Plant growth monitoring model construction method, device and equipment
By acquiring and processing the multispectral images of the drone, determining the target irradiance range, and building a plant growth monitoring model, the problem of insufficient accuracy in existing technology under cloudy and overcast conditions is solved, and the High accuracy monitoring.
Patent Information
- Application Number
- CN202311676583.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-12-07
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2043-12-07
AI Technical Summary
The accuracy of the existing plant growth monitoring model under cloudy and overcast conditions cannot be guaranteed, and it depends heavily on weather conditions, which affects the reliability of plant growth monitoring.
By obtaining multi-spectral images of the target plant during each breeding period, determine the irradiance range for each breeding period, and select images within the target irradiance range for each breeding period during each breeding period. , using linear regression algorithm to build a growth monitoring model to reduce dependence on weather conditions.
The accuracy of the plant growth monitoring model under non-sunny weather conditions is improved, and the consistency of irradiance is ensured, thereby ensuring the accuracy and reliability of the model construction.
Smart Images

Figure CN117765385B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of plant growth monitoring, and in particular to a method, device and equipment for constructing a plant growth monitoring model. Background Art
[0002] The construction of a plant growth monitoring model helps to obtain timely and accurate plant growth information during production, thereby effectively guiding production management to achieve high yield, high efficiency and high quality of crop production.
[0003] The UAV multispectral remote sensing technology, which has the advantages of fast, non-destructive, accurate, flexible and low cost, has developed rapidly in the field of plant monitoring. Specifically, from 10 am to 2 pm on a sunny day, multispectral image data is collected by using a multispectral sensor carried by a UAV, and the corresponding vegetation index is extracted based on the collected multispectral image data. A plant growth monitoring model is constructed based on the extracted vegetation index and ground measured data to realize plant growth monitoring.
[0004] However, in the field environment, the weather is uncontrollable, and the data acquisition conditions of the above growth monitoring model limit the application of the growth monitoring model to a certain extent. That is, the construction of the above growth monitoring model is extremely dependent on the weather, and the accuracy of the growth monitoring model constructed under cloudy and overcast conditions cannot be guaranteed. Summary of the invention
[0005] The present application provides a method, device and equipment for constructing a plant growth monitoring model, so as to realize the construction of a growth monitoring model under the premise of reducing dependence on weather.
[0006] In a first aspect, the present application provides a method for constructing a plant growth monitoring model, comprising:
[0007] Step 1: Obtain multi-spectral images of the plant canopy of the target plant by drone in each growth period, and each multi-spectral image of the plant canopy by drone has corresponding irradiance;
[0008] Step 2, based on the irradiance corresponding to the multi-spectral images of the plant canopy taken by drones during each growth period, determine the irradiance range for each growth period;
[0009] Step 3, based on the irradiance range of each growth period, determine the target irradiance range, the target irradiance range is a range where the irradiance ranges of each growth period overlap, or the target irradiance range is a range where the distance between the irradiance range of each growth period and the target irradiance range is less than a preset distance;
[0010] Step 4, determining the corresponding plant drone canopy multispectral images whose irradiance is within the target irradiance range from the multi-stage plant drone canopy multispectral images in each growth period;
[0011] Step 5: Based on the multispectral images of the plant canopy taken by drones within the target irradiance range in each growth period and the measured growth indicators of the plants in each growth period, a growth monitoring model of the target plant is constructed through a linear regression algorithm.
[0012] In a possible implementation, step 1 includes:
[0013] During each growth period of the target plant, a multi-spectral image of the plant drone canopy is collected by a multi-spectral sensor carried by a drone, wherein each period of plant drone canopy multi-spectral image includes corresponding multiple groups of plant drone canopy multi-spectral images, each group of plant drone canopy multi-spectral images has the irradiance at the time of acquisition, and each group of plant drone canopy multi-spectral images represents a combination of original images of each single band.
[0014] In one possible implementation,
[0015] The irradiance of each set of plant drone canopy multispectral images is acquired by the downlink light sensor carried on the said drone and automatically stored in the raw image of each band;
[0016] The irradiance corresponding to each period of plant drone canopy multispectral imagery is obtained by taking the average of the irradiances of the corresponding multiple groups of plant drone canopy multispectral images, where the irradiance of each group of plant drone canopy multispectral images is represented by the irradiance of the original image in the near-infrared band.
[0017] In a possible implementation, the step 2 includes:
[0018] Determine the irradiance corresponding to each phase of multi-spectral drone canopy images of plants in each growth period;
[0019] The maximum and minimum irradiances were determined among the irradiances corresponding to the multi-spectral images of plant canopies taken by drones during each growth period.
[0020] Based on the maximum and minimum irradiances corresponding to the multi-spectral images of plant canopies taken by drones during multiple periods of each growth period, the irradiance range for each growth period is determined.
[0021] In a possible implementation, step 5 includes:
[0022] Perform image preprocessing on the multispectral image of the plant canopy by drone within the target irradiance range in each growth period to obtain the radiance image corresponding to each growth period;
[0023] Based on the radiance image corresponding to each growth period and the irradiance of the multispectral image of the plant drone canopy within the target irradiance range in each growth period, the reflectance image corresponding to each growth period is obtained;
[0024] Determine the vegetation index corresponding to each growth period based on the reflectance value in the reflectance image corresponding to each growth period;
[0025] Based on the vegetation index, measured leaf area index and measured canopy chlorophyll content corresponding to each growth period, a growth monitoring model of the target plant is constructed through a linear regression algorithm;
[0026] Among them, the reflectivity value of the pixel point in the reflectivity image is obtained by the following formula:
[0027] ρ(λ)=L(λ)*π / EDLS(λ),
[0028] Where λ represents the band, ρ(λ) represents the reflectance value of the pixel in the reflectance image of each band, and L(λ) represents the radiance of the pixel in the radiance image corresponding to each band (unit: W / (sr*m 2 )), EDLS(λ) represents the irradiance of each band of the multispectral image of the plant canopy within the target irradiance range in each growth period (unit: W / m 2 ).
[0029] In a possible implementation, the reflectivity value includes the reflectivity between the visible light band and the near infrared band;
[0030] The vegetation index includes at least one of the optimized soil conditioned vegetation index, the triangular vegetation index, the enhanced vegetation index, the red edge chlorophyll index, and the normalized difference red edge index;
[0031] The calculation formula of vegetation index is as follows:
[0032] OSAVI=(R NIR -R red ) / (R NIR +R red +0.1),
[0033] TVI=60(R NIR -R green )-100(R red -R redge ),
[0034] EVI=2.5*(R NIR -R red ) / (R NIR +6*R red -7.5*R blue+1),
[0035] CI=(R NIR / R redge )-1,
[0036] NDRE=(R NIR -R redge ) / (R NIR +R redge ),
[0037] Where OSAVI stands for Optimized Soil Adjusted Vegetation Index, TVI stands for Triangular Vegetation Index, EVI stands for Enhanced Vegetation Index, CI stands for Red Edge Chlorophyll Index, NDRE stands for Normalized Difference Red Edge Index; R NIR Represents the near-infrared reflectivity, R red Represents the red band reflectivity, R green Represents the green band reflectance, R redge Represents the red edge band reflectivity, R blue Represents the blue band reflectance.
[0038] In a possible implementation manner, the measured plant growth indicators in each growth period include the measured leaf area index and the measured canopy chlorophyll content.
[0039] In a second aspect, the present application provides a plant growth monitoring model construction device, comprising:
[0040] The acquisition module is used to obtain multi-period plant drone canopy multispectral images of the target plant in each growth period. Each period of plant drone canopy multispectral images has corresponding irradiance;
[0041] The first determination module is used to determine the irradiance range of each growth period based on the irradiance corresponding to the multi-spectral images of the plant canopy taken by drones during multiple growth periods;
[0042] A second determination module is used to determine a target irradiance range based on the irradiance range of each growth period, wherein the target irradiance range is a range in which the irradiance ranges of each growth period overlap, or the target irradiance range is a range in which the distance from the irradiance range of each growth period is less than a preset distance;
[0043] The third determination module is used to determine the corresponding plant drone canopy multispectral images whose irradiance is within the target irradiance range from the multi-stage plant drone canopy multispectral images in each growth period;
[0044] The processing module is used to construct a growth monitoring model of the target plant through a linear regression algorithm based on the multispectral image of the plant drone canopy within the target irradiance range in each growth period and the measured growth indicators of the plants in each growth period.
[0045] In a possible implementation manner, the acquisition module is specifically used to:
[0046] During each growth period of the target plant, a multi-spectral image of the plant drone canopy is collected by a multi-spectral sensor carried by a drone, wherein each period of plant drone canopy multi-spectral image includes corresponding multiple groups of plant drone canopy multi-spectral images, each group of plant drone canopy multi-spectral images has the irradiance at the time of acquisition, and each group of plant drone canopy multi-spectral images represents a combination of original images of each single band.
[0047] In one possible implementation,
[0048] The irradiance of each set of plant drone canopy multispectral images is acquired by the downlink light sensor carried on the said drone and automatically stored in the raw image of each band;
[0049] The irradiance corresponding to each period of plant drone canopy multispectral imagery is obtained by taking the average of the irradiances of the corresponding multiple groups of plant drone canopy multispectral images, where the irradiance of each group of plant drone canopy multispectral images is represented by the irradiance of the original image in the near-infrared band.
[0050] In a possible implementation manner, the first determining module is specifically configured to:
[0051] Determine the irradiance corresponding to each phase of multi-spectral drone canopy images of plants in each growth period;
[0052] The maximum and minimum irradiances were determined among the irradiances corresponding to the multi-spectral images of plant canopies taken by drones during each growth period.
[0053] Based on the maximum and minimum irradiances corresponding to the multi-spectral images of plant canopies taken by drones during multiple periods of each growth period, the irradiance range for each growth period is determined.
[0054] In a possible implementation manner, the processing module is specifically used to:
[0055] Perform image preprocessing on the multispectral image of the plant canopy by drone within the target irradiance range in each growth period to obtain the radiance image corresponding to each growth period;
[0056] Based on the radiance image corresponding to each growth period and the irradiance of the multispectral image of the plant drone canopy within the target irradiance range in each growth period, the reflectance image corresponding to each growth period is obtained;
[0057] Determine the vegetation index corresponding to each growth period based on the reflectance value in the reflectance image corresponding to each growth period;
[0058] Based on the vegetation index, measured leaf area index and measured canopy chlorophyll content corresponding to each growth period, a growth monitoring model of the target plant is constructed through a linear regression algorithm;
[0059] Among them, the reflectivity value of the pixel point in the reflectivity image is obtained by the following formula:
[0060] ρ(λ)=L(λ)*π / EDLS(λ),
[0061] Where λ represents the band, ρ(λ) represents the reflectance value of the pixel in the reflectance image of each band, and L(λ) represents the radiance of the pixel in the radiance image corresponding to each band (unit: W / (sr*m 2 )), EDLS(λ) represents the irradiance of each band of the multispectral image of the plant canopy within the target irradiance range in each growth period (unit: W / m 2 ).
[0062] In a possible implementation, the reflectivity value includes the reflectivity between the visible light band and the near infrared band;
[0063] The vegetation index includes at least one of the optimized soil conditioned vegetation index, the triangular vegetation index, the enhanced vegetation index, the red edge chlorophyll index, and the normalized difference red edge index;
[0064] The calculation formula of vegetation index is as follows:
[0065] OSAVI=(R NIR -R red ) / (R NIR +R red +0.1),
[0066] TVI=60(R NIR -R green )-100(R red -R redge ),
[0067] EVI=2.5*(R NIR -R red ) / (R NIR +6*R red -7.5*R blue+1),
[0068] CI=(R NIR / R redge )-1,
[0069] NDRE=(R NIR -R redge ) / (R NIR +R redge ),
[0070] Where OSAVI stands for Optimized Soil Adjusted Vegetation Index, TVI stands for Triangular Vegetation Index, EVI stands for Enhanced Vegetation Index, CI stands for Red Edge Chlorophyll Index, NDRE stands for Normalized Difference Red Edge Index; R NIR Represents the near-infrared reflectivity, R red Represents the red band reflectivity, R green Represents the green band reflectance, R redge Represents the red edge band reflectivity, R blue Represents the blue band reflectance.
[0071] In a possible implementation manner, the measured plant growth indicators in each growth period include the measured leaf area index and the measured canopy chlorophyll content.
[0072] In a third aspect, the present application provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for constructing a plant growth monitoring model as described in any one of the first aspects is implemented.
[0073] In a fourth aspect, the present application provides a non-transitory computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method for constructing a plant growth monitoring model as described in any one of the first aspects is implemented.
[0074] The method, device and equipment for constructing a plant growth monitoring model provided in the present application first obtain multi-period plant drone canopy multispectral images of the target plant in each growth period, and each period of the plant drone canopy multispectral images has a corresponding irradiance; based on the irradiance corresponding to the multi-period plant drone canopy multispectral images in each growth period, the irradiance range of each growth period can be determined, and then based on the irradiance range of each growth period, the target irradiance range is determined, and the target irradiance range is a range in which the irradiance ranges of each growth period overlap, or the target irradiance range is a range in which the distance between the irradiance range of each growth period is less than a preset distance; after determining the target irradiance range, the corresponding plant drone canopy multispectral images with irradiance within the target irradiance range are determined in the multi-period plant drone canopy multispectral images in each growth period, and based on the plant drone canopy multispectral images within the target irradiance range in each growth period, and the measured growth indicators of the plants in each growth period, a growth monitoring model of the target plant is constructed through a linear regression algorithm. The solution of the present application ensures that the irradiance of the subsequent multispectral images of the plant drone canopy used to construct the growth monitoring model is relatively close by determining the multispectral images of the plant drone canopy within the target irradiance range in each growth period. Since irradiance is an important factor affecting reflectivity, the construction of the growth monitoring model is realized based on the multispectral images of the plant drone canopy within the target irradiance range in each growth period, and the accuracy of the constructed growth monitoring model can be guaranteed even in non-sunny conditions. The solution of the present application reduces the dependence on weather conditions while also being able to more accurately realize the construction of the growth monitoring model. BRIEF DESCRIPTION OF THE DRAWINGS
[0075] In order to more clearly illustrate the technical solutions in the present application or the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0076] Figure 1 A flow chart of a method for constructing a plant growth monitoring model provided in an embodiment of the present application;
[0077] Figure 2 A schematic diagram of the mean and standard deviation of irradiance in each band in the data set provided in the embodiment of the present application;
[0078] Figure 3 A schematic diagram of the relationship between irradiance and reflectivity provided in an embodiment of the present application;
[0079] Figure 4A schematic diagram for comparing the growth monitoring models provided in the embodiments of the present application;
[0080] Figure 5 A schematic diagram of the structure of a plant growth monitoring model construction device provided in an embodiment of the present application;
[0081] Figure 6 A schematic diagram of the physical structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0082] In order to make the purpose, technical solutions and advantages of this application clearer, the technical solutions in this application will be clearly and completely described below in conjunction with the drawings in this application. Obviously, the described embodiments are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0083] Timely and accurate acquisition of plant growth information is the key to guiding production management and achieving high-yield, high-efficiency and high-quality plant production. Rapid, non-destructive and accurate remote sensing technology has great development prospects in plant growth monitoring.
[0084] UAV remote sensing has the advantages of strong flexibility, low cost, simple structure, high temporal and spatial resolution, and is gradually becoming an effective supplement to satellite remote sensing and ground remote sensing. Optical sensors carried on UAVs are widely used in the growth monitoring of plants such as corn, rice and wheat because they have rich spectral information combined with plant spectral characteristics.
[0085] Since optical sensors are sensitive to changes in solar incident irradiance, usually 10 a.m. to 2 p.m. on a sunny day with stable and strong incident irradiance is considered the best time to acquire multispectral images from drones. However, under field conditions, especially in the summer in central and northeastern my country, there is a lot of cloudy weather. During the critical growth period of plants, the lack of data or low quality will affect the accuracy of the construction of plant growth monitoring models.
[0086] Effective radiation correction methods can weaken the impact of incident irradiance changes on the accuracy of optical images. The radiation correction of multispectral remote sensing images of unmanned aerial vehicles mostly adopts the substitute site calibration method, also known as the standard ground object radiation correction method. This method is considered to be the most accurate under sunny conditions with high ambient light intensity and stable. The general steps are to use the reference target images of known reflectivity taken before and after the flight mission to establish the relationship between image radiance and reflectance, and then use this established relationship to perform radiation correction on all radiance images. However, the premise for such methods to obtain higher accuracy is that the light intensity during the mission remains unchanged or changes very little, and it is not applicable under cloudy or overcast conditions with large changes in light intensity and state. In recent years, with the improvement of the battery life of unmanned aerial vehicles, long-term and large-area unmanned aerial vehicle flight missions have become more and more popular. Even under sunny conditions, the light intensity changes greatly, and new methods are needed to adapt to the development of unmanned aerial vehicles and the monitoring needs of crops. Therefore, downlink light sensors that can record the solar incident irradiance in real time, are easy to carry and do not need to be repeatedly installed are widely used in unmanned aerial vehicle radiation correction.
[0087] However, the radiation-corrected plant reflectance varies on different dates and times of sunny days, indicating that there are other factors besides weather and time that affect the accuracy of drone multispectral data. However, in current research, there is no consensus on the impact of time and weather on drone multispectral data. How to solve the limitations of weather and time on optical sensors is an urgent problem to be solved in drone multispectral crop monitoring.
[0088] During the data processing, it was found that there is a correlation between solar incident irradiance and plant canopy reflectance, and further verification showed that there is indeed a very significant correlation between the two. Based on this, the embodiment of the present application provides a method for constructing a plant growth monitoring model, using whether the irradiance in different growth periods of the plant is consistent as a standard for whether the drone multispectral image can be used as a growth monitoring model. Compared with the method of constructing a growth monitoring model by collecting drone multispectral images from 10 am to 2 pm on a sunny day, in addition to reducing dependence on weather conditions, it also has higher accuracy. The scheme of the embodiment of the present application will be introduced below in conjunction with the accompanying drawings.
[0089] Figure 1 The flowchart of the method for constructing a plant growth monitoring model provided in the embodiment of the present application is as follows: Figure 1 As shown, the method may include:
[0090] S11, obtain multi-period plant drone canopy multispectral images of the target plant in each growth period, and each period of plant drone canopy multispectral images has corresponding irradiance.
[0091] The target plant is a plant that needs to be monitored for growth, for example, it may be corn or other possible plants, which is not limited in the embodiments of the present application.
[0092] The target plants have different growth periods, for example, they may include seedling stage, three-leaf stage, jointing stage, etc. For each growth period of the target plant, multi-period drone canopy multispectral images of the plant in the growth period can be obtained, and each period of drone canopy multispectral images of the plant has corresponding irradiance.
[0093] In a possible implementation, during each growth period of the target plant, a multi-spectral sensor mounted on a drone can be used to collect multiple periods of plant drone canopy multispectral images, wherein each period of plant drone canopy multispectral images includes corresponding multiple groups of plant drone canopy multispectral images, each group of plant drone canopy multispectral images has the irradiance at the time of acquisition, and each group of plant drone canopy multispectral images represents a combination of original images of each single band.
[0094] For each set of plant drone canopy multispectral images, the irradiance is acquired by the downlink light sensor mounted on the drone and automatically stored in the original image of each band. Specifically, the irradiance information of each set of plant drone canopy multispectral images is recorded by the downlink light sensor in the original image of each band. Through built-in information extraction, the irradiance value recorded by the downlink light sensor can be obtained.
[0095] The irradiance corresponding to each period of plant drone canopy multispectral images is obtained by taking the average of the irradiances of the corresponding multiple groups of plant drone canopy multispectral images, where the irradiance of each group of plant drone canopy multispectral images is represented by the irradiance of the original image in the near-infrared band. In a possible implementation, the irradiance corresponding to each period of plant drone canopy multispectral images can also be obtained by taking the average of the irradiances of the corresponding multiple groups of plant drone canopy multispectral images after removing outliers.
[0096] Multiple blocks can be pre-set for planting target plants, and each block has different planting parameters, such as nitrogen application amount. Taking corn as the target plant as an example, a corn nitrogen gradient experiment can be set to make the leaf area index and canopy chlorophyll content of corn planted in different blocks different, so as to analyze the accuracy and precision of the subsequent growth monitoring model based on the differences in the leaf area index and canopy chlorophyll content of corn in different blocks.
[0097] For any growth period among the multiple growth periods, multiple periods of plant drone canopy multispectral images can be obtained during the growth period of the target plant, and each period of plant drone canopy multispectral images includes plant drone canopy multispectral images corresponding to each regional block. For example, for any growth period, by using drone-mounted multispectral images, plant drone canopy multispectral images are collected at three different time periods each day for three consecutive days (i.e., the sampling day, the day before the sampling day, and the day after the sampling day) during the growth period, and 9 periods of plant drone canopy multispectral images within the growth period can be obtained. For any period of plant drone canopy multispectral images, including plant drone canopy multispectral images corresponding to each regional block collected by the drone-mounted multispectral sensor, each plant drone canopy multispectral image corresponding to each regional block has a corresponding irradiance, which is obtained by the downlink light sensor carried by the drone.
[0098] For each period of plant drone canopy multispectral images in each growth period, the drone canopy multispectral images of plants corresponding to each regional block are obtained by shooting multiple regional blocks with a drone-mounted multispectral sensor. It should be noted that in the process of shooting based on the drone-mounted multispectral sensor, it is necessary to shoot under the condition of satisfying the preset irradiance condition. That is, for any growth period, the drone canopy multispectral images of plants corresponding to each of the multiple regional blocks are collected under the condition of satisfying the preset irradiance condition. The preset irradiance condition may include a first preset irradiance condition and / or a second preset irradiance condition. Among them, the first preset irradiance condition refers to the condition that the irradiance change is less than the preset threshold, that is, the condition that the irradiance change is small. For example, it can be collected on a sunny day or on a cloudy day with relatively uniform clouds. When the first preset irradiance condition is met, the irradiance values of each regional block are slightly different. The second preset irradiance condition refers to the condition that the irradiance is not lower than the preset irradiance threshold, that is, the irradiance of the drone canopy multispectral images of plants corresponding to each of the multiple regional blocks collected is not lower than the preset irradiance threshold. For example, try not to collect plant drone canopy multispectral images at dusk, rainy weather, or other times when the irradiance is significantly lower, so as to ensure that the irradiance of the collected plant drone canopy multispectral images is greater than the preset irradiance threshold. The preset irradiance threshold can be set as needed, for example, it can be set to 0.5W / m 2 , 0.6W / m 2 Etc., the embodiments of the present application are not limited to this.
[0099] Optionally, during the process of shooting with a drone equipped with a multispectral sensor, the drone flight parameters and the multispectral sensor parameters may be pre-set, and the drone flight parameters and the multispectral sensor parameters remain unchanged during different growth periods.
[0100] Optionally, for any growth period, the time period for the drone to carry out shooting with a multispectral sensor can include multiple different time periods, for example, it can include several different time periods such as 10 a.m., 12 p.m., and 2 p.m. The drone can be equipped with a multispectral sensor to take pictures on days with different weather conditions, such as but not limited to cloudy, overcast, sunny, etc. By collecting multispectral images of plant drone canopies in different time periods and under different weather conditions, the data used to build the growth monitoring model can be enriched.
[0101] In the process of shooting with a multispectral sensor mounted on a drone, multiple sets of original plant drone canopy multispectral images may be taken at one time. For any area block in multiple area blocks, one set can be selected from the multiple sets of original plant drone canopy multispectral images as the initial plant drone canopy multispectral image corresponding to the area block. Among them, it needs to be satisfied that the initial plant drone canopy multispectral image corresponding to the area block includes the area block, and the area block is in a non-edge position in the corresponding initial plant drone canopy multispectral image.
[0102] For any area block among the multiple area blocks, after determining the initial plant drone canopy multispectral image corresponding to the area block, the area of interest can be selected, and the area of interest is selected based on a single area block.
[0103] Specifically, for any area block among the multiple area blocks, the initial plant drone canopy multispectral image corresponding to the area block can be intercepted according to the position of the area block in the corresponding initial plant drone canopy multispectral image, so as to obtain the plant drone canopy multispectral image corresponding to the area block. The plant drone canopy multispectral image corresponding to the area block is the part of the initial plant drone canopy multispectral image corresponding to the area block that includes the area block.
[0104] In the above embodiment, the process of how to obtain the multispectral image of the plant drone canopy corresponding to any area block is introduced. Based on the above implementation method, the multispectral image of the plant drone canopy corresponding to each of the multiple area blocks in any growth period can be obtained.
[0105] For any growth period among the multiple growth periods, the multi-period plant drone canopy multispectral images within the growth period all have irradiance when they are acquired, and the irradiance of each period of plant drone canopy multispectral images can be determined based on the irradiance of plant drone canopy multispectral images corresponding to each of the multiple regional blocks in each period. Specifically, the irradiance of each period of plant drone canopy multispectral images can be determined based on the average value between the irradiances of plant drone canopy multispectral images corresponding to each of the multiple regional blocks in each period. For example, the average value between the irradiances of plant drone canopy multispectral images corresponding to each of the multiple regional blocks in each period can be directly determined as the irradiance of plant drone canopy multispectral images in each period; for example, the average value between the irradiances of plant drone canopy multispectral images corresponding to each of the multiple regional blocks in each period can be retained to one or two decimal places, etc., to obtain the irradiance of each period of plant drone canopy multispectral images, and so on.
[0106] Optionally, in an embodiment of the present application, each area block corresponds to a multispectral image of a plant drone canopy, and each group of images includes multiple images of different bands. The irradiance corresponding to the image of each band may be the same or different. The irradiance corresponding to the multispectral image of the plant drone canopy can be jointly determined based on the irradiance corresponding to the images of these multiple different bands. It is also possible to use the irradiance of the image of the near-infrared band as the irradiance corresponding to the multispectral image of the plant drone canopy among the images of these multiple bands. In the visible-near infrared region, the near-infrared band plays an important role in vegetation remote sensing. Due to the influence of the internal structure of the leaves, the reflection of vegetation in the near-infrared region is extremely obvious. The near-infrared band has the greatest impact on the monitoring effect of plants. Therefore, the irradiance of the image of the near-infrared band can be used as the irradiance corresponding to the multispectral image of the plant drone canopy.
[0107] S12, based on the irradiance corresponding to the multi-spectral images of the plant canopy taken by drones during multiple periods in each growth period, determine the irradiance range for each growth period.
[0108] First, the irradiance corresponding to each phase of the multi-spectral images of the plant drone canopy in each growth period is determined. The specific implementation method can refer to the relevant introduction in the above embodiment.
[0109] After obtaining the irradiance corresponding to each plant drone canopy multispectral image in each growth period, the irradiance range of each growth period can be determined. For any growth period, the irradiance corresponding to each plant drone canopy multispectral image in that growth period is within the irradiance range of that growth period.
[0110] Specifically, the maximum irradiance and the minimum irradiance are determined in the irradiances corresponding to the multi-spectral images of the plant drone canopy in each growth period, wherein for any growth period, the maximum irradiance is the maximum value of the irradiances corresponding to the multi-spectral images of the plant drone canopy in the growth period, and the minimum irradiance is the minimum value of the irradiances corresponding to the multi-spectral images of the plant drone canopy in the growth period. Then, based on the maximum irradiance and the minimum irradiance of the irradiances corresponding to the multi-spectral images of the plant drone canopy in each growth period, the irradiance range of each growth period is determined. Among them, the irradiance range of each growth period is a range with the corresponding maximum irradiance and minimum irradiance as endpoints.
[0111] S13, based on the irradiance range of each growth period, determine the target irradiance range, the target irradiance range is a range where the irradiance ranges of each growth period overlap, or the target irradiance range is a range where the distance between the irradiance range of each growth period and the target irradiance range is less than a preset distance.
[0112] The target irradiance range is determined based on the irradiance ranges of each growth period. In a possible implementation, the irradiance ranges of multiple growth periods may be intersected to obtain the target irradiance range, which is the range where the irradiance ranges of each growth period intersect.
[0113] In a possible implementation, a range in which the distance from the irradiance range of each growth period is less than a preset distance can be determined and used as the target irradiance range. For the irradiance range of any growth period, the distance between the irradiance range of the growth period and the target irradiance range may include the distance between the left endpoint of the irradiance range of the growth period and the left endpoint of the target irradiance range, and may also include the distance between the right endpoint of the irradiance range of the growth period and the right endpoint of the target irradiance range.
[0114] S14, determining the corresponding plant drone canopy multispectral images whose irradiance is within the target irradiance range from the multi-stage plant drone canopy multispectral images in each growth period.
[0115] After obtaining the target irradiance range, for each growth period, a corresponding plant drone canopy multispectral image with an irradiance within the target irradiance range can be determined from the multi-period plant drone canopy multispectral images within the growth period. The plant drone canopy multispectral images with irradiance within the target irradiance range in each growth period are used to reflect the irradiance of the corresponding growth period. Since the irradiance of each period of the plant drone canopy multispectral images finally determined is within the target irradiance range, the irradiance consistency between each growth period is guaranteed.
[0116] S15, based on the multispectral images of the plant canopy taken by drones within the target irradiance range in each growth period and the measured growth indicators of plants in each growth period, a growth monitoring model of the target plants was constructed through a linear regression algorithm.
[0117] The measured growth indicators of plants in each growth period include the measured leaf area index and the measured canopy chlorophyll content. The measured leaf area index and the measured canopy chlorophyll content are relevant parameters of the target plants in the key growth period. These parameters need to be combined to jointly construct the subsequent growth monitoring model.
[0118] The following is an example of determining the leaf area index and canopy chlorophyll content corresponding to the target plant in multiple area blocks during any growth period.
[0119] For any of the multiple blocks, multiple sampling points can be selected in the block, a certain number of target plants are selected at each sampling point, and then the leaf area coefficient method is used to measure the leaf area of the target plants, thereby calculating the leaf area index. The canopy chlorophyll content can be calculated using the spectrophotometer method.
[0120] For any target plant in the area block, the measured leaf area index of the target plant can be calculated by the following formula (1):
[0121] LAI=S1 / S2 (1)
[0122] Wherein, LAI represents the measured leaf area index of the target plant, S1 represents the total leaf area of the target plant, and S2 represents the area occupied by the target plant.
[0123] The total leaf area of the target plant is equal to the sum of the areas of all the leaves on the target plant. For any leaf on the target plant, the leaf area coefficient method can be used to obtain the leaf area of the leaf. There are two cases: if the leaf is an expanded leaf, the expanded leaf area = leaf length * leaf width * 0.75; if the leaf is an unexpanded leaf, the unexpanded leaf area = leaf length * leaf width * 0.5. By measuring the length (i.e., the length from the leaf base to the leaf tip) and width (the width of the widest part of the leaf), the total leaf area of the target plant can be obtained.
[0124] The area occupied by the target plant can be obtained by dividing the area of the region block to which the target plant belongs and the number of target plants included in the region block. After obtaining the total leaf area and the area occupied by the target plant, the leaf area index of the target plant can be obtained.
[0125] Then, for any area block, after sampling multiple target plants in the area block, the measured leaf area index of each of the multiple target plants is averaged after removing the outliers, and the measured leaf area index corresponding to the target plants in the area block can be obtained. The measured leaf area index in each growth period can be obtained by averaging the measured leaf area index corresponding to the target plants in multiple area blocks.
[0126] For any area block among the multiple area blocks, multiple sampling points can be selected in the area block, a certain number of target plants are selected at each sampling point, and then the canopy chlorophyll content is calculated using a spectrophotometer method.
[0127] The determination of canopy chlorophyll content starts with routine sampling. For any target plant in the area block, each fresh leaf of the target plant is placed in a fresh-keeping bag. When determining the pigment content, the leaf veins are removed, and n leaf discs are punched out with a 0.5 mm diameter puncher. The discs are divided into 3 replicates, and 6 small discs are placed in each replicate. The discs are placed in 15 ml of 95% ethanol extract, and then placed in a dark place for 48 hours until all the leaves turn white. The absorbance of the leaf extract at 645 nm, 663 nm and 470 nm is measured with a UV-visible spectrophotometer, and then the chlorophyll concentrations of chlorophyll a (Chla) and chlorophyll b (Chla) in the leaves are calculated. Then, the chlorophyll content (LCC, μg / cm 2 ), and finally the canopy chlorophyll content of the target plant is obtained.
[0128] The specific calculation method can be found in the following formula (2) to formula (4):
[0129]
[0130] LCC=(chla+b)*VT / S5 (3)
[0131] chla+b=8.04*A663+20.29*A645 (4)
[0132] Among them, CCC (unit g / m 2 ) represents the canopy chlorophyll content of the target plant, n represents the number of leaf discs punched by the puncher, n is a positive integer (e.g., 18), LCC (unit: μg / cm 2 ) is the leaf chlorophyll content, S3 is the leaf area, S4 is the unit land area, chla+b (unit mg / l) is the chlorophyll concentration, VT (unit ml) is the volume of leaf extract, S5 (unit cm 2 ) is the area of the leaf disc, A663 and A645 represent the absorbance values of the leaf extract of the target plant at 663 nm and 645 nm, respectively.
[0133] For any area block, after sampling multiple target plants in the area block, the canopy chlorophyll content of each target plant is averaged after removing outliers, and the measured canopy chlorophyll content corresponding to the target plant in the area block can be obtained. The measured canopy chlorophyll content in each growth period can be obtained by averaging the measured canopy chlorophyll content corresponding to the target plants in multiple area blocks.
[0134] After obtaining the measured leaf area index and the measured canopy chlorophyll content, it is also necessary to obtain the vegetation index corresponding to each growth period in order to construct a growth monitoring model.
[0135] First, the multispectral image of the plant drone canopy within the target irradiance range in each growth period is preprocessed to obtain the radiance image corresponding to each growth period. Specifically, the multispectral image of the plant drone canopy within the target irradiance range in each growth period can be preprocessed based on the image preprocessing function (for example, the raw_image_to_radiance function in the MicaSense module) to obtain the radiance image corresponding to each growth period, and the pixel value of each pixel in the radiance image is used to indicate the corresponding radiance. Among them, the image preprocessing of the multispectral image of the plant drone canopy within the target irradiance range in each growth period based on the image preprocessing function can include exposure and gain equalization processing, halo correction processing and lens distortion correction processing. Exposure and gain normalization is to place the imaging sensor in a dark room and set different exposure times for shooting. The extracted remote sensing image pixel brightness value (DN value) is used as the noise image, and the original image is subtracted from the noise image for exposure and gain normalization processing; halo correction processing is to take the average brightness value of the adjacent pixels of the spot as the spot brightness, or use the cubic convolution method to correct the spot brightness; lens distortion correction processing is to use enough black and white chessboard images, use the least squares method to calculate the intrinsic and extrinsic parameters of the lens, and calculate the distortion-related parameters according to the remaining point coordinates for correction.
[0136] Then, based on the radiance image corresponding to each growth period and the irradiance of the plant drone canopy multispectral image within the target irradiance range in each growth period, the reflectance image corresponding to each growth period is obtained. The reflectance image includes the reflectance value. The calculation formula of the reflectance value can be seen in the following formula (5):
[0137] ρ(λ)=L(λ)*π / EDLS(λ) (5)
[0138] Where λ represents the band, ρ(λ) represents the reflectance value of the pixel in the reflectance image of each band, and L(λ) represents the radiance of the pixel in the radiance image corresponding to each band (in watts / (steradian*m) 2), W / (sr*m 2 )), EDLS(λ) represents the irradiance of each band of the multispectral image of the plant canopy within the target irradiance range during each growth period (in watts / m 2 , W / m 2 ).
[0139] In the embodiment of the present application, the reflectivity value includes the reflectivity between the visible light band and the near infrared band. Based on the reflectivity value in the reflectivity image corresponding to each growth period, the vegetation index corresponding to each growth period can be determined.
[0140] Among them, the vegetation index may include, for example, the Optimized Soil Adjusted Vegetation Index (OSAVI), the Triangular Vegetation Index (TVI), the Enhanced Vegetation Index (EVI), the Red Edge Chlorophyll Index (CI), the Normalized Difference Red Edge Index (NDRE), and the like.
[0141] The specific calculation method of the vegetation index can be found in the following formulas (6) to (10):
[0142] OSAVI=(R NIR -R red ) / (R NIR +R red +0.1) (6)
[0143] TVI=60(R NIR -R green )-100(R red -R redge ) (7)
[0144] EVI=2.5*(R NIR -R red ) / (R NIR +6*R red -7.5*R blue +1) (8)
[0145] CI=(R NIR / R redge )-1 (9)
[0146] NDRE=(R NIR -R redge ) / (R NIR +R redge ) (10)
[0147] Where OSAVI stands for Optimized Soil Adjusted Vegetation Index, TVI stands for Triangular Vegetation Index, EVI stands for Enhanced Vegetation Index, CI stands for Red Edge Chlorophyll Index, NDRE stands for Normalized Difference Red Edge Index; R NIR Represents the near-infrared reflectivity, Rred Represents the red band reflectivity, R green Represents the green band reflectance, R redge Represents the red edge band reflectivity, R blue Represents the blue band reflectance.
[0148] After obtaining the vegetation index, measured leaf area index and measured canopy chlorophyll content corresponding to each growth period, a growth monitoring model of the target plant can be constructed based on the vegetation index, measured leaf area index and measured canopy chlorophyll content corresponding to each growth period through a linear regression algorithm. Specifically, a function fitting can be performed with the vegetation index as the horizontal coordinate and the measured leaf area index and measured canopy chlorophyll content as the vertical coordinate, and the fitted curve is the growth monitoring model.
[0149] Taking corn as the target plant as an example, the following introduces the analysis and verification process of the correlation between solar incident irradiance and corn canopy reflectivity, as well as the construction process of the growth monitoring model.
[0150] (1) Test introduction
[0151] Experimental location: Comprehensive Experimental Base of Crop Science Institute, Chinese Academy of Agricultural Sciences, Horqin District, Tongliao City, Inner Mongolia Autonomous Region.
[0152] Corn Type: Spring Corn.
[0153] Sowing date: May 5; Small trumpet-shaped period: June 30, 53 days from the sowing date; Large trumpet-shaped period: July 19, 72 days from the sowing date; Grain formation period: August 9, 93 days from the sowing date.
[0154] First, corn was planted in multiple regional blocks and a corn nitrogen gradient experiment was conducted.
[0155] Corn is planted at a density of 75,000 plants / hm2 in multiple areas. 2 ; Variety: Jingnongke 728; Nitrogen gradient: a total of ten gradients, respectively 0, 45, 90, 135, 180, 225, 270, 315, 360, 540 kg / hm 2 Pure nitrogen, corresponding to ten area blocks.
[0156] Then, obtain the drone multispectral image data.
[0157] A drone equipped with a multispectral sensor is used to photograph corn in multiple area blocks to obtain corresponding multispectral image data, i.e., the initial plant drone canopy multispectral image of each area block. The multispectral sensor can capture images of multiple bands at the same time. For example, in the embodiment of the present application, the multispectral sensor simultaneously captures images of five bands, which include a blue band, a green band, a red band, a near-infrared band, and a red edge band. The parameters of these five bands can be found in Table 1 below:
[0158] Table 1
[0159] Band Central wavelength / nm Band width / nm blue 475 20 green 560 20 red 668 10 Near Infrared 840 40 Red Edge 717 10
[0160] The multispectral sensor is equipped with DLS, which is an airborne radiation correction sensor that measures solar incident irradiance and flight angle information, and can store data directly in the meta-tag image file format (TIFF) image.
[0161] The flight parameters of the drone were set as follows: longitudinal overlap rate of 85%, lateral overlap rate of 85%, flight altitude of 50m, flight speed of 1.2m / s, and automatic exposure. Parameters such as flight area, heading, and flight altitude were consistent between missions. The absolute position of the drone can be recorded using the position / altitude measurements obtained by the drone's embedded global positioning system (GPS) / inertial measurement unit (IMU) device. Flight tests were conducted at the corresponding test sites before and after the jointing stage of corn. The drone images were saved in TIFF format, and the multispectral sensor was set to automatically take photos according to the planned route.
[0162] Table 2 illustrates the relevant information on experimental design and multispectral imaging data acquisition:
[0163] Table 2
[0164]
[0165]
[0166] Figure 2 The schematic diagram of the mean and standard deviation of irradiance in each band in the data set provided in the embodiment of the present application is as follows: Figure 2As shown, the first four periods are used to construct the relationship between reflectance and illumination model; the last four periods are used to construct the growth monitoring model of corn from the small trumpet stage to the grain formation stage. Among them, 6 / 30 / 12, 7 / 19 / 10, and 8 / 09 / 14 are used to construct the growth monitoring model under the same irradiance, and 6 / 30 / 12, 7 / 20 / 14, and 8 / 09 / 14 are used to construct the growth monitoring model under sunny days and inconsistent irradiance conditions.
[0167] This experiment includes different treatments (including setting different nitrogen gradients for nitrogen experiments), mainly for repetition, and the data presents a certain range of changes, which means that most possible situations are covered, thereby improving the accuracy of the experiment. Therefore, this dataset can provide support for corn growth monitoring using irradiance as an environmental condition.
[0168] (2) Verification of the correlation between irradiance and reflectivity
[0169] The drone multispectral image dataset in the analysis experiment (including the drone canopy multispectral images of plants corresponding to multiple regional blocks at different times) was selected, and the drone multispectral image dataset was processed through the above steps, that is, the corresponding irradiance was extracted based on the initial plant drone canopy multispectral image corresponding to each regional block, and then the region of interest was extracted to obtain the drone canopy multispectral image of the plant corresponding to each regional block. After image preprocessing, the radiance image corresponding to the drone canopy multispectral image of each plant was obtained, and the reflectance value was obtained in combination with the corresponding irradiance. Finally, the irradiance and reflectance in the analysis experiment were used to construct polynomial, linear, exponential, and logarithmic models to obtain the relationship between irradiance and reflectance. Table 3 illustrates the relationship between irradiance and reflectance under different models such as polynomial, linear, exponential, and logarithmic:
[0170] Table 3
[0171] type Model formula <![CDATA[R 2 ]]> Polynomial <h2 style=";text-align:left;direction:ltr"><![CDATA[y=1.346x <h2 style=";text-align:left;direction:ltr"> 2 <h2 style=";text-align:left;direction:ltr"> -1.7769x+1.2175]]><h2 style=";text-align:left;direction:ltr"> 0.8490** Linear y=-0.3532x+0.8762 0.7267** index <![CDATA[y=0.8951e -0.501x ]]> 0.7426** Logarithm y=-0.184ln(x)+0.5625 0.7913** Power <![CDATA[y=0.5743x -0.26 ]]> 0.8059**
[0172] Table 3 shows the functional relationship between irradiance and reflectance obtained based on different model fittings, where x represents irradiance, y represents reflectance value, and R 2 Represents the absolute coefficient between irradiance and reflectance values. R 2 The calculation formula is as follows:
[0173]
[0174] Where n represents the sample size, which is a positive integer; x i represents the irradiance of the i-th sample, y i represents the reflectance value of the i-th sample, Represents the mean of the reflectance values of n samples.
[0175] Figure 3 A schematic diagram of the relationship between irradiance and reflectivity provided in the embodiment of the present application, such as Figure 3 As shown, the model is constructed based on the four phases of data in the analysis test in Table 2. The four phases of data in the analysis test involve data from 2022 / 7 / 31 and 2022 / 8 / 01. By default, the growth state of the plant does not change from 2022 / 7 / 31 to 2022 / 8 / 01. In each phase of data, the irradiance and corresponding reflectance values of the multispectral image of the plant drone canopy corresponding to each regional block are extracted. The method of extracting the irradiance and corresponding reflectance values of the multispectral image of the plant drone canopy corresponding to each regional block can be found in the relevant introduction of the above embodiment, which will not be repeated here.
[0176] Then, with the horizontal axis as irradiance and the vertical axis as reflectance value, fitting can be performed to obtain the correlation between irradiance and reflectance value. Figure 3 As shown in the figure, the irradiance and reflectance values of the multispectral images of the plant drone canopy corresponding to the ten regional blocks in the four phases of data in the analysis experiment are respectively illustrated, and the fitting curve between the irradiance and reflectance values is obtained by fitting, that is, Figure 3 Example of dashed lines in .
[0177] In theory, if the growth state of the plant does not change, the reflectance value corresponding to the multispectral image of the plant canopy by drone will remain unchanged. Figure 3 It can be seen that the reflectivity value changes with the change of irradiance. No matter whether a polynomial model, linear model, exponential model, logarithmic model or power model is constructed between irradiance and reflectivity, there is a very significant correlation between irradiance and reflectivity. Therefore, it can be determined that the incident irradiance is an important factor affecting the accuracy of reflectivity, which further proves that the growth monitoring model is accurate when the target irradiance corresponding to different growth periods is consistent.
[0178] (3) Construction of growth monitoring model
[0179] Figure 4 A schematic diagram for comparing the growth monitoring models provided in the embodiments of the present application is shown in FIG. Figure 4 As shown, the changing trends of leaf area index (LAI), canopy chlorophyll content (CCC) and enhanced vegetation index (EVI) of multiple regional blocks in the corn seedling stage to jointing stage as the number of days from the sowing date are illustrated.
[0180] exist Figure 4 In the figure, different curves represent different area blocks. Figure 4 (a) shows the variation trend of the leaf area index of corn in multiple blocks with the sowing date, where the horizontal axis is the number of days from the sowing date and the vertical axis is the leaf area index of corn; Figure 4 (b) shows the variation trend of corn canopy chlorophyll content in multiple regional blocks with sowing date, where the horizontal axis is the number of days from sowing date and the vertical axis is the corn canopy chlorophyll content; Figure 4 (c) shows the variation trend of the enhanced vegetation index with the sowing date under the growth monitoring model constructed under the condition of sunny day when the target irradiance corresponding to different growth periods is inconsistent. The horizontal axis is the number of days from the sowing date, and the vertical axis is the enhanced vegetation index. Figure 4 (d) shows the variation trend of the enhanced vegetation index with the sowing date under the growth monitoring model constructed when the target irradiance corresponding to different growth periods is consistent. The horizontal axis is the number of days from the sowing date, and the vertical axis is the enhanced vegetation index.
[0181] Table 4 illustrates the correlation coefficients between different vegetation index models and leaf area index and canopy chlorophyll content at different growth stages:
[0182] Table 4
[0183]
[0184] The formula for calculating the correlation coefficient can be found in the following formula (12):
[0185]
[0186] n represents the sample size, X i represents the i-th sample size in variable one, represents the average value of n samples in variable one; Y i represents the i-th sample size in variable 2, represents the average value of n samples in variable two; r represents the correlation coefficient between variable one and variable two.
[0187] As shown in Table 4, the correlation coefficients between different vegetation index models and leaf area index (LAI) and canopy chlorophyll content (CCC) of corn at the seedling stage to the jointing stage are illustrated, including the correlation coefficients between the vegetation index model and the leaf area index and canopy chlorophyll content under similar target irradiances, and the correlation coefficients between the vegetation index model and the leaf area index and canopy chlorophyll content under different target irradiances. Figure 4 As can be seen from the examples in Table 4, when the difference between the target irradiances corresponding to different growth periods is less than the preset irradiance threshold, the corn growth monitoring model constructed based on the multispectral images of plant canopies corresponding to multiple regional blocks in different growth periods has higher accuracy.
[0188] In summary, the plant growth monitoring model construction method, device and equipment provided in the embodiments of the present application first obtain multi-stage plant drone canopy multispectral images of the target plant in each growth period, and each stage of the plant drone canopy multispectral image has a corresponding irradiance; based on the irradiance corresponding to the multi-stage plant drone canopy multispectral images in each growth period, the irradiance range of each growth period can be determined, and then based on the irradiance range of each growth period, the target irradiance range is determined, and the target irradiance range is a range in which the irradiance ranges of each growth period overlap, or the target irradiance range is a range in which the distance between the irradiance range of each growth period is less than a preset distance; after determining the target irradiance range, the corresponding plant drone canopy multispectral images with irradiance within the target irradiance range are determined in the multi-stage plant drone canopy multispectral images in each growth period, and based on the plant drone canopy multispectral images within the target irradiance range in each growth period, and the measured growth indicators of the plants in each growth period, a growth monitoring model of the target plant is constructed through a linear regression algorithm. The scheme of the embodiment of the present application ensures that the irradiance of the subsequent multi-spectral images of the plant drone canopy used to construct the growth monitoring model is relatively close by determining the multi-spectral images of the plant drone canopy within the target irradiance range in each growth period. Since the light is strong and stable from 10 am to 2 pm on a sunny day, it is undeniable that 10 am to 2 pm on a sunny day is the best time to obtain drone optical images. However, for crops with a long growth period, such as wheat, even if it is sunny, the light intensity in the early and late growth stages is greatly different, so the scheme of the embodiment of the present application can not only reduce the dependence on the weather, but also to a certain extent improve the accuracy of plant growth monitoring on sunny days.
[0189] The plant growth monitoring model construction device provided in the present application is described below. The plant growth monitoring model construction device described below and the plant growth monitoring model construction method described above can be referenced to each other.
[0190] Figure 5 The schematic diagram of the structure of the plant growth monitoring model construction device provided in the embodiment of the present application is as follows: Figure 5 As shown, the device comprises:
[0191] An acquisition module 51 is used to acquire multi-period plant drone canopy multispectral images of target plants in each growth period, and each period of plant drone canopy multispectral images has corresponding irradiance;
[0192] The first determination module 52 is used to determine the irradiance range of each growth period based on the irradiance corresponding to the multi-spectral image of the plant canopy by the drone in each growth period;
[0193] A second determination module 53 is used to determine a target irradiance range based on the irradiance ranges of each growth period, wherein the target irradiance range is a range where the irradiance ranges of each growth period intersect, or the target irradiance range is a range where the distance between the irradiance range of each growth period and the target irradiance range is less than a preset distance;
[0194] The third determination module 54 is used to determine the corresponding plant drone canopy multispectral images whose irradiance is within the target irradiance range from the multi-stage plant drone canopy multispectral images in each growth period;
[0195] The processing module 55 is used to construct a growth monitoring model of the target plant through a linear regression algorithm based on the multispectral image of the plant drone canopy within the target irradiance range in each growth period and the measured growth indicators of the plants in each growth period.
[0196] In a possible implementation manner, the acquisition module 51 is specifically configured to:
[0197] During each growth period of the target plant, a multi-spectral image of the plant drone canopy is collected by a multi-spectral sensor carried by a drone, wherein each period of plant drone canopy multi-spectral image includes corresponding multiple groups of plant drone canopy multi-spectral images, each group of plant drone canopy multi-spectral images has the irradiance at the time of acquisition, and each group of plant drone canopy multi-spectral images represents a combination of original images of each single band.
[0198] In one possible implementation,
[0199] The irradiance of each set of plant drone canopy multispectral images is acquired by the downlink light sensor carried on the said drone and automatically stored in the raw image of each band;
[0200] The irradiance corresponding to each period of plant drone canopy multispectral imagery is obtained by taking the average of the irradiances of the corresponding multiple groups of plant drone canopy multispectral images, where the irradiance of each group of plant drone canopy multispectral images is represented by the irradiance of the original image in the near-infrared band.
[0201] In a possible implementation manner, the first determining module 52 is specifically configured to:
[0202] Determine the irradiance corresponding to each phase of multi-spectral drone canopy images of plants in each growth period;
[0203] The maximum and minimum irradiances were determined among the irradiances corresponding to the multi-spectral images of plant canopies taken by drones during each growth period.
[0204] Based on the maximum and minimum irradiances corresponding to the multi-spectral images of plant canopies taken by drones during multiple periods of each growth period, the irradiance range for each growth period is determined.
[0205] In a possible implementation manner, the processing module 55 is specifically used to:
[0206] Perform image preprocessing on the multispectral image of the plant canopy by drone within the target irradiance range in each growth period to obtain the radiance image corresponding to each growth period;
[0207] Based on the radiance image corresponding to each growth period and the irradiance of the multispectral image of the plant drone canopy within the target irradiance range in each growth period, the reflectance image corresponding to each growth period is obtained;
[0208] Determine the vegetation index corresponding to each growth period based on the reflectance value in the reflectance image corresponding to each growth period;
[0209] Based on the vegetation index, measured leaf area index and measured canopy chlorophyll content corresponding to each growth period, a growth monitoring model of the target plant is constructed through a linear regression algorithm;
[0210] Among them, the reflectivity value of the pixel point in the reflectivity image is obtained by the following formula:
[0211] ρ(λ)=L(λ)*π / EDLS(λ),
[0212] Where λ represents the band, ρ(λ) represents the reflectance value of the pixel in the reflectance image of each band, and L(λ) represents the radiance of the pixel in the radiance image corresponding to each band (unit: W / (sr*m 2 )), EDLS(λ) represents the irradiance of each band of the multispectral image of the plant canopy within the target irradiance range in each growth period (unit: W / m 2 ).
[0213] In a possible implementation, the reflectivity value includes the reflectivity between the visible light band and the near infrared band;
[0214] The vegetation index includes at least one of the optimized soil conditioned vegetation index, the triangular vegetation index, the enhanced vegetation index, the red edge chlorophyll index, and the normalized difference red edge index;
[0215] The calculation formula of vegetation index is as follows:
[0216] OSAVI=(R NIR -R red ) / (R NIR +R red+0.1),
[0217] TVI=60(R NIR -R green )-100(R red -R redge ),
[0218] EVI=2.5*(R NIR -R red ) / (R NIR +6*R red -7.5*R blue +1),
[0219] CI=(R NIR / R redge )-1,
[0220] NDRE=(R NIR -R redge ) / (R NIR +R redge ),
[0221] Where OSAVI stands for Optimized Soil Adjusted Vegetation Index, TVI stands for Triangular Vegetation Index, EVI stands for Enhanced Vegetation Index, CI stands for Red Edge Chlorophyll Index, NDRE stands for Normalized Difference Red Edge Index; R NIR Represents the near-infrared reflectivity, R red Represents the red band reflectivity, R green Represents the green band reflectance, R redge Represents the red edge band reflectivity, R blue Represents the blue band reflectance.
[0222] In a possible implementation manner, the measured plant growth indicators in each growth period include the measured leaf area index and the measured canopy chlorophyll content.
[0223] Figure 6 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 6As shown, the electronic device may include: a processor 610, a communication interface 620, a memory 630 and a communication bus 640, wherein the processor 610, the communication interface 620 and the memory 630 communicate with each other through the communication bus 640. The processor 610 may call the logic instructions in the memory 630 to execute the method for building a plant growth monitoring model, the method comprising: step 1, obtaining multi-period plant drone canopy multispectral images of the target plant in each growth period, each period of the plant drone canopy multispectral image has a corresponding irradiance; step 2, based on the irradiance corresponding to the multi-period plant drone canopy multispectral images in each growth period, determining the irradiance range of each growth period; step 3, based on the irradiance range of each growth period, determining the target irradiance range, the target irradiance range being the irradiance of each growth period. There is an overlapping range of illumination ranges, or the target irradiance range is a range in which the distance between the target irradiance range and the irradiance range of each growth period is less than the preset distance; step 4, determining the corresponding plant drone canopy multispectral images with irradiance within the target irradiance range in the multi-period plant drone canopy multispectral images in each growth period; step 5, based on the plant drone canopy multispectral images within the target irradiance range in each growth period, and the measured growth indicators of the plants in each growth period, a growth monitoring model of the target plant is constructed through a linear regression algorithm.
[0224] In addition, the logic instructions in the above-mentioned memory 630 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art, and the computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a disk or an optical disk.
[0225] On the other hand, the present application also provides a computer program product, which includes a computer program, which can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the plant growth monitoring model construction method provided by the above methods, and the method includes: step 1, obtaining multi-period plant drone canopy multispectral images of the target plant in each growth period, and each period of the plant drone canopy multispectral image has a corresponding irradiance; step 2, based on the irradiance corresponding to the multi-period plant drone canopy multispectral images in each growth period, determining the irradiance range of each growth period; step 3, based on the irradiance of each growth period Illumination range, determine the target irradiance range, the target irradiance range is the range in which the irradiance ranges of each growth period overlap, or the target irradiance range is the range in which the distance between the irradiance range of each growth period is less than the preset distance; step 4, determine the corresponding plant drone canopy multispectral images with irradiance within the target irradiance range in the multi-period plant drone canopy multispectral images in each growth period; step 5, based on the plant drone canopy multispectral images within the target irradiance range in each growth period, and the measured growth indicators of the plants in each growth period, construct the growth monitoring model of the target plant through a linear regression algorithm.
[0226] On the other hand, the present application also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the plant growth monitoring model construction method provided by the above methods, the method comprising: step 1, obtaining multi-period plant drone canopy multispectral images of the target plant in each growth period, each period of the plant drone canopy multispectral image having a corresponding irradiance; step 2, based on the irradiance corresponding to the multi-period plant drone canopy multispectral images in each growth period, determining the irradiance range of each growth period; step 3, based on the irradiance range of each growth period, determining the target irradiance range, The target irradiance range is a range in which the irradiance ranges of each growth period overlap, or the target irradiance range is a range in which the distance between the irradiance range of each growth period is less than a preset distance; step 4, determining the corresponding plant drone canopy multispectral images whose irradiance is within the target irradiance range in the multi-period plant drone canopy multispectral images in each growth period; step 5, based on the plant drone canopy multispectral images within the target irradiance range in each growth period, and the measured growth indicators of the plants in each growth period, a growth monitoring model of the target plant is constructed through a linear regression algorithm.
[0227] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.
[0228] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.
[0229] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.
Claims
1. A method for constructing a plant growth monitoring model, characterized in that: include: Step 1: Obtain multi-spectral images of the plant canopy of the target plant by drone in each growth period, and each multi-spectral image of the plant canopy by drone has corresponding irradiance; Step 2, based on the irradiance corresponding to the multi-spectral images of the plant canopy taken by drones during each growth period, determine the irradiance range for each growth period; Step 3, based on the irradiance range of each growth period, determine the target irradiance range, the target irradiance range is a range where the irradiance ranges of each growth period overlap, or the target irradiance range is a range where the distance between the irradiance range of each growth period and the target irradiance range is less than a preset distance; Step 4, determining the corresponding plant drone canopy multispectral images whose irradiance is within the target irradiance range from the multi-stage plant drone canopy multispectral images in each growth period; Step 5: Based on the multispectral images of the plant canopy taken by drones within the target irradiance range in each growth period and the measured growth indicators of the plants in each growth period, a growth monitoring model of the target plant is constructed through a linear regression algorithm.
2. The method according to claim 1, characterized in that The step 1 comprises: During each growth period of the target plant, a multi-spectral image of the plant drone canopy is collected by a multi-spectral sensor carried by a drone, wherein each period of plant drone canopy multi-spectral image includes corresponding multiple groups of plant drone canopy multi-spectral images, each group of plant drone canopy multi-spectral images has the irradiance at the time of acquisition, and each group of plant drone canopy multi-spectral images represents a combination of original images of each single band.
3. The method according to claim 2, characterized in that The irradiance of each set of plant drone canopy multispectral images is acquired by the downlink light sensor carried on the said drone and automatically stored in the raw image of each band; The irradiance corresponding to each period of plant drone canopy multispectral imagery is obtained by taking the average of the irradiances of the corresponding multiple groups of plant drone canopy multispectral images, where the irradiance of each group of plant drone canopy multispectral images is represented by the irradiance of the original image in the near-infrared band.
4. The method according to claim 1, characterized in that: The step 2 comprises: Determine the irradiance corresponding to each phase of multi-spectral drone canopy images of plants in each growth period; The maximum and minimum irradiances were determined among the irradiances corresponding to the multi-spectral images of plant canopies taken by drones during each growth period. Based on the maximum and minimum irradiances corresponding to the multi-spectral images of plant canopies taken by drones during multiple growth periods, the irradiance range for each growth period is determined.
5. The method according to any one of claims 1 to 4, characterized in that: The step 5 comprises: Perform image preprocessing on the multispectral image of the plant canopy by drone within the target irradiance range in each growth period to obtain the radiance image corresponding to each growth period; Based on the radiance image corresponding to each growth period and the irradiance of the multispectral image of the plant drone canopy within the target irradiance range in each growth period, the reflectance image corresponding to each growth period is obtained; Determine the vegetation index corresponding to each growth period based on the reflectance value in the reflectance image corresponding to each growth period; Based on the vegetation index, measured leaf area index and measured canopy chlorophyll content corresponding to each growth period, a growth monitoring model of the target plant is constructed through a linear regression algorithm; Among them, the reflectivity value of the pixel point in the reflectivity image is obtained by the following formula: ρ(λ)=L(λ)*π / EDLS(λ), Where λ represents the band, ρ(λ) represents the reflectance value of the pixel in the reflectance image of each band, and L(λ) represents the radiance of the pixel in the radiance image corresponding to each band (unit: W / (sr*m 2 )), EDLS(λ) represents the irradiance of each band of the multispectral image of the plant canopy within the target irradiance range in each growth period (unit: W / m 2 ).
6. The method according to claim 5, characterized in that The reflectivity value includes the reflectivity between the visible light band and the near infrared band; The vegetation index includes at least one of the optimized soil conditioned vegetation index, the triangular vegetation index, the enhanced vegetation index, the red edge chlorophyll index, and the normalized difference red edge index; The calculation formula of vegetation index is as follows: OSAVI=(R NIR -R red ) / (R NIR +R red +0.1), TVI=60(R NIR -R green )-100(R red -R redge ), EVI=2.5*(R NIR -R red ) / (R NIR +6*R red -7.5*R blue +1), CI=(R NIR / R redge )-1, NDRE=(R NIR -R redge ) / (R NIR +R redge ), Where OSAVI stands for Optimized Soil Adjusted Vegetation Index, TVI stands for Triangular Vegetation Index, EVI stands for Enhanced Vegetation Index, CI stands for Red Edge Chlorophyll Index, NDRE stands for Normalized Difference Red Edge Index; R NIR Represents the near-infrared reflectivity, R red Represents the red band reflectivity, R green Represents the green band reflectance, R redge Represents the red edge band reflectivity, R blue Represents the blue band reflectance.
7. The method according to any one of claims 1 to 4, characterized in that: The measured growth indicators of plants in each growth period include the measured leaf area index and the measured canopy chlorophyll content.
8. A plant growth monitoring model construction device, characterized in that: include: The acquisition module is used to obtain multi-period plant drone canopy multispectral images of the target plant in each growth period. Each period of plant drone canopy multispectral images has corresponding irradiance; The first determination module is used to determine the irradiance range of each growth period based on the irradiance corresponding to the multi-spectral images of the plant canopy taken by drones during multiple growth periods; A second determination module is used to determine a target irradiance range based on the irradiance range of each growth period, wherein the target irradiance range is a range in which the irradiance ranges of each growth period overlap, or the target irradiance range is a range in which the distance from the irradiance range of each growth period is less than a preset distance; The third determination module is used to determine the corresponding plant drone canopy multispectral images whose irradiance is within the target irradiance range from the multi-stage plant drone canopy multispectral images in each growth period; The processing module is used to construct a growth monitoring model of the target plant through a linear regression algorithm based on the multispectral image of the plant drone canopy within the target irradiance range in each growth period and the measured growth indicators of the plants in each growth period.
9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the plant growth monitoring model construction method according to any one of claims 1 to 7 is implemented.
10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for constructing a plant growth monitoring model as described in any one of claims 1 to 7 is implemented.
