Intelligent evaluation method and operation system for water and fertilizer utilization efficiency

Abstract

The invention relates to the technical field of water and fertilizer application evaluation, and particularly discloses an intelligent evaluation method for water and fertilizer utilization efficiency and an operation system, and the method comprises the steps: obtaining a crop image of a crop corresponding to a fertilization parameter; identifying the crop image, outputting a growth evaluation report, and constructing a mapping relation from the fertilization parameters to the growth evaluation report for evaluating the fertilization parameters; fertilization parameters are recorded, meanwhile, crop images are obtained through a visual monitoring point, the crop images are recognized, the growth conditions before and after fertilization are determined, then the growth difference before and after fertilization is determined, and the mapping relation from the fertilization parameters to the growth difference is constructed; the growth difference of the fertilization parameters can be predicted before the fertilization behavior, and the prediction accuracy is high.

Description

Technical Field

[0001] The present invention relates to the technical field of water and fertilizer application evaluation, and in particular to an intelligent evaluation method and an operating system for water and fertilizer utilization efficiency. Background Art

[0002] Smart agricultural integrated water and fertilizer management is an agricultural management model that deeply integrates modern information technology with agricultural science. It uses relevant technical means to determine water and fertilizer demand information by comprehensively considering the soil fertility (such as the content of various nutrients in the soil, soil texture, etc.), the actual needs of crops at different growth stages (such as the different requirements for water and nutrients during the seedling stage, flowering stage, and fruiting stage), and meteorological conditions (including factors such as temperature, precipitation, and daylight duration). Under the existing technical background, the correspondence between various conditions and water and fertilizer demand information is determined through big data, and its effectiveness is very high.

[0003] However, in actual applications, staff often adjust the water and fertilizer application process according to actual conditions, and it is difficult for staff to know the effects of different water and fertilizer application information. Therefore, how to provide an evaluation scheme for the water and fertilizer application process to assist staff in adjusting the water and fertilizer application process is the technical problem that the technical solution of the present invention intends to solve. Summary of the Invention

[0004] The purpose of the present invention is to provide an intelligent evaluation method and operating system for water and fertilizer utilization efficiency to solve the problems raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions: An intelligent evaluation method for water and fertilizer utilization efficiency, the method comprising: Obtain crop distribution information and select visual monitoring points based on the crop distribution information; Record fertilization parameters in real time, generate acquisition instructions directed to visual monitoring points based on the fertilization parameters, and obtain crop images of crops corresponding to the fertilization parameters; the fertilization parameters include fertilization point, fertilization type, and fertilization amount; Identify the crop image to determine the crop vein, determine an initial inspection point based on the crop vein, and perform regional extension based on the initial inspection point to determine a detection area; The determined detection areas are screened to obtain a regional connection map, which is input into a trained neural network recognition model to output a growth evaluation report, and a mapping relationship between fertilization parameters and the growth evaluation report is constructed for evaluating fertilization parameters.

[0006] As a further solution of the present invention, the step of obtaining crop distribution information and selecting visual monitoring points according to the crop distribution information includes: Acquire remote sensing images of crop areas, identify them, and locate crops; Randomly select visual monitoring points in the crop area and simultaneously determine the collection direction of the visual monitoring points; the collection direction is horizontal, which uses a preset collection height and the direction with the largest number of crops within the collection wide angle; Eliminate the area corresponding to the harvested crops in the crop area to obtain an updated crop area; Randomly select visual monitoring points in the updated crop area, and simultaneously determine the collection direction of the visual monitoring points. Repeat this process until each crop is collected by at least one visual monitoring point, obtaining a selection plan. During the random selection process, the selection probability of each location in the crop area is determined by the distance between the location and the area center, and the selection probability is proportional to the distance. The loop is executed until the number of selected solutions reaches a preset threshold, and the solution with the least number of visual monitoring points is selected from all selected solutions as the final solution.

[0007] As a further solution of the present invention, the steps of recording fertilization parameters in real time, generating acquisition instructions directed to visual monitoring points according to the fertilization parameters, and obtaining crop images of crops corresponding to the fertilization parameters include: When fertilization behavior is monitored, record the fertilization point, type of fertilization and amount of fertilization; Determine the affected area based on the fertilization location, type and amount of fertilization; Query the crops in the affected area as collected crops; Query the visual monitoring points corresponding to the collected crops, generate collection instructions pointing to the visual monitoring points, and obtain crop images of the crops corresponding to the fertilization parameters.

[0008] As a further solution of the present invention, the steps of identifying the crop image, determining the crop vein, determining the initial inspection point based on the crop vein, and performing regional extension based on the initial inspection point to determine the inspection area include: Acquiring a planting time of the crop, and acquiring a standard image according to the planting time; Locating markers in the standard image and the crop image, and calibrating the standard image and the crop image based on the markers; the markers are preset image features; Based on the same differential step size, the difference matrices of the standard image and the crop image are calculated respectively, and the two difference matrices are traversed to determine and fit the crop vein fall; Determine an initial inspection point based on the crop vein when it is grounded, and perform regional extension based on the initial inspection point to determine a detection area; The calculation process of the difference matrix is ​​as follows: ; Wherein, the difference between the quadratic root of the sum of the squares of n and m and b is less than a preset threshold.

[0009] As a further solution of the present invention, the steps of determining the initial inspection point based on the crop vein being grounded, and performing regional extension based on the initial inspection point to determine the inspection area include: Traversing the pixel points in the crop vein according to a preset direction, and when there is no next pixel point within a preset angle range, marking the current pixel point as an initial inspection point; Taking the initial inspection point as the center, the color value difference of adjacent pixels is calculated in sequence in a preset direction; When the color value difference is less than a preset color value threshold, mark the corresponding pixel point; Counting the pixel points to determine the extension line; A center point is determined according to the extension line, and a detection area is determined according to the center point.

[0010] As a further solution of the present invention: the steps of determining the center point according to the extension line and determining the detection area according to the center point include: Obtaining the pixel length of each extension line, and determining the direction weight according to the pixel weight; Determine the direction of the center point according to the direction weight, and intercept the average length of the pixel lengths of all extended lines in the direction of the center point to determine the center point; Determine the area to be inspected based on the center point, and calculate the color value mean and standard deviation in the area to be inspected in real time; wherein the area size of the area to be inspected is an incremental value, and the incremental direction is determined by the difference between adjacent pixel points; The color value mean and standard deviation are compared with the preset numerical conditions to determine the detection area.

[0011] As a further solution of the present invention, the detection area determined by screening is obtained to obtain a regional connection map, the regional connection map is input into a trained neural network recognition model, a growth evaluation report is output, and a mapping relationship between fertilization parameters and the growth evaluation report is constructed. The steps for evaluating the fertilization parameters include: Inputting the detection area into a preset contour recognition model to determine the validity of the detection area; the validity includes invalidity and non-invalidity; Select the non-invalid detection area, input the preset statistical layer, and obtain the regional connection map; Inputting the regional connection map into a trained neural network recognition model and outputting a growth evaluation report; Read the growth evaluation reports before and after fertilization, calculate the growth difference, use it as the label of the water and fertilizer parameters corresponding to the crop, build a sample set, and train a water and fertilizer efficiency evaluation model based on the sample set to obtain the predicted growth difference of water and fertilizer parameters.

[0012] The technical solution of the present invention also provides an intelligent evaluation and operating system for water and fertilizer utilization efficiency, the system comprising: Monitoring point selection module, used to obtain crop distribution information and select visual monitoring points based on the crop distribution information; The crop image acquisition module is used to record fertilization parameters in real time, generate acquisition instructions directed to visual monitoring points based on the fertilization parameters, and acquire crop images of crops corresponding to the fertilization parameters; the fertilization parameters include fertilization point, fertilization type, and fertilization amount; a detection area determination module, configured to identify the crop image, determine the crop vein, determine an initial inspection point based on the crop vein, and perform regional extension based on the initial inspection point to determine a detection area; The fertilization parameter evaluation module is used to screen the determined detection area, obtain the regional connection map, input the regional connection map into the trained neural network recognition model, output the growth evaluation report, and construct a mapping relationship from the fertilization parameters to the growth evaluation report for evaluating the fertilization parameters.

[0013] As a further solution of the present invention: the monitoring point selection module includes: Crop positioning unit, used to obtain remote sensing images of crop areas, identify remote sensing images, and locate crops; A direction determination unit is used to randomly select visual monitoring points in the crop area and simultaneously determine the collection direction of the visual monitoring points; the collection direction is a horizontal direction, which uses a preset collection height and the direction with the largest number of crops within the collection wide angle; The region updating unit is used to remove the region corresponding to the harvested crops in the crop region to obtain an updated crop region; A selection scheme generating unit is configured to randomly select visual monitoring points in the updated crop area, simultaneously determine the collection direction of the visual monitoring points, and execute the process repeatedly until each crop is collected by at least one visual monitoring point, thereby obtaining a selection scheme. During the random selection process, the selection probability of each location in the crop area is determined by the distance between the location and the area center, and the selection probability is proportional to the distance. The selection scheme screening unit is used to execute cyclically until the number of selection schemes reaches a preset number threshold, and select the selection scheme with the least visual monitoring points among all the selection schemes as the final scheme.

[0014] As a further solution of the present invention: the crop image acquisition module includes: The behavior monitoring unit is used to record the fertilization point, fertilization type and fertilization amount when fertilization behavior is detected; An impact area determination unit is used to determine the impact area based on the fertilization point, fertilization type and fertilization amount; The crop query unit is used to query the crops in the affected area as collection crops; The instruction sending unit is used to query the visual monitoring point corresponding to the collected crop, generate a collection instruction pointing to the visual monitoring point, and obtain the crop image of the crop corresponding to the fertilization parameter.

[0015] Compared with the existing technology, the beneficial effects of the present invention are: the present invention records fertilization parameters, and at the same time, obtains crop images through visual monitoring points, identifies the crop images, determines the growth conditions before and after fertilization, and then determines the growth differences before and after fertilization, and constructs a mapping relationship from fertilization parameters to growth differences. When facing prediction needs, the growth differences of fertilization parameters can be predicted before fertilization, and the prediction accuracy is high. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention.

[0017] Figure 1 This is a flow chart of the intelligent evaluation method for water and fertilizer use efficiency.

[0018] Figure 2 This is a structural block diagram of the intelligent evaluation operating system for water and fertilizer utilization efficiency. DETAILED DESCRIPTION

[0019] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0020] Figure 1 This is a flowchart of an intelligent evaluation method for water and fertilizer utilization efficiency. In an embodiment of the present invention, an intelligent evaluation method for water and fertilizer utilization efficiency includes: Step S100: obtaining crop distribution information and selecting visual monitoring points according to the crop distribution information; Obtain crop distribution information within the water and fertilizer management area. Crop distribution information refers to which crops are located at each location. Visual monitoring points are selected based on the crop distribution information, and cameras are installed at the visual monitoring points to photograph the crops.

[0021] Step S200: Recording fertilization parameters in real time, generating acquisition instructions directed to visual monitoring points based on the fertilization parameters, and acquiring crop images of crops corresponding to the fertilization parameters; the fertilization parameters include fertilization points, fertilization types, and fertilization amounts; Each time fertilizer is applied, the fertilization parameters are recorded. The fertilization parameters are what fertilizer is applied at what location and how much fertilizer is applied. That is, the fertilization parameters include the fertilization point, fertilization type and fertilization amount. The relevant visual monitoring points are located according to the fertilization parameters, and acquisition instructions pointing to the visual monitoring points are generated. The image acquisition equipment at the visual monitoring points can obtain crop images. It is worth mentioning that the process of determining the relevant visual monitoring points is to determine the crops affected by the fertilization parameters, and then obtain the visual monitoring points that can capture the crops as the relevant visual monitoring points.

[0022] Step S300: Identifying the crop image, determining the crop vein, determining an initial inspection point based on the crop vein, and performing regional extension based on the initial inspection point to determine a detection area; By identifying the acquired crop images, crop veins can be determined; the crop veins are image features corresponding to the stems of the crop during its growth process; generally, crop veins during the flowering and fruiting stages are easier to obtain; the crop veins have a standard tree-like structure, with the tails of each branch as the initial inspection points, and then the area is extended with the initial inspection points as the center to obtain the detection area; the detection area generally corresponds to the fruiting products at a certain stage, such as the image area corresponding to the flowers during the flowering stage, and the image area corresponding to the fruits during the fruiting stage.

[0023] Step S400: Screening the determined detection areas to obtain a regional connection map, inputting the regional connection map into a trained neural network recognition model, outputting a growth evaluation report, and constructing a mapping relationship between fertilization parameters and the growth evaluation report for evaluating the fertilization parameters; There may be errors in the regional extension process and the accuracy is low. At this time, it is necessary to screen the detection area first, and then connect the screened detection areas to obtain a regional connection map. The regional connection map is a graph structure, which corresponds to the distribution of fruits. Similarly, if it is the flowering period, it is the distribution of flowers, and if it is the fruiting period, it is the distribution of fruits. The staff will create a sample set in advance and establish a mapping model (neural network recognition model) from the regional connection map to the growth evaluation report. When the regional connection map is obtained, the regional connection map is input into the mapping model to obtain the growth evaluation report. It should be noted that the environmental data and crop images both contain time tags, and each analysis process is carried out under the same time correspondence.

[0024] Regarding step S100, the steps of obtaining crop distribution information and selecting visual monitoring points according to the crop distribution information include: Acquire remote sensing images of crop areas, identify them, and locate crops; Randomly select visual monitoring points in the crop area and simultaneously determine the collection direction of the visual monitoring points; the collection direction is horizontal, which uses a preset collection height and the direction with the largest number of crops within the collection wide angle; Eliminate the area corresponding to the harvested crops in the crop area to obtain an updated crop area; Randomly select visual monitoring points in the updated crop area, and simultaneously determine the collection direction of the visual monitoring points. Repeat this process until each crop is collected by at least one visual monitoring point, obtaining a selection plan. During the random selection process, the selection probability of each location in the crop area is determined by the distance between the location and the area center, and the selection probability is proportional to the distance. The loop is executed until the number of selected solutions reaches a preset threshold, and the solution with the least number of visual monitoring points is selected from all selected solutions as the final solution.

[0025] In an example of the technical solution of the present invention, the process of selecting visual monitoring points is explained. Since the distribution of crops is generally regular, such as planting in rows with equal intervals, it is certainly feasible to manually set visual monitoring points, such as setting two in a row. However, this method is not necessarily the best method. It has a strong regularity, but the cost utilization rate may be low. In fact, only a limited number of cameras that can capture all crops are needed.

[0026] There are numerous combinations of cameras capable of capturing all crops, and many feasible solutions exist. However, acquiring images of all crops using the fewest number of visual monitoring points is a complex mathematical problem that is beyond the capabilities of those skilled in the art. In fact, crops are not necessarily planted at strictly equal spacing; spacing may change during growth, making it unrealistic to solve this mathematical problem. Therefore, the above content provides a relatively simple process for selecting the optimal solution, as follows: Acquire remote sensing images of the crop area, identify the remote sensing images, locate the crops, randomly select visual monitoring points in the crop area, and simultaneously determine the collection direction of the visual monitoring points. The collection direction is horizontal, and its height is a preset value, generally set at a height that can obtain the full view of the crops. The wide angle of the collection equipment at the visual monitoring point is also fixed, while the horizontal direction is a very wide range and can be set within a range of 360 degrees. The technical solution of the present invention determines the specific angle by using a preset collection height and the direction with the largest number of crops within the collection wide angle.

[0027] After selecting the visual monitoring point and determining its collection direction, query the crops it can collect, and eliminate the areas corresponding to the collected crops in the crop area (generally based on the outline of the crop, create equidistant entities according to the preset length. Of course, the circumscribed circle of the crop outline can also be obtained, which is determined by the staff independently) to obtain the updated crop area. For the updated crop area, the process of determining the visual monitoring point and its collection direction is executed again based on it, and then the crop area is updated. When each crop is collected by at least one visual monitoring point, all selected visual monitoring points are counted as a selection plan.

[0028] Consider the above content as a large loop. By executing the loop a preset number of times, a preset number of selection schemes can be obtained. Among these selection schemes, the selection scheme with the least visual monitoring points is selected as the final scheme. This is the optimization process. The obtained scheme is a better scheme, not necessarily the optimal scheme. In fact, although it is not necessarily the optimal scheme, it is good enough compared to the cost of investment, and the cost utilization rate is extremely high.

[0029] Step S200, the steps of recording fertilization parameters in real time, generating acquisition instructions directed to visual monitoring points according to the fertilization parameters, and acquiring crop images of crops corresponding to the fertilization parameters include: When fertilization behavior is monitored, record the fertilization point, type of fertilization and amount of fertilization; Determine the affected area based on the fertilization location, type and amount of fertilization; Query the crops in the affected area as collected crops; Query the visual monitoring points corresponding to the collected crops, generate collection instructions pointing to the visual monitoring points, and obtain crop images of the crops corresponding to the fertilization parameters.

[0030] Monitor fertilization behavior in real time. When fertilization behavior is detected, record the fertilization point, type and amount of fertilization. In fact, in the existing fertilization management architecture, fertilization behavior is directly uploaded by the staff. When the fertilization behavior uploaded by the staff is received, it is considered that the fertilization behavior has been monitored. Determine the affected area based on the fertilization point, type and amount of fertilization. The specific process is to determine a radius based on the fertilization type and amount of fertilization, and create a circular area with the fertilization point as the center, called the affected area; query the crops in the affected area as the collected crops, query the visual monitoring points corresponding to the collected crops, generate collection instructions pointing to the visual monitoring points, and obtain crop images of the crops corresponding to the fertilization parameters.

[0031] Regarding step S300, the steps of identifying the crop image, determining the crop vein, determining the initial inspection point based on the crop vein, and performing regional extension based on the initial inspection point to determine the inspection area include: Acquiring a planting time of the crop, and acquiring a standard image according to the planting time; Locating markers in the standard image and the crop image, and calibrating the standard image and the crop image based on the markers; the markers are preset image features; Based on the same differential step size, the difference matrices of the standard image and the crop image are calculated respectively, and the two difference matrices are traversed to determine and fit the crop vein fall; An initial inspection point is determined based on the crop vein being implemented, and a region is extended based on the initial inspection point to determine a detection area.

[0032] In one example of the technical solution of the present invention, the planting time of the crop is obtained, and a standard image is obtained based on the planting time. The standard image is known data, and markers in the standard image and the crop image are located. The standard image and the crop image are calibrated based on the markers. The markers are generally objects related to the crop, such as soil or a limiting frame (even soil is acceptable). The calibration process only calibrates the color value, which is used to convert the crop image into a crop image under standard lighting conditions (the lighting conditions of the standard image default to standard lighting conditions).

[0033] The same calculation process is used for standard images and crop images. The calculation process is to calculate the difference between pixels of adjacent steps. The difference reflects the degree of difference between each pixel and the surrounding pixels. The pixel points with the largest difference are selected to obtain the crop pulse. The description of the large difference needs to be determined by all the differences (difference matrix), such as the mean of all the differences. By analyzing the standard image and the crop image simultaneously, two results can be obtained. The two results can be verified with each other. The final crop pulse can be obtained by fitting the two results.

[0034] Finally, the crop veins are analyzed to determine the detection area.

[0035] The calculation process of the difference matrix is ​​as follows: ; Wherein, the difference between the quadratic root of the sum of the squares of n and m and b is less than a preset threshold.

[0036] In the above content, and In fact, included , they are calculated separately as boundary values; in the difference matrix obtained by the above process, each value represents the maximum difference between the corresponding pixel point and the surrounding pixel points, which can be compared to the concept of gradient.

[0037] Furthermore, the steps of determining an initial inspection point based on the crop vein being located, and performing regional extension based on the initial inspection point to determine the inspection area include: Traversing the pixel points in the crop vein according to a preset direction, and when there is no next pixel point within a preset angle range, marking the current pixel point as an initial inspection point; Read pixels in the crop vein in sequence according to the preset direction. The reading process is as follows: with the preset direction as the center line, determine an angle range. If a pixel exists within the angle range, read the next pixel. If no pixel exists within the angle range, traverse to the end and use the end point as the initial inspection point. Taking the initial inspection point as the center, the color value difference of adjacent pixels is calculated in sequence in a preset direction; This direction is different from the traversal direction. The traversal direction is generally a large direction, such as from bottom to top. The direction in the process of calculating color value difference is a direction group. For example, a direction group is determined with an angle interval of 30 degrees. In all directions in the direction group, the color value difference of adjacent pixels needs to be calculated.

[0038] When the color value difference is less than a preset color value threshold, mark the corresponding pixel point; Counting the pixel points to determine the extension line; When the color value difference is small, the corresponding pixel points are marked; based on the direction statistics of the marked pixel points, the extension line can be obtained.

[0039] A center point is determined according to the extension line, and a detection area is determined according to the center point.

[0040] The center point can be determined by analyzing the extension line, and the detection area can be obtained by executing the extension process with the center point as the starting point.

[0041] Specifically, the steps of determining a center point according to the extension line and determining a detection area according to the center point include: Obtaining the pixel length of each extension line, and determining the direction weight according to the pixel weight; Determine the direction of the center point according to the direction weight, and intercept the average length of the pixel lengths of all extended lines in the direction of the center point to determine the center point; Determine the area to be inspected based on the center point, and calculate the color value mean and standard deviation in the area to be inspected in real time; wherein the area size of the area to be inspected is an incremental value, and the incremental direction is determined by the difference between adjacent pixel points; The above content provides a center point determination process, which includes two stages: one is to determine the direction, and the other is to determine the position; specifically, the pixel length of each extension line is first calculated, and by comparing the pixel lengths, the influence of each extension line on the final center point can be determined, which is represented by the direction weight; according to the direction weight, on the one hand, the direction can be determined, and on the other hand, a mean length can be determined. By intercepting the mean length in the determined direction, the center point can be obtained.

[0042] Determine the area to be inspected based on the center point, and calculate the color value mean and standard deviation in the area to be inspected in real time; wherein the area size of the area to be inspected is an incremental value, and the incremental direction is determined by the difference between adjacent pixel points; Starting from the center point, the area is continuously expanded to the surrounding area. When the difference between adjacent pixels is less than the preset threshold, the extension is performed in that direction.

[0043] Comparing the color value mean and standard deviation with preset numerical conditions to determine the detection area; The goal of determining the detection area is to make the detection area correspond to the fruit (including the flower). Therefore, in the most appropriate detection area, the color values ​​of each pixel should be similar, that is, the standard deviation should be small. Based on this, during the extension process, the color value mean and standard deviation are continuously calculated, and the final detection area can be determined by comparing the standard deviation with the preset numerical conditions.

[0044] Regarding step S400, the detection area determined by screening is obtained to obtain a regional connection map, the regional connection map is input into the trained neural network recognition model, a growth evaluation report is output, and a mapping relationship between fertilization parameters and the growth evaluation report is constructed. The steps for evaluating fertilization parameters include: Inputting the detection area into a preset contour recognition model to determine the validity of the detection area; the validity includes invalidity and non-invalidity; The generation process of the detection area is an extension process, and its shape is not limited. Therefore, there is a high possibility that there will be invalid detection areas in the generated detection area, especially in the extension process starting from the initial inspection point. It is very likely that the entire pulse will be detected against the direction of the pulse. At this time, the detection area is invalid; the shapes of valid detection areas are limited, such as flowers and fruits.

[0045] Select the non-invalid detection area, input the preset statistical layer, and obtain the regional connection map; Select and connect the non-invalid detection areas, input the preset statistical layer, and you can get a connection diagram represented in the form of a layer.

[0046] Inputting the regional connection map into a trained neural network recognition model and outputting a growth evaluation report; With the help of the existing neural network recognition model, a mapping relationship between the regional connection map and the growth evaluation report can be established. When the regional connection map is obtained, the growth evaluation report can be directly output.

[0047] Read the growth evaluation reports before and after fertilization, calculate the growth difference, use it as the label of the water and fertilizer parameters corresponding to the crop, build a sample set, and train the water and fertilizer efficiency evaluation model based on the sample set to obtain the predicted growth difference of water and fertilizer parameters; Since the technical solution of the present invention ultimately aims to obtain the relationship between the fertilization process and the growth, water and fertilizer parameters are used as features, and the growth evaluation report is used as a label to construct a sample set. Based on the sample set, a water and fertilizer efficiency evaluation model is trained to obtain the predicted growth differences of water and fertilizer parameters. Therefore, when encountering a fertilization process that needs to be evaluated, the growth of water and fertilizer parameters can be predicted before fertilization to obtain the growth differences of crops with extremely high accuracy. Combined with the existing crop growth, the predicted status can be obtained.

[0048] It should be noted that the fertilization behavior occurs between time A and time B, and it corresponds to the difference from time A to time B. This is because the samples in the sample set are the growth differences between the water and fertilizer parameters of the fertilization behavior and before and after the fertilization behavior. In addition, the growth differences actually include the growth differences in different crop periods. The water and fertilizer efficiency evaluation model is essentially a clustering model, which clusters all growth differences corresponding to each water and fertilizer parameter into one category. In actual application, the most matching growth difference can be selected according to the current crop period.

[0049] Figure 2 This is a structural block diagram of an intelligent water and fertilizer utilization efficiency evaluation operating system. In an embodiment of the present invention, an intelligent water and fertilizer utilization efficiency evaluation operating system is provided. The system 10 includes: A monitoring point selection module 11 is used to obtain crop distribution information and select visual monitoring points based on the crop distribution information; The crop image acquisition module 12 is used to record fertilization parameters in real time, generate acquisition instructions directed to visual monitoring points based on the fertilization parameters, and acquire crop images of crops corresponding to the fertilization parameters; the fertilization parameters include fertilization point, fertilization type, and fertilization amount; A detection area determination module 13 is configured to identify the crop image, determine the crop vein, determine an initial inspection point based on the crop vein, and perform regional extension based on the initial inspection point to determine a detection area; The fertilization parameter evaluation module 14 is used to screen the determined detection area, obtain the regional connection map, input the regional connection map into the trained neural network recognition model, output the growth evaluation report, and construct a mapping relationship from the fertilization parameters to the growth evaluation report for evaluating the fertilization parameters.

[0050] Furthermore, the monitoring point selection module 11 includes: Crop positioning unit, used to obtain remote sensing images of crop areas, identify remote sensing images, and locate crops; A direction determination unit is used to randomly select visual monitoring points in the crop area and simultaneously determine the collection direction of the visual monitoring points; the collection direction is a horizontal direction, which uses a preset collection height and the direction with the largest number of crops within the collection wide angle; The region updating unit is used to remove the region corresponding to the harvested crops in the crop region to obtain an updated crop region; A selection scheme generating unit is configured to randomly select visual monitoring points in the updated crop area, simultaneously determine the collection direction of the visual monitoring points, and execute the process repeatedly until each crop is collected by at least one visual monitoring point, thereby obtaining a selection scheme. During the random selection process, the selection probability of each location in the crop area is determined by the distance between the location and the area center, and the selection probability is proportional to the distance. The selection scheme screening unit is used to execute cyclically until the number of selection schemes reaches a preset number threshold, and select the selection scheme with the least visual monitoring points among all the selection schemes as the final scheme.

[0051] Specifically, the crop image acquisition module 12 includes: The behavior monitoring unit is used to record the fertilization point, fertilization type and fertilization amount when fertilization behavior is detected; An impact area determination unit is used to determine the impact area based on the fertilization point, fertilization type and fertilization amount; The crop query unit is used to query the crops in the affected area as collection crops; The instruction sending unit is used to query the visual monitoring point corresponding to the collected crop, generate a collection instruction pointing to the visual monitoring point, and obtain the crop image of the crop corresponding to the fertilization parameter.

[0052] 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 in the scope of protection of the present invention.

Claims

1. An intelligent evaluation method for water and fertilizer utilization efficiency, characterized in that: The method comprises: Obtain crop distribution information and select visual monitoring points based on the crop distribution information; Record fertilization parameters in real time, generate acquisition instructions directed to visual monitoring points based on the fertilization parameters, and obtain crop images of crops corresponding to the fertilization parameters; the fertilization parameters include fertilization point, fertilization type, and fertilization amount; Identify the crop image to determine the crop vein, determine an initial inspection point based on the crop vein, and perform regional extension based on the initial inspection point to determine a detection area; The determined detection areas are screened to obtain a regional connection map, which is input into a trained neural network recognition model to output a growth evaluation report, and a mapping relationship between fertilization parameters and the growth evaluation report is constructed for evaluating fertilization parameters.

2. The intelligent evaluation method for water and fertilizer utilization efficiency according to claim 1, characterized in that: The steps of obtaining crop distribution information and selecting visual monitoring points according to the crop distribution information include: Acquire remote sensing images of crop areas, identify them, and locate crops; Randomly select visual monitoring points in the crop area and simultaneously determine the collection direction of the visual monitoring points; the collection direction is horizontal, which uses a preset collection height and the direction with the largest number of crops within the collection wide angle; Eliminate the area corresponding to the harvested crops in the crop area to obtain an updated crop area; Randomly select visual monitoring points in the updated crop area, and simultaneously determine the collection direction of the visual monitoring points. Repeat this process until each crop is collected by at least one visual monitoring point, obtaining a selection plan. During the random selection process, the selection probability of each location in the crop area is determined by the distance between the location and the area center, and the selection probability is proportional to the distance. The loop is executed until the number of selected solutions reaches a preset threshold, and the solution with the least number of visual monitoring points is selected from all selected solutions as the final solution.

3. The intelligent evaluation method for water and fertilizer utilization efficiency according to claim 1, characterized in that: The steps of recording fertilization parameters in real time, generating acquisition instructions directed to visual monitoring points according to the fertilization parameters, and obtaining crop images of crops corresponding to the fertilization parameters include: When fertilization behavior is monitored, record the fertilization point, type of fertilization and amount of fertilization; Determine the affected area based on the fertilization location, type and amount of fertilization; Query the crops in the affected area as collected crops; Query the visual monitoring points corresponding to the collected crops, generate collection instructions pointing to the visual monitoring points, and obtain crop images of the crops corresponding to the fertilization parameters.

4. The intelligent evaluation method for water and fertilizer utilization efficiency according to claim 3, characterized in that: The steps of identifying the crop image, determining the crop vein, determining the initial inspection point based on the crop vein, and performing regional extension based on the initial inspection point to determine the inspection area include: Acquiring a planting time of the crop, and acquiring a standard image according to the planting time; Locating markers in the standard image and the crop image, and calibrating the standard image and the crop image based on the markers; the markers are preset image features; Based on the same differential step size, the difference matrices of the standard image and the crop image are calculated respectively, and the two difference matrices are traversed to determine and fit the crop vein fall; Determine an initial inspection point based on the crop vein when it is grounded, and perform regional extension based on the initial inspection point to determine a detection area; The calculation process of the difference matrix is ​​as follows: ； Wherein, the difference between the quadratic root of the sum of the squares of n and m and b is less than a preset threshold.

5. The intelligent evaluation method for water and fertilizer utilization efficiency according to claim 4, characterized in that: The steps of determining the initial inspection point based on the crop vein being grounded, and performing regional extension based on the initial inspection point to determine the inspection area include: Traversing the pixel points in the crop vein according to a preset direction, and when there is no next pixel point within a preset angle range, marking the current pixel point as an initial inspection point; Taking the initial inspection point as the center, the color value difference of adjacent pixels is calculated in sequence in a preset direction; When the color value difference is less than a preset color value threshold, mark the corresponding pixel point; Counting the pixel points to determine the extension line; A center point is determined according to the extension line, and a detection area is determined according to the center point.

6. The intelligent evaluation method for water and fertilizer utilization efficiency according to claim 5, characterized in that: The steps of determining a center point according to the extension line and determining a detection area according to the center point include: Obtaining the pixel length of each extension line, and determining the direction weight according to the pixel weight; Determine the direction of the center point according to the direction weight, and intercept the average length of the pixel lengths of all extended lines in the direction of the center point to determine the center point; Determine the area to be inspected based on the center point, and calculate the color value mean and standard deviation in the area to be inspected in real time; wherein the area size of the area to be inspected is an incremental value, and the incremental direction is determined by the difference between adjacent pixel points; The color value mean and standard deviation are compared with the preset numerical conditions to determine the detection area.

7. The intelligent evaluation method for water and fertilizer utilization efficiency according to claim 1, characterized in that: The detection area determined by the screening is obtained to obtain a regional connection map, the regional connection map is input into the trained neural network recognition model, a growth evaluation report is output, and a mapping relationship between fertilization parameters and the growth evaluation report is constructed. The steps for evaluating the fertilization parameters include: Inputting the detection area into a preset contour recognition model to determine the validity of the detection area; the validity includes invalidity and non-invalidity; Select the non-invalid detection area, input the preset statistical layer, and obtain the regional connection map; Inputting the regional connection map into a trained neural network recognition model and outputting a growth evaluation report; Read the growth evaluation reports before and after fertilization, calculate the growth difference, use it as the label of the water and fertilizer parameters corresponding to the crop, build a sample set, and train a water and fertilizer efficiency evaluation model based on the sample set to obtain the predicted growth difference of water and fertilizer parameters.

8. An intelligent evaluation and operating system for water and fertilizer utilization efficiency, characterized in that: The system comprises: Monitoring point selection module, used to obtain crop distribution information and select visual monitoring points based on the crop distribution information; The crop image acquisition module is used to record fertilization parameters in real time, generate acquisition instructions directed to visual monitoring points based on the fertilization parameters, and acquire crop images of crops corresponding to the fertilization parameters; the fertilization parameters include fertilization point, fertilization type, and fertilization amount; a detection area determination module, configured to identify the crop image, determine the crop vein, determine an initial inspection point based on the crop vein, and perform regional extension based on the initial inspection point to determine a detection area; The fertilization parameter evaluation module is used to screen the determined detection area, obtain the regional connection map, input the regional connection map into the trained neural network recognition model, output the growth evaluation report, and construct a mapping relationship from the fertilization parameters to the growth evaluation report for evaluating the fertilization parameters.

9. The intelligent evaluation and operating system for water and fertilizer utilization efficiency according to claim 8, characterized in that: The monitoring point selection module includes: Crop positioning unit, used to obtain remote sensing images of crop areas, identify remote sensing images, and locate crops; A direction determination unit is used to randomly select visual monitoring points in the crop area and simultaneously determine the collection direction of the visual monitoring points; the collection direction is a horizontal direction, which uses a preset collection height and the direction with the largest number of crops within the collection wide angle; The region updating unit is used to remove the region corresponding to the harvested crops in the crop region to obtain an updated crop region; A selection scheme generating unit is configured to randomly select visual monitoring points in the updated crop area, simultaneously determine the collection direction of the visual monitoring points, and execute the process repeatedly until each crop is collected by at least one visual monitoring point, thereby obtaining a selection scheme. During the random selection process, the selection probability of each location in the crop area is determined by the distance between the location and the area center, and the selection probability is proportional to the distance. The selection scheme screening unit is used to execute cyclically until the number of selection schemes reaches a preset number threshold, and select the selection scheme with the least visual monitoring points among all the selection schemes as the final scheme.

10. The intelligent evaluation and operating system for water and fertilizer utilization efficiency according to claim 8, characterized in that: The crop image acquisition module includes: The behavior monitoring unit is used to record the fertilization point, fertilization type and fertilization amount when fertilization behavior is detected; An impact area determination unit is used to determine the impact area based on the fertilization point, fertilization type and fertilization amount; The crop query unit is used to query the crops in the affected area as collection crops; The instruction sending unit is used to query the visual monitoring point corresponding to the collected crop, generate a collection instruction pointing to the visual monitoring point, and obtain the crop image of the crop corresponding to the fertilization parameter.

Images