A method and device for predicting crop yield based on three-dimensional growth model
By constructing a crop yield prediction method based on a three-dimensional growth model, using two-dimensional photos and planting parameters for the whole growth period, the limitations of crop growth models in the prior art are solved, and the accuracy and versatility of yield prediction are improved.
Patent Information
- Application Number
- CN202411983769.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-31
- Publication Date
- 2025-04-08
AI Technical Summary
There are limitations in the construction and application of existing crop growth models, including limited sample square scale, large time span of field observation data acquisition, lack of accurate soil and crop data, and dependence on forecasting future weather, resulting in insufficient accuracy and universality of yield prediction.
A crop yield prediction method based on a three-dimensional growth model was adopted. By obtaining the two-dimensional RGB photo collection sequence of the whole growth period of the crop, a three-dimensional model sequence of the whole growth period of each individual was constructed, and combined with the whole growth period planting parameter group sequence, the target crop full growth period growth model was constructed for yield prediction.
The accuracy of crop yield prediction is improved, the time series data and three-dimensional phenotype data of the whole crop growth period are fully taken into account, the universality and accuracy of the model is enhanced, and the shortcomings of the existing technology are made up.
Smart Images
Figure CN119783908A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of yield prediction, and in particular to a crop yield prediction method and device based on a three-dimensional growth model. Background Art
[0002] The crop growth model is a digital simplified expression of the actual crop growth process. It is based on the crop, environment, and cultivation technology as a whole, and applies the principles and methods of system analysis to theoretically summarize and quantitatively analyze the physiological processes of crop growth and development, photosynthetic production, organogenesis, and yield formation, as well as their feedback relationship with the environment and technology. Then, the corresponding mathematical model is established to perform dynamic quantitative simulation of the crop growth process, so as to achieve the purpose of yield prediction and nutritional regulation. Obtaining growth monitoring data for the entire growth period of crops is a preliminary condition for building a crop growth model.
[0003] Although various crop growth models have been constructed, most of them are limited to controlled experiments at the sample scale in terms of model construction, and the time span of field observation data collection is relatively large. In addition, the acquisition of three-dimensional measured data of crops throughout the growth period as input data for training models is rarely used in the construction of crop growth models due to difficulties in collection and large amounts of data. In terms of model application, there are also certain limitations: first, the main factor determining changes in crop yields depends on the ability to predict future weather; second, there is a lack of accurate soil and crop data input, and there is no sufficiently accurate environmental data to run and verify the model; third, the model originated from a single-point experiment, and many assumptions are based on uniform production conditions in the field. In fact, multiple limiting factors in the farmland will work at the same time, reducing the versatility and accuracy of the model. These limitations directly affect the accuracy and versatility of the construction of crop growth models. Summary of the invention
[0004] The purpose of this application is to provide a crop yield prediction method and device based on a three-dimensional growth model, which can improve the accuracy of yield prediction.
[0005] To achieve the above objectives, this application provides the following solutions: In a first aspect, the present application provides a crop yield prediction method based on a three-dimensional growth model, comprising: Based on the 2D RGB photo set sequence of multiple individuals of the target crop during their entire growth period, a 3D model sequence of each individual during its entire growth period is constructed; Obtaining a planting parameter group sequence of multiple individuals of a target crop during the entire growth period; Based on the full growth period planting parameter group sequence and the full growth period three-dimensional model sequence of each individual, constructing a full growth period growth model of the target crop; Obtain the planting parameter group of the target crop to be tested in the current cycle; The planting parameter group of the target crop to be tested individual in the current cycle is input into the planting parameter group of the crop to be tested in the current cycle to obtain the yield prediction data of the individual to be tested.
[0006] Optionally, the step of constructing a three-dimensional model sequence of the full growth period of each individual according to a two-dimensional RGB photo set sequence of multiple individuals of the target crop during the full growth period includes: Get any variety of the target crop as the current variety; Determine any individual of the current species as the current individual; Let the number of iterations i = 1; Determine the two-dimensional RGB photo set corresponding to the i-th historical moment in the two-dimensional RGB photo set sequence of the entire reproductive period of the current individual as the current two-dimensional RGB photo set; Reconstruct the 3D model of the i-th historical moment based on the current 2D RGB photo set; Increase the value of the iteration number i by 1, and return to the step "determine the two-dimensional RGB photo set corresponding to the i-th historical moment in the two-dimensional RGB photo set sequence of the full growth period of the current individual as the current two-dimensional RGB photo set" until all historical moments in the full growth period of the individual are traversed to obtain the three-dimensional model sequence of the full growth period of the current individual; Update the current individual and return to step "set iteration number i=1" until all individuals in the current variety are traversed to obtain a three-dimensional model sequence of the entire growth period of different individuals of the same variety; Update the current variety and return to step "determine any individual of the current variety as the current individual" until all varieties are traversed to obtain a sequence of three-dimensional models of the entire growth period of individuals of different varieties.
[0007] Optionally, the current two-dimensional RGB photo set includes a plurality of two-dimensional RGB photos of the corresponding individual; The multiple two-dimensional RGB photos in the current two-dimensional RGB photo set all include corresponding individuals and identifiers set at the corresponding individuals; The position of the markers is the same when taking all 2D RGB photos; Different current two-dimensional RGB photos in the current two-dimensional RGB photo set correspond to different shooting angles; The overlap rate of two adjacent two-dimensional RGB photos is not less than a preset overlap rate.
[0008] Optionally, the logo is a Rubik's Cube; The preset overlap rate is 80%.
[0009] Optionally, reconstructing a 3D model of the i-th historical moment based on the current 2D RGB photo set includes: Extract the 3D point position in each 2D RGB photo; Use 3D point positions to register and stitch 2D RGB photos to generate a dense point cloud; Use normal vectors to perform ground correction on the crop dense point cloud to remove the influence of terrain; Correct the size of the crop dense point cloud based on the identification size; Remove soil background information and logo background information; De-noise the dense point cloud of crops to avoid interference from noise points; The RGB information and XYZ coordinates of the denoised dense point cloud are retained, and the result is exported as a PLY format point cloud to obtain the three-dimensional model at the i-th historical moment.
[0010] Optionally, any element in the full growth period planting parameter group sequence is a planting parameter group of the target crop at a corresponding historical moment; The planting parameter group includes: light parameters, temperature parameters, water parameters, fertilizer parameters, air parameters and farming method parameters; The light parameters include: solar radiation and albedo; The temperature parameters include: maximum temperature, minimum temperature, temperature difference and effective accumulated temperature; The water parameters include: precipitation, relative humidity, surface runoff, field capacity, soil moisture content, soil texture and soil drainage; The fertilizer parameters include: total phosphorus content, total nitrogen content, hydrolyzed nitrogen content, effective phosphorus content, available phosphorus content and available potassium content; The air parameters include: air pressure and wind speed; The farming method parameters include: sowing date, row spacing, irrigation date, irrigation amount, irrigation method, fertilizer type, fertilization date, fertilizer amount, fertilization method and harvest date.
[0011] Optionally, constructing a full growth period growth model of a target crop based on the full growth period planting parameter group sequence and the full growth period three-dimensional model sequence of each individual includes: A data set is constructed based on the individual's full growth period planting parameter group sequence and the full growth period three-dimensional model sequence; any data pair in the data set is the input and output of the same individual at the same historical moment; the input is the planting parameter group; the output includes a first yield data group and a second yield data group; the first yield data group is determined according to the three-dimensional model of the corresponding individual at the corresponding historical moment; The input quantity is used as input, and the output quantity is used as output, and the basic model is trained or fitted to obtain a growth model for the target crop throughout its growth period.
[0012] Optionally, the first yield data set includes: leaf structure parameters of the aboveground part and root structure parameters of the underground part; The second yield data set includes: leaf area index, organ biomass, potential yield, limiting yield, aboveground leaf yield, underground root yield and root element content.
[0013] Optionally, the basic model is a linear regression model, a machine learning model or a deep learning model.
[0014] In a second aspect, the present application provides a computer device, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-mentioned crop yield prediction method based on the three-dimensional growth model.
[0015] According to the specific embodiments provided in this application, this application discloses the following technical effects: The present application provides a crop yield prediction method and device based on a three-dimensional growth model, which is realized based on the reconstruction of the three-dimensional model of individual crops when real data of the entire growth period of the crops can be obtained, wherein the most critical factors are the real data of the entire growth period of the crops and the reconstruction of the three-dimensional model of the individual crops. A three-dimensional point cloud model of the individual crops is constructed by collecting two-dimensional photos of the individual crops; a crop growth model is constructed by collecting measured field data of the entire growth period of the crops, and the input, output and modeling method are debugged and optimized, so as to finally obtain a stable growth model suitable for the crop for crop yield prediction, fully consider the importance of real time series data and three-dimensional phenotypic data of the entire growth period of the crops, and comprehensively and reliably evaluate the crop growth model, thereby making up for the shortcomings of the existing crop yield prediction methods, and can more intuitively reflect the real situation of crop growth and effectively improve the prediction accuracy of the model. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0017] Figure 1 This is a flow chart of a crop yield prediction method based on a three-dimensional growth model in one embodiment of the present application.
[0018] Figure 2 This is a schematic diagram of the process of collecting two-dimensional images of beets in one embodiment of the present application.
[0019] Figure 3 Schematic diagram of the process of reconstructing a three-dimensional model of a beet individual in one embodiment of the present application.
[0020] Figure 4This is a schematic diagram of an individual three-dimensional model of the critical growth period of beet in one embodiment of the present application. DETAILED DESCRIPTION
[0021] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0022] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.
[0023] In an exemplary embodiment, Figure 1 As shown, a crop yield prediction method based on a three-dimensional growth model is provided, comprising: Step 101: Based on a sequence of two-dimensional RGB photographs of multiple individuals of a target crop during their entire growth period, a sequence of three-dimensional models of each individual during its entire growth period is constructed.
[0024] Step 101 includes: Step 101 - 1: Obtain any variety of the target crop as the current variety.
[0025] Step 101-2: Determine any individual of the current species as the current individual.
[0026] Step 101 - 3: Set the number of iterations i=1.
[0027] Step 101-4: Determine that the two-dimensional RGB photo set corresponding to the i-th historical moment in the sequence of two-dimensional RGB photo sets of the entire reproductive period of the current individual is the current two-dimensional RGB photo set. The current two-dimensional RGB photo set includes multiple two-dimensional RGB photos of the corresponding individual. The multiple two-dimensional RGB photos in the current two-dimensional RGB photo set all include the corresponding individual and the identifier set at the corresponding individual. The positions of the identifiers are the same when all two-dimensional RGB photos are taken. The shooting angles corresponding to different current two-dimensional RGB photos in the current two-dimensional RGB photo set are different. The overlap rate of two adjacent two-dimensional RGB photos is not less than the preset overlap rate. The identifier is a magic cube. The preset overlap rate is 80%.
[0028] Step 101-5: Reconstruct the three-dimensional model of the i-th historical moment based on the current two-dimensional RGB photo set. Step 101-5 includes: extracting the 3D point position in each two-dimensional RGB photo. Using the 3D point position to align and stitch the two-dimensional RGB photos, and then generate a dense point cloud. Use the normal vector to perform ground correction on the crop dense point cloud to remove the influence of the terrain. Correct the size of the crop dense point cloud based on the logo size. Remove background information such as soil and logos. De-noise the crop dense point cloud to avoid interference from noise. Keep the RGB information and XYZ coordinates of the dense point cloud, export the result as a PLY format point cloud, and obtain the three-dimensional model of the i-th historical moment.
[0029] Step 101-6: Increase the value of the iteration number i by 1, and return to step 101-4 until all historical moments in the individual's entire growth period are traversed to obtain the three-dimensional model sequence of the current individual's entire growth period.
[0030] Step 101-7: Update the current individual and return to step 101-3 until all individuals in the current variety are traversed to obtain a three-dimensional model sequence of the entire growth period of different individuals of the same variety.
[0031] Step 101-8: Update the current variety and return to step 101-2 until all varieties are traversed to obtain a sequence of three-dimensional models of individuals of different varieties throughout their growth period.
[0032] The process of collecting 2D photos of individual crops during the whole growth period is as follows: During the whole growth period of the crop, under the premise of conditions permitting, collect 2D RGB photos of individual crops every day. Collection standard: Before collecting images, first remove the debris around the individual crops to avoid affecting the operation. Place a magic cube next to the individual crops to serve as a calibration object to correct the size of the individual crops. Take two circles of photos around the individual crops. The first circle is tilted downward, controlled within the range of 0-90° from the top viewing angle, and the crop individuals are photographed 360° around; the second circle is tilted upward, controlled within the range of 90°-180° from the top viewing angle, and the crop individuals are also photographed 360° around. It is worth noting that the overlap rate between photos should be kept at no less than 80% when taking photos in motion. In other words, assuming that the field of view (FOA) of the photo is 30°, the number of photos taken in one circle of mobile photography cannot be less than 60. The significance of setting the overlap rate between photos to no less than 80% is to better stitch the 2D photos when reconstructing the 3D model.
[0033] The reconstruction process of the 3D model of individual crops is as follows: 3D reconstruction of individual crop 2D photos is based on the Structure From Motion (SFM) technology. First, the image features of the collected 2D photos are extracted to form photos, so as to estimate the position of 3D points. The 3D point positions are used to align and stitch the pictures, and further generate dense point clouds. The RGB information and XYZ coordinates of the dense point cloud are retained, and the result is exported as a PLY format point cloud. The reconstructed 3D point cloud of the crop is preprocessed: first, ground correction is performed, Python is used to obtain the normal vector of the plant vertical to the ground, and the orientation of the plant is corrected according to the rotation of the normal vector to remove the influence of the terrain; second, the size of the individual crop is corrected, that is, the point cloud is imported into ClondCompare, and the size of the reconstructed 3D point cloud of the crop is manually corrected using the magic cube to make it consistent with the actual plant size; third, background information is removed, and Python is used to remove the background such as soil and calibration object magic cube, and only the 3D point cloud of the plant part is retained; finally, denoising is performed, and the noise around the plant point cloud is removed, and only the 3D point cloud of the plant is retained as the complete 3D model of the individual crop after reconstruction.
[0034] Step 102: Obtain a sequence of planting parameter groups for the entire growth period of multiple individuals of the target crop. Any element in the sequence of planting parameter groups for the entire growth period is a planting parameter group for the target crop at a corresponding historical moment. The planting parameter group includes: light parameters, temperature parameters, water parameters, fertilizer parameters, air parameters and tillage method parameters. Light parameters include: solar radiation and albedo. Temperature parameters include: maximum temperature, minimum temperature, temperature difference and effective accumulated temperature. Water parameters include: precipitation, relative humidity, surface runoff, field water holding capacity, soil moisture content, soil texture and soil drainage. Fertilizer parameters include: total phosphorus content, total nitrogen content, hydrolyzed nitrogen content, effective phosphorus content, available phosphorus content and available potassium content. Air parameters include: air pressure and wind speed. Tillage method parameters include: sowing date, row spacing, irrigation date, irrigation amount, irrigation method, fertilizer type, fertilization date, fertilizer amount, fertilization method and harvest date.
[0035] Step 103: construct a full growth period growth model of the target crop based on the full growth period planting parameter group sequence and the full growth period three-dimensional model sequence of each individual.
[0036] Step 103 includes: constructing a data set based on the individual's full growth period planting parameter group sequence and full growth period three-dimensional model sequence. Any data pair in the data set is the input and output of the same individual at the same historical moment. The input is the planting parameter group. The output includes a first yield data group and a second yield data group. The first yield data group is determined based on the three-dimensional model of the corresponding individual at the corresponding historical moment. With the input as input and the output as output, the basic model is trained or fitted to obtain a full growth period growth model of the target crop.
[0037] The first yield data group includes: aboveground leaf structure parameters and underground root structure parameters. The second yield data group includes: leaf area index, organ biomass, potential yield, limited yield, aboveground leaf yield, underground root yield and root element content. The methods for determining the structural parameters of aboveground leaves and underground roots include the overall method and the layered method: the overall method uses the phenotypic parameters of the established crop individual model, such as length, height, width, convex hull area, and convex hull volume, as output; the layered method is to vertically equidistantly layer the established crop individual model, and extract the corresponding plant phenotypic parameters in each layer as output.
[0038] The process of building a growth model is divided into three levels: building growth models for the same individual at different periods, building growth models for different individuals of the same breed, and building growth models for different breeds. In order to ensure the accuracy and applicability of the model, the three processes are progressive. The method of building a growth model uses linear regression, machine learning, and deep learning methods. Taking linear regression as an example, the model construction formula is as follows: .
[0039] Among them, Y represents the output, , , and Represents the input quantity, , , , Represents coefficients (or weights). The method for establishing the relationship between input and output can be linear regression, machine learning, deep learning or other methods. The ratio of training and validation samples of the model is 7:3. The inputs, modeling methods and outputs listed above are not all used in the model, but are optimally selected in the construction of models for different varieties to obtain the most suitable variables and models for the variety. After selecting the inputs, modeling methods and outputs, the growth model is continuously optimized, and finally a relatively stable crop growth model is obtained.
[0040] Step 104: Obtain the planting parameter group of the target crop to be tested in the current cycle.
[0041] Step 105: Input the planting parameter group of the target crop to be tested in the current cycle into the planting parameter group of the current cycle of the crop to be tested to obtain the yield prediction data of the individual to be tested. According to the step-by-step modeling method, the prediction of crop yield is also divided into three levels: yield prediction of the same individual in different growth periods, yield prediction of different individuals of the same variety, and yield prediction of different varieties; by obtaining the measured data of crops in the whole growth period and constructing a three-dimensional model of crop individuals in the whole growth period, the accuracy of crop growth model construction and yield prediction is comprehensively and reliably evaluated, and this method is suitable for large-scale promotion. Specifically, the construction of the growth model of this application has two advantages: one is to use the measured data at intervals of 1 day during the whole growth period as model training, and the other is to use the three-dimensional point cloud training model reconstructed from the real scene. These two advantages are relatively lacking in the existing growth model, especially the time series data of the real scene of the whole growth period of the crop as the input training model, which will provide solid data support for the construction of an accurate crop growth model, which is of great significance to improving the model simulation and prediction accuracy of crops, improving the versatility of crop model migration and crop planting management decision-making.
[0042] As one of the main raw materials for sugar production, sugar beet is of great significance to the development of rural economy and animal husbandry. At present, how to increase the yield and sugar content has become the main problem of sugar beet production. During the growth cycle, the growth status of sugar beet is directly related to the yield and sugar content of sugar beet. Therefore, timely monitoring the growth status of sugar beet throughout the growth period and constructing its growth model are very important for accurate evaluation of sugar beet yield and sugar content. Below, taking sugar beet as an example, this embodiment is specifically described.
[0043] Through the three-dimensional surround photography technology, two-dimensional RGB images are collected to reconstruct the three-dimensional model of individual beets, and then the beet growth model for the entire growth period is constructed and the yield prediction is realized.
[0044] 1. Collection of two-dimensional photos of individual sugar beets throughout the growth period.
[0045] During the entire growth period of sugar beets (including the seedling stage, the rapid growth stage of leaf clusters, the tuber expansion stage and the sugar accumulation stage), if conditions permit, a two-dimensional RGB photo set of sugar beets is collected every other day.
[0046] Acquisition standard: Before acquiring images, first clear the debris around the individual beet to avoid affecting the operation. Place a Rubik's Cube next to the individual beet as a calibration object to calibrate the size of the individual beet. Take two circles around the individual beet ( Figure 2). The first circle is tilted downward, controlled within the top viewing angle range of 0-90°, and 360° surround tilt photography is performed on individual beets; the second circle is tilted upward, controlled within the top viewing angle range of 90°-180°, and 360° surround tilt photography is also performed on individual beets. It is worth noting that the overlap rate between photos must be kept at no less than 80% when taking mobile photos. In other words, assuming that the field of view (FOA) of the photo is 30°, the number of photos taken in one circle of mobile photography cannot be less than 60. The significance of setting the overlap rate between photos to no less than 80% is to be able to better stitch two-dimensional photos when reconstructing a three-dimensional model.
[0047] 2. Reconstruction of individual three-dimensional model of beets.
[0048] The principle of three-dimensional reconstruction of two-dimensional photos of individual beets is based on the Structure From Motion (SFM) technology. First, the image features of the collected two-dimensional photos are extracted to form photos, so as to estimate the positions of 3D points. The 3D point positions are used to align and stitch the pictures, and further generate dense point clouds. The RGB information and XYZ coordinates of the dense point cloud are retained, and the results are exported as PLY format point clouds. The reconstructed three-dimensional point cloud of the crop is preprocessed: first, ground correction is performed, Python is used to obtain the normal vector of the plant vertical to the ground, and the orientation of the plant is corrected according to the rotation of the normal vector to remove the influence of the terrain; second, the size of the individual crop is corrected, that is, the point cloud is imported into ClondCompare, and the size of the reconstructed three-dimensional point cloud of the crop is manually corrected using the magic cube to make it consistent with the actual plant size; third, background information is removed, and Python is used to remove the background such as soil and calibration object magic cube, and only the three-dimensional point cloud of the plant part is retained; finally, denoising is performed to remove the noise around the plant point cloud, and only the three-dimensional point cloud of the plant is retained as the complete three-dimensional model of the individual crop after reconstruction. The reconstruction process of the three-dimensional model of the individual beet is as follows: Figure 3 shown.
[0049] 3. Construction of the growth model of sugar beet throughout its entire growth period.
[0050] The two-dimensional photos of the whole growth period of beet collected every day in step 1 are reconstructed using the individual three-dimensional model reconstruction method of beet in step 2 to construct the growth model of the whole growth period of beet. Figure 4 .
[0051] The construction process of the beet growth model is divided into three levels: the construction of growth models for the same individual at different periods, the construction of growth models for different individuals of the same variety, and the construction of growth models for different varieties. In order to ensure the accuracy and applicability of the model, the three processes are progressive. Since the effective accumulated temperature of beets is related to their growth height, in the embodiment we only use the effective accumulated temperature as the input, and the canopy height of the leaves above the ground of beets as the output. The modeling method uses linear regression to construct and predict the growth model at three levels. The construction formula of the model is as follows: .
[0052] Where Y represents the overall canopy height of the aboveground leaves of beet, Represents the effective accumulated temperature of sugar beets.
[0053] 4. Forecast of sugar beet production.
[0054] According to the step-by-step modeling approach, the prediction of beet yield is also divided into three levels: yield prediction of the same individual at different growth periods, yield prediction of different individuals of the same variety, and yield prediction of different varieties.
[0055] In an exemplary embodiment, a computer device is provided, which may be a server or a terminal. The computer device includes a processor, a memory, an input / output interface (I / O for short) and a communication interface. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface is connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a crop yield prediction method based on a three-dimensional growth model is implemented.
[0056] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.
[0057] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.
[0058] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.
[0059] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM).
[0060] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.
[0061] The technical features of the above embodiments may be arbitrarily combined. To make the description concise, 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.
[0062] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.
Claims
1. A crop yield prediction method based on a three-dimensional growth model, characterized in that: include: Based on the 2D RGB photo set sequence of multiple individuals of the target crop during their entire growth period, a 3D model sequence of each individual during its entire growth period is constructed; Obtaining a planting parameter group sequence of multiple individuals of a target crop during the entire growth period; Based on the full growth period planting parameter group sequence and the full growth period three-dimensional model sequence of each individual, constructing a full growth period growth model of the target crop; Obtain the planting parameter group of the target crop to be tested in the current cycle; The planting parameter group of the target crop to be tested individual in the current cycle is input into the planting parameter group of the crop to be tested in the current cycle to obtain the yield prediction data of the individual to be tested.
2. The crop yield prediction method based on a three-dimensional growth model according to claim 1, characterized in that: The method of constructing a three-dimensional model sequence of the full growth period of each individual according to a two-dimensional RGB photo set sequence of multiple individuals of the target crop during the full growth period includes: Get any variety of the target crop as the current variety; Determine any individual of the current species as the current individual; Let the number of iterations i = 1; Determine the two-dimensional RGB photo set corresponding to the i-th historical moment in the two-dimensional RGB photo set sequence of the entire reproductive period of the current individual as the current two-dimensional RGB photo set; Reconstruct the 3D model of the i-th historical moment based on the current 2D RGB photo set; Increase the value of the iteration number i by 1, and return to step "determine the two-dimensional RGB photo set corresponding to the i-th historical moment in the two-dimensional RGB photo set sequence of the full growth period of the current individual as the current two-dimensional RGB photo set" until all historical moments in the full growth period of the individual are traversed to obtain the three-dimensional model sequence of the full growth period of the current individual; Update the current individual and return to step "set iteration number i=1" until all individuals in the current variety are traversed to obtain a three-dimensional model sequence of the entire growth period of different individuals of the same variety; Update the current variety and return to step "determine any individual of the current variety as the current individual" until all varieties are traversed to obtain a sequence of three-dimensional models of the entire growth period of individuals of different varieties.
3. The crop yield prediction method based on a three-dimensional growth model according to claim 2, characterized in that: The current two-dimensional RGB photo set includes a plurality of two-dimensional RGB photos of the corresponding individual; The multiple two-dimensional RGB photos in the current two-dimensional RGB photo set all include corresponding individuals and identifiers set at the corresponding individuals; The position of the markers is the same when taking all 2D RGB photos; Different current two-dimensional RGB photos in the current two-dimensional RGB photo set correspond to different shooting angles; The overlap rate of two adjacent two-dimensional RGB photos is not less than a preset overlap rate.
4. The crop yield prediction method based on a three-dimensional growth model according to claim 3, characterized in that: The logo is a Rubik's Cube; The preset overlap rate is 80%.
5. The crop yield prediction method based on a three-dimensional growth model according to claim 3, characterized in that: Reconstruct the 3D model of the i-th historical moment based on the current 2D RGB photo set, including: Extract the 3D point position in each 2D RGB photo; Use 3D point positions to register and stitch 2D RGB photos to generate a dense point cloud; Use normal vectors to perform ground correction on the crop dense point cloud to remove the influence of terrain; Correct the size of the crop dense point cloud based on the identification size; Remove soil background information and logo background information; De-noise the dense point cloud of crops to avoid interference from noise points; The RGB information and XYZ coordinates of the denoised dense point cloud are retained, and the result is exported as a PLY format point cloud to obtain the three-dimensional model at the i-th historical moment.
6. The crop yield prediction method based on a three-dimensional growth model according to claim 2, characterized in that: Any element in the full growth period planting parameter group sequence is a planting parameter group of the target crop at a corresponding historical moment; The planting parameter group includes: light parameters, temperature parameters, water parameters, fertilizer parameters, air parameters and farming method parameters; The light parameters include: solar radiation and albedo; The temperature parameters include: maximum temperature, minimum temperature, temperature difference and effective accumulated temperature; The water parameters include: precipitation, relative humidity, surface runoff, field capacity, soil moisture content, soil texture and soil drainage; The fertilizer parameters include: total phosphorus content, total nitrogen content, hydrolyzed nitrogen content, effective phosphorus content, available phosphorus content and available potassium content; The air parameters include: air pressure and wind speed; The farming method parameters include: sowing date, plant spacing, irrigation date, irrigation amount, irrigation method, fertilizer type, fertilization date, fertilizer amount, fertilization method and harvest date.
7. The crop yield prediction method based on a three-dimensional growth model according to claim 6, characterized in that: The method of constructing a full growth period growth model of a target crop based on the full growth period planting parameter group sequence and the full growth period three-dimensional model sequence of each individual includes: A data set is constructed based on the individual's full growth period planting parameter group sequence and the full growth period three-dimensional model sequence; any data pair in the data set is the input and output of the same individual at the same historical moment; the input is the planting parameter group; the output includes a first yield data group and a second yield data group; the first yield data group is determined according to the three-dimensional model of the corresponding individual at the corresponding historical moment; The input quantity is used as input, and the output quantity is used as output, and the basic model is trained or fitted to obtain a growth model for the target crop throughout its growth period.
8. The crop yield prediction method based on a three-dimensional growth model according to claim 7, characterized in that: The first yield data set includes: leaf structure parameters of the aboveground part and root structure parameters of the underground part; The second yield data set includes: leaf area index, organ biomass, potential yield, limiting yield, aboveground leaf yield, underground root yield and root element content.
9. The crop yield prediction method based on a three-dimensional growth model according to claim 8, characterized in that: The basic model is a linear regression model, a machine learning model or a deep learning model.
10. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the crop yield prediction method based on a three-dimensional growth model as described in any one of claims 1 to 9.