example, the actual weights can be increased for segments located at the edge of fields or those with insufficient irrigation cover.

[0090] In Example 6, Figure 6 The diagram shows the fourth sub-process flowchart of the corn topdressing management method based on image recognition provided in this embodiment of the invention. The following details the step of calculating the reference value of each abnormal block using the true weight and writing it into the corresponding block:

[0091] S401: Collect environmental data within the monitoring area and construct several triggering rules.

[0092] Collect environmental data within the monitoring area, including rainfall and temperature, and construct multiple triggering rules based on the environmental data.

[0093] S402: When the real-time value of the environmental data meets the triggering rule, update the abnormal block and the true weight.

[0094] Real-time environmental data is collected. When the real-time value meets a trigger rule, the anomaly block and the true weight are updated. For example, a trigger rule might be rainfall greater than 50 mm. When the rainfall in the monitored area exceeds 50 mm, the anomaly block and the true weight are redefined after the rainfall ends.

[0095] Figure 7 This diagram illustrates the structural block diagram of a corn topdressing system based on image recognition provided in an embodiment of the present invention. The corn topdressing management system 1 based on image recognition includes:

[0096] The delineation module 11 is used to delineate the monitoring area where corn topdressing is required, divide the monitoring area into several blocks, collect real-scene images within the monitoring area, select target blocks based on preset evaluation rules, construct test scenarios, and collect multiple sets of reflection signals using light source groups and light intensity sensors pre-installed in the blocks. The light source groups include near-infrared light and red light, and each light ray corresponds to a set of reflection signals.

[0097] The correction module 12 is used to define blocks that enhance near-infrared light reflection signals and weaken red light reflection signals as abnormal blocks, create topdressing tasks, and send them to preset terminals. It uses sensing devices deployed in the abnormal blocks to collect nutrient data and correct the topdressing tasks and abnormal blocks.

[0098] The summary module 13 is used to draw a planar distribution map of the monitoring area, set the initial weight of the blocks and mark them on the planar distribution map, traverse the neighborhood of each block in turn, determine whether there are abnormal blocks in the neighborhood, if so, use the preset set value to offset the initial weight of the central block to obtain the true weight, summarize the offset results of all blocks, and generate a weight distribution map.

[0099] The writing module 14 is used to collect the processing results of the topdressing task, wherein the processing results include at least: fertilizer application amount and growth change, generate standard values, calculate the reference value of each abnormal block through the real weight, and write it into the corresponding block.

[0100] Figure 8 This diagram illustrates the composition of the delineation module 11 in the corn topdressing management system based on image recognition provided in an embodiment of the present invention. The delineation module 11 includes:

[0101] The creation unit 111 is used to create a state recognition model and label risk features in the real-world image, wherein the risk features include at least: stunted growth, yellow leaves, and curled leaves;

[0102] Training unit 112 is used to integrate all real-world images and risk features, generate a training set, and train the state recognition model.

[0103] Recording unit 113 is used to record the time corresponding to the test scene, extract the light intensity parameter in the reflected signal, and establish the correspondence between time and light intensity parameter;

[0104] The drawing unit 114 is used to draw a light intensity change curve based on the correspondence, with time as the horizontal axis and light intensity parameter as the vertical axis, wherein each ray in the light source group corresponds to a light intensity change curve.

[0105] Figure 9 This diagram illustrates the structural composition of the correction module 12 in the corn topdressing management system based on image recognition provided in an embodiment of the present invention. The correction module 12 includes:

[0106] The segmentation unit 121 is used to segment the nutrient data into several individual items and create a fluctuation range that corresponds one-to-one with each individual item;

[0107] The activation unit 122 is used to define the corresponding single item as an abnormal item and activate the pre-edited emergency response rules when the nutrient data exceeds the fluctuation range.

[0108] Figure 10 This diagram illustrates the structural composition of the aggregation module 13 in the corn topdressing management system based on image recognition provided in an embodiment of the present invention. The aggregation module 13 includes:

[0109] Setting unit 131 is used to set the influencing factors of the initial weight, wherein the influencing factors include at least: planting location and growth status;

[0110] The adjustment unit 132 is used to adjust the actual weight of each block based on the influencing factors.

[0111] Figure 11 This diagram illustrates the structural composition of the writing module 14 in the corn topdressing management system based on image recognition provided in an embodiment of the present invention. The writing module 14 includes:

[0112] Construction unit 141 is used to collect environmental data within the monitoring area and construct several triggering rules;

[0113] The update unit 142 is used to update the abnormal block and the real weight when the real-time value of the environmental data meets the triggering rule.

[0114] The delineation module 11 is mainly used to complete step S100, the correction module 12 is mainly used to complete step S200, the summarization module 13 is mainly used to complete step S300, and the writing module 14 is mainly used to complete step S400.

[0115] The creation unit 111 is mainly used to complete step S101, the training unit 112 is mainly used to complete step S102, the recording unit 113 is mainly used to complete step S103, and the drawing unit 114 is mainly used to complete step S104.

[0116] The segmentation unit 121 is mainly used to complete step S201, and the activation unit 122 is mainly used to complete step S202.

[0117] The setting unit 131 is mainly used to complete step S301, and the adjustment unit 132 is mainly used to complete step S302.

[0118] The building unit 141 is mainly used to complete step S401, and the updating unit 142 is mainly used to complete step S402.

[0119] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0120] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these modifications and improvements all fall within the scope of protection of the present invention. Therefore, the scope of protection of this patent should be determined by the appended claims.

[0121] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A corn topdressing management method based on image recognition, characterized in that, The method includes: The monitoring area for corn topdressing is delineated, and the monitoring area is divided into several blocks. Real-world images of the monitoring area are collected. Based on the preset evaluation rules, target blocks are selected, test scenarios are constructed, and multiple sets of reflection signals are collected using light source groups and light intensity sensors pre-installed in the blocks. The light source groups include near-infrared light and red light, and each light ray corresponds to a set of reflection signals. Blocks with enhanced near-infrared light reflection signals and weakened red light reflection signals are defined as abnormal blocks. Topdressing tasks are created and sent to preset terminals. Nutrient data is collected using sensing devices deployed in the abnormal blocks, and the topdressing tasks and abnormal blocks are corrected. Draw a planar distribution map of the monitoring area, set the initial weights of the blocks, and mark them on the planar distribution map. Iterate through the neighborhood of each block in turn, determine whether there are abnormal blocks in the neighborhood, and if so, use the preset set value to offset the initial weight of the central block to obtain the true weight. Summarize the offset results of all blocks to generate a weight distribution map. The processing results of the topdressing task are collected, wherein the processing results include at least: fertilizer application amount and growth changes, standard values ​​are generated, reference values ​​for each abnormal block are calculated through the true weight, and written into the corresponding block.

2. The corn topdressing management method based on image recognition according to claim 1, characterized in that, The steps of delineating the monitoring area requiring corn topdressing, dividing the monitoring area into several blocks, and collecting real-scene images within the monitoring area include: Create a state recognition model to label risk features in real-world images, wherein the risk features include at least: stunted growth, yellowing leaves, and curled leaves; All real-world images and risk features are integrated to generate a training set, which is then used to train the state recognition model.

3. The corn topdressing management method based on image recognition according to claim 1, characterized in that, The step of acquiring multiple sets of reflection signals using light source groups and light intensity sensors pre-installed in the blocks includes: Record the time corresponding to the test scene, extract the light intensity parameters in the reflected signal, and establish the correspondence between time and light intensity parameters; Based on the aforementioned correspondence, a light intensity variation curve is plotted with time as the horizontal axis and light intensity parameter as the vertical axis, wherein each ray in the light source group corresponds to a light intensity variation curve.

4. The corn topdressing management method based on image recognition according to claim 1, characterized in that, The steps of collecting nutrient data using sensing devices deployed in the anomaly block and correcting the topdressing task and the anomaly block include: The nutrient data is divided into several individual items, and a fluctuation range corresponding to each individual item is created; When the nutrient data exceeds the fluctuation range, the corresponding item is defined as an anomaly, and the pre-edited emergency response rules are activated.

5. The corn topdressing management method based on image recognition according to claim 3, characterized in that, The steps of drawing a planar distribution map of the monitoring area and setting the initial weights of the blocks include: Factors influencing the setting of initial weights include, at a minimum, planting location and growth vigor. Based on the aforementioned influencing factors, the actual weight of each block is adjusted.

6. The corn topdressing management method based on image recognition according to claim 5, characterized in that, The step of calculating the reference value for each anomalous block using the true weight and writing it into the corresponding block includes: Collect environmental data within the monitoring area and construct several triggering rules; When the real-time value of the environmental data meets the triggering rules, the abnormal block and the true weight are updated.

7. A corn topdressing management system based on image recognition, characterized in that, The system includes: The delineation module is used to delineate the monitoring area where corn topdressing is required. The monitoring area is divided into several blocks, real-scene images of the monitoring area are collected, target blocks are selected based on preset evaluation rules, test scenarios are constructed, and multiple sets of reflection signals are collected using light source groups and light intensity sensors pre-installed in the blocks. The light source groups include near-infrared light and red light, and each light ray corresponds to a set of reflection signals. The correction module is used to define blocks that enhance near-infrared light reflection signals and weaken red light reflection signals as abnormal blocks, create topdressing tasks, and send them to preset terminals. The module uses sensing devices deployed in the abnormal blocks to collect nutrient data and correct the topdressing tasks and abnormal blocks. The summary module is used to draw a planar distribution map of the monitoring area, set the initial weight of the blocks and mark them on the planar distribution map, traverse the neighborhood of each block in turn, determine whether there are abnormal blocks in the neighborhood, if so, use the preset set value to offset the initial weight of the central block to obtain the true weight, summarize the offset results of all blocks, and generate a weight distribution map. The writing module is used to collect the processing results of the topdressing task, wherein the processing results include at least: fertilizer application amount and growth changes, generate standard values, calculate the reference value of each abnormal block through the real weight, and write it into the corresponding block.

8. The corn topdressing management system based on image recognition according to claim 7, characterized in that, The delineation module includes: A creation unit is used to create a state recognition model and label risk features in real-world images, wherein the risk features include at least: stunted growth, yellowing leaves, and curled leaves; The training unit is used to integrate all real-world images and risk features, generate a training set, and train the state recognition model. The recording unit is used to record the time corresponding to the test scene, extract the light intensity parameters in the reflected signal, and establish the correspondence between time and light intensity parameters. The drawing unit is used to draw a light intensity change curve based on the correspondence, with time as the horizontal axis and light intensity parameter as the vertical axis, wherein each ray in the light source group corresponds to a light intensity change curve.

9. The corn topdressing management system based on image recognition according to claim 7, characterized in that, The correction module includes: The segmentation unit is used to segment the nutrient data into several individual items and create a fluctuation range that corresponds one-to-one with each individual item; The activation unit is used to define the corresponding item as an anomaly when the nutrient data exceeds the fluctuation range, and to activate the pre-edited emergency response rules.

10. The corn topdressing management system based on image recognition according to claim 9, characterized in that, The aggregation module includes: The setting unit is used to set the influencing factors of the initial weights, wherein the influencing factors include at least: planting location and growth status; The adjustment unit is used to adjust the actual weight of each block based on the influencing factors.

Images