CN112293247A

CN112293247A - Remote sensing technology-based image acquisition system for screening rice varieties

Info

Publication number: CN112293247A
Application number: CN202011001538.6A
Authority: CN
Inventors: 吴贤婷; 朱仁山; 龚龑; 彭漪; 方圣辉
Original assignee: Wuhan University WHU
Current assignee: Wuhan University WHU
Priority date: 2020-09-22
Filing date: 2020-09-22
Publication date: 2021-02-02

Abstract

The invention relates to an image acquisition system for screening rice varieties based on a remote sensing technology, which comprises a test field, a flying carrier and an image acquisition tool; the test field comprises a plurality of cells, each cell is planted with a rice variety, and all areas of the test field are reflected on the image acquired by the image acquisition tool at one time; each cell covers at least 40 multiplied by 40 pixels on the image collected by the image collecting tool; there are gaps between cells. By using the system of the invention, the test and the relation between the flying heights of the flying carrier can be properly arranged according to the performance of the image acquisition tool. When the MAC image acquisition system is used, under the shooting height of 50m, the area of a cell is compressed to be 1m minimum²About 60 plants. The design ensures that reliable reflectivity characteristic data is extracted, greatly reduces labor cost, increases the number of cells in a test field per unit area, and increases the sieveAnd (4) selecting flux.

Description

Remote sensing technology-based image acquisition system for screening rice varieties

Technical Field

The invention relates to the field of intelligent agriculture, in particular to an image acquisition system for screening rice varieties based on a remote sensing technology

Background

With the development of remote sensing technology, the development of unmanned aerial vehicles and the improvement of camera resolution, reflection characteristic information of plant canopies is obtained through remote sensing influence, Vegetation Indexes (VI) are obtained through calculation, and the VI are connected with corresponding physiological parameters (such as nitrogen content) so as to be used for agricultural production and research. Lukas Prey et al placed the spectrometer approximately 1m above the wheat canopy, and measured the reflectance of the canopy and used to evaluate the corresponding physiological data. Zheng Heng Biao and so on of the Nanjing agriculture university use an unmanned aerial vehicle carried camera to shoot images of crops, and analyze and evaluate physiological parameters of the crops based on the images, and show that certain correlation exists between some physiological parameters and VI calculated by the images shot by the unmanned aerial vehicle.

The research is only limited to research and discussion for evaluating theoretical possibility of remote sensing and unmanned aerial vehicle technology in plant parameter monitoring, and the remote sensing and unmanned aerial vehicle technology is not actually applied to crop breeding, nor does the research provide inspiration and help for specific steps and parameters in an actual breeding process.

Generally, in the rice breeding and screening process, a plurality of rice varieties are planted in the same test field, and the rice varieties with required characters are screened by comparison under the same water supply, fertility and climate conditions. In order to ensure the survival of the plants and the effectiveness of the sample size, each rice line needs a certain number of plants which are planted together to form a cell. And screening based on remote sensing and unmanned aerial vehicle technology has extra requirements on the size of a cell. At present, no published literature is available to study how to rationally plan test field size, cell size and distribution, and unmanned aerial vehicle flight altitude.

Therefore, condition parameters need to be effectively optimized based on the characteristics of remote sensing and unmanned aerial vehicle technologies, and an image acquisition system suitable for rice variety screening is constructed, so that images capable of extracting reflectivity characteristic information of each cell can be obtained, and the method is used for agricultural breeding research and practice, especially high-throughput omics analysis and large population breeding screening.

Disclosure of Invention

In order to solve the problems, the invention provides an image acquisition system for screening rice varieties based on a remote sensing technology, which is characterized by comprising a test field, a flying carrier and an image acquisition tool;

the flying carrier carries the image acquisition tool to fly over the test field at a specified time, and acquires the image of the test field;

the test field comprises a plurality of cells, each cell is planted with a rice variety, and the size of the test field is set to be that when the flying carrier flies over the test field, all areas of the test field are reflected on the image acquired by the image acquisition tool at one time;

each cell covers at least 40 x 40 pixels on the image acquired by the image acquisition tool;

and an interval is arranged between every two cells.

In a specific embodiment, the relationship among the flying height H of the flying vehicle when acquiring the image of the test field, the longitudinal maximum length a of the test field, and the transverse maximum length B of the test field satisfies the following inequality:

A/H is less than or equal to X/H, and B/H is less than or equal to Y/H,

wherein X and Y are the maximum longitudinal length and the maximum transverse length which can be acquired by the image acquisition tool under the known height h.

In a preferred embodiment, the flying vehicle has a flying height of 50-200m when acquiring the image of the test field.

In a preferred embodiment, each of said cells is such as to cover at least a minimum square having a side length 40 times the spatial resolution of said image acquired by said image acquisition tool at height H.

In a preferred embodiment, the spacing between every two of said cells is not less than 40 cm. At least 40cm of interval is needed between the cells to avoid the problem that the reflectivity characteristic information cannot be extracted for each cell due to the fact that the cells are overlapped with each other in the late growth period of the rice.

In one embodiment, the image capture tool is equipped with one or more cameras with bandpass filters.

In a specific embodiment, the center wavelength of the band pass filter is 490,520,550,570,670,680,700,720,800,850,900 or 950 nm.

In a preferred embodiment, a ground calibration target is also provided in the test field to provide relatively stable reflectivity in the visible to near infrared wavelengths. For example, 6 ground calibration targets may be provided, each providing a ground calibration target that provides a relatively stable reflectance of visible to near-infrared wavelengths of 0.03, 0.12, 0.24, 0.36, 0.56, and 0.80, respectively.

In a preferred embodiment, a color difference cell is also set in the test field. For example, a purple rice cell can be provided, with purple rice varieties containing higher anthocyanins than chlorophyll, and thus being advantageous in data processing for distinguishing reflectance characteristics from other varieties. NDRE values are between 1 for cool blue 0 and warm red, so warmer colors represent higher chlorophyll content, nitrogen accumulation and photosynthetic rate, and colder colors, in contrast.

By using the system of the invention, the relationship among the size of the test field, the size and the interval of the cells and the flying height of the flying carrier can be properly arranged according to the performance of the image acquisition tool. When the MAC image acquisition system is used, the area of a cell can be compressed to be 16m at the minimum under the shooting height of 200m²Left and right; at a shooting height of 50m, we can compress the area of a cell to a minimum of 1m²About 60 plants. The separation distance between cells ensures that the situation that adjacent cells cannot be distinguished is avoided. The design ensures that reliable reflectivity characteristic data is extracted, greatly reduces labor cost, increases the number of cells in a test field in unit area, and increases screening flux.

Drawings

Fig. 1 is an image of a test field with no empty rows left between cells.

FIG. 2 is an image of a test field with an empty row (total spacing about 40cm) between cells;

fig. 3 is RGB and NDRE images of 6 growth stages of 51 rice varieties collected using an unmanned aerial vehicle, wherein: a is RGB image of 6 birth periods, a is TS period, b is JS period, c is PIS period, d is BS period, e is FHS period, f is MRS period; b is an NDRE image of 6 birth periods, a is TS period, B is JS period, c is PIS period, d is BS period, e is FHS period, and f is MRS period.

FIG. 4 is a graph of 6 fertility data relationships for 51 rice varieties and constructed regression models, in which: a is the relationship between SPAD and NDRE in 6 growth periods, and n is 306; b is a regression model between SPAD and NDRE constructed by 5 fertility data except TS stage, and R is²Is more than 0.81, n is 255; c is the relationship between N% and NDRE for 6 growth periods, N306; d is a regression model between SPAD and NDRE constructed by 5 fertility data except TS stage, R²＞0.61，n＝255。

FIG. 5 is a graph of the relationship between SPAD, N% and NDRE during FHS, where: a is the relationship between SPAD and NDRE in all rice varieties (n is 51); b is the relationship between SPAD and NDRE in early-maturing (EM) rice cultivars (n ═ 34); c is the relationship between SPAD and NDRE in late-maturing (LM) rice cultivars (n ═ 17); d is the relationship between N% and NDRE in all rice varieties (N ═ 51); e is the relationship between N% and NDRE in early-maturing (EM) rice varieties (N ═ 34); f is the relationship between N% and NDRE in late-maturing (LM) rice cultivars (N ═ 17). P < 0.001.

Fig. 6 is a non-linear regression model between N% × LAI and NDRE constructed from data from 6 growth stages of 42 rice varieties grown in south hai.

Detailed Description

1. Groping for shooting height, test field size and cell planning

A test field is selected, a plurality of rice varieties are planted in the test field, the row spacing is 20cm, the plant spacing is 16cm, and an unmanned aerial vehicle (S1000, Xinjiang) is used for carrying a Mini-MCA image acquisition system to shoot the test field in the air. The image pixels shot by the Mini-MCA image acquisition system are 1280 multiplied by 1024, and the image coverage is 69.375m multiplied by 55.417m when the image is shot at the height of 100 m. The spatial resolution was 54.17 mm.

In our experimental process, it is found that certain errors are brought about in the image stitching process, and the errors cause deviation in VI calculation and estimation of corresponding plant physiological parameters, so that the image of the whole experimental field should be included in the picture acquired by the image acquisition system at one time.

Therefore, the height we take using drone flight should be at or above the height that happens to have the entire test field taken on one picture. Based on this, assuming that the test field is rectangular, 69.375m × 55.417m, the shooting height should be 100m or more. Accordingly, if the test field length of the same aspect ratio is 50m, the flying height H should satisfy the formula H/50 ≧ 100/69.375, H being at least 72 m.

During experimentation we found that in order to effectively analyze the reflectance images of cells, each cell needs to cover at least 40 x 40 pixels. From the above-mentioned characteristics of MCA images, we conclude that if the image is taken at a height of 100m, the side length of the cell should be no less than 40 × 54.17mm, 2166.8mm, which is about 2 m. Accordingly, when the photographing height is 50m, the side length of the cell should not be less than about 1m, and when the photographing height is 200m, the side length of the cell should not be less than 4 m.

We have studied about the inter-cell spacing problem. When the row spacing and the plant spacing of the rice between the cells are 20cm multiplied by 16cm, the photos of the rice after growing up are shown in figure 1, and the adjacent cells are interwoven together and can not be distinguished from each other, even the situation that the high plant cell shields the adjacent short plant cell because of the difference of the plant heights occurs. Therefore, as a result of adjusting the intervals between the cells, as shown in fig. 2, when the cells are spaced apart by a distance equal to or more than one blank row (i.e., the distance between the boundary plants is about 40cm), the images of the cells can be distinguished, and the reflectivity characteristics can be extracted for each cell.

2. Screening examples of high NUE Rice varieties

2.1 plant Material and planting

50 varieties (indica rice, Australia rice and varieties between the indica rice and the Australia rice) with higher NUE are selected from 3000 genome projects, 51 rice varieties are added, and the rice varieties are planted in the experimental and research base (30.3756 degrees N, 114.7448 degrees E) of Wuhan university in Hubei province. The Huzhou rice is sown in 2017 in 5-month and 10-day rice and transplanted in 5-month and 31-day rice. The rice in the water of the hilly side germinates in 12-month and 10-month period in 2017, and is transplanted in 1-month and 6-month period in 2018. 60 plants are planted in each variety, 10 plants are planted in one row, 6 rows are formed, the row spacing is 20cm, the plant spacing is 16cm, a cell is formed, and the interval between every two cells is 40 cm.

41 indica rice varieties are selected from Chinese breeding projects, 42 rice varieties are added, and the rice varieties are planted in Wuhan university hybrid rice experiment and research bases (18 degrees 03 '147.1' N,110 degrees 03 '34.9' E) in Hainan Ling water. The rice is sown in 2017, 12 and 10 days, and transplanted in 2018, 1 and 5 days. 60 plants, 40 plants and one row are planted in each variety, the total row is 24 rows, the row spacing is 20cm, the plant spacing is 16cm, a cell is formed, and the interval between every two cells is 40 cm.

375Kg of compound fertilizer (the ratio of nitrogen, phosphorus and potassium is 15-15-15) is applied per hectare for conventional paddy field management. At each development stage of each experiment, a UAV drone was arranged to acquire images of all rice fields, and each field was repeatedly measured 5 times.

2.2 image acquisition

Images of the target study area were acquired using the Mini-MCA system installed on the unmanned aerial vehicle (S1000, majiang), and images were collected every five days from the time of transplantation until the crop was mature. The Mini-MCA comprises an array of 12 individual miniature digital cameras. 10bit SXGA data can be generated for each sensor channel, and the image resolution can reach 1m/130 hectare. Each camera is equipped with a custom bandpass filter, each bandpass filter having a center wavelength of 490,520,550,570,670,680,700,720,800,850,900, 950nm, respectively.

The MCA system is connected to the UAV through a gimbal frame to prevent the influence of the movement of the UAV, and the inaccurate matching effect of the cameras is controlled through 12 cameras matched before flying. Each UAV flight is performed in sunny, cloudless sky conditions for a time between 10am and 2pm, which is the minimum change in the solar altitude. In the Hubei experiment, the UAV flight height is 50m above the target block, and the spatial resolution is about 2.7 cm. In the water-of-land experiment, the UAV flight height is 200m above a target block, and the spatial resolution is about 10.8 cm.

The whole growth cycle of rice can be divided into 6 typical developmental stages, including: tillering Stage (TS), Jointing Stage (JS), ear differentiation stage (PIS), Booting Stage (BS), heading stage (FHS) and milk stage (MRS). An image was taken at each representative time of the session, and the nitrogen content (N%) and chlorophyll content (SPAD value) were then measured using conventional methods.

The image digital quantization value (DN) is converted to a surface reflectivity (ρ λ) using an empirical linear correction method. Image radiance correction was performed through a standard of 6 ground calibration targets placed in the camera field of view first on each flight. The area of the study and all ground calibration targets are contained in the same photograph. Herein, the ground calibration target provides relatively stable reflectivities of 0.03, 0.12, 0.24, 0.36, 0.56, and 0.80 for visible to near-infrared wavelengths, respectively. Since a linear relationship is assumed between DN and ρ λ, the reflectance equation for a rice variety can be:

ρ_λ＝DN_λ×Gain_λ+offset_λ

(λ＝490，520，550，570，670，680，700，720，800，850，900and 900nm) (1)

where ρ is_λAnd DN_λDigitally quantizes the surface reflectivity at wavelength λ and the image for the given pixel. Gain λ and Offset λ are the camera Gain and Offset of the camera at wavelength λ. Gain_λAnd Offset_λThe p value and DN value may be calculated using a least squares method.

2.3 calculation of VI

Data analysis and Statistical description were performed by IBM SPSS Statistics (Statistical Product and Service Solutions 22.0, IBM, Armonk, NY, United States). GraphPad software (Versi) was usedon 5.0.，Harvey Motulsky&Arthur Christopoulos, San Diego, California, USA). And (3) carrying out statistical evaluation on the data sets of nitrogen content (N%), chlorophyll content (SPAD) and Leaf Area Index (LAI) according to requirements, and displaying normal distribution. The poisson correlation coefficient (r) is used as a result of the correlation analysis. Analysis and comparison of corrected R²And p-value, regression analysis was performed. The best fit curve is converted to an equation as a regression model to represent the correlation between N%, SPAD, LAI × N% and Normalized Difference Red Edge (NDRE) or other Vegetation Index (VI). The calculation formula of each VI is shown in table 1.

TABLE 1 VI equations of calculation and references

2.4 correlation between several VI and Nitrogen content and chlorophyll

The former collected canopy spectral reflectance data using a spectrometer 1m above the rice and analyzed data for 6 key stages in the growth cycle to determine which growth stage was the best stage for the selected VI for assessment of chlorophyll and nitrogen content.

The above 6 stages of 42 rice varieties were evaluated by correlation analysis and regression analysis (table 2). The association of NDRE with chlorophyll is generally better than the association with nitrogen at each growth stage. However, NDRE showed correlation with chlorophyll (R) in the BS phase²0.0798) and nitrogen content (R)²0.0003). Furthermore, for chlorophyll, NDRE is in MRS phase (R)²0.4647) and JS period (R)²0.4627) is better than the PIS phase (R)²0.2748) and TS period (R)²0.1614). For the nitrogen content, the other 5 periods except FHS showed poor correlation with NDRE (R)²Less than 0.4). FHS shows a strong correlation with chlorophyll (0.6557) and nitrogen content (0.4919). As FHS is the period when the mature morphological structure and basic biomass of rice plants are completely established, it can be reasonably concluded that for the analysis of VI by the chromatograph measurement method in the close range, TSRapid morphological changes in plant development during periods JS and PIS, scion and inflorescence inception during periods PIS and BS, and canopy and grain color changes during MRS have the potential to interfere with reflectance data collection.

TABLE 26 regression analysis of SPAD, N% and NDRE in the growth phases

*：p<0.05；**：p<0.01；***：p<0.001。

2.5 NDRE real-time model reversible analysis of growth differences between rice varieties

In order to analyze the entire growth cycle from the transplanting period to the harvesting period, 51 rice varieties were planted in rectangular cells (1.2m × 1.6m), image data were collected with UAVs every 5-7 days (determined according to sunlight conditions), and as a result, as shown in fig. 3, images of six growth periods each exhibited a good reflectance spectrum of the plant, and effective reflectance characteristic information could be extracted for each cell. A purple rice variety was included as an internal control in the test group, and since it contained higher anthocyanin than chlorophyll, it was advantageous to distinguish it from other varieties in reflectance characteristics at the time of data processing. NDRE values are between 1 for cool blue 0 and warm red, so warmer colors represent higher chlorophyll content, nitrogen accumulation and photosynthetic rate, and colder colors, in contrast. In the birth cycles of all varieties, from TS period, JS period to PIS period, the NDRE value gradually increases, and rapidly decreases after BS period.

The NDRE values for each growth period for 51 rice varieties ranged as follows: TS stage (0.4121-0.5473), JS stage (0.4555-0.6173), PIS stage (0.3762-0.5762), BS stage (0.3506-0.5394), FHS stage (0.1931-0.4134), MRS stage (0.1487-0.3343). Among them, TS, JS, PIS and BS observed the highest NDRE in rice variety #33(Qingtai Ai), FHS and MRS observed the highest NDRE in rice variety #1(LY 9348). The lowest NDRE was observed in #17(ARC11777, TS), #4(Luohong 4B, JS and PIS), #16(MaMaGu, BS), #7(ZuiHou, FHS) and #28(MoMi, MRS).

High NDRE values above 0.5 were observed for all rice varieties in JS, PIS and BS phases, which are related to rice development, since JS phase is a period during vegetative growth when biomass is rapidly accumulated and PIS/BS phase is a period of transformation from vegetative to reproductive growth, which indicates that leaf and stem growth requires more energy than later flower and seed production. However, the exact maximum NDRE value, the time to reach the maximum NDRE value and the time to fall back from the maximum NDRE value vary widely among 51 rice varieties at the same time or at different times for the same variety. This indicates that chlorophyll content, photosynthetic rate, nitrogen uptake, transport, accumulation and ability to maintain nitrogen levels vary among varieties and at different times throughout the reproductive cycle. Thus, NDRE can be used as a parameter to measure and evaluate the real-time changes in chlorophyll and nitrogen accumulation.

2.6 optimization of prediction model and establishment of model I

To determine why the association of NDRE with chlorophyll and nitrogen varied during different growth periods, scatter plots were drawn on a total of 306 data from 6 growth periods for 51 rice varieties for analysis (fig. 4A and C). After transplanting, rice plants develop from plantlets (40cm high, 5-6 leaves) to large plants (120cm, 16-18 leaves). Growth of biomass and canopy modification is based on accumulation of chlorophyll and nitrogen.

In general, chlorophyll and nitrogen content is positively correlated with biomass prior to maturation. Thus, changes in chlorophyll and nitrogen content at each time period are expected to be as follows: TS (transport stream)<JS/PIS/BS. However, after maturation, chlorophyll begins to break down and the leaves age rapidly and yellow. Thus, chlorophyll and nitrogen are expected to exhibit a decline pattern from JS, PIS, BS to FHS, MRS. However, unlike expectations, TS was higher in both chlorophyll and nitrogen than other periods (fig. 4A and C). Considering the small biomass, narrow leaves, small plants at this stage, chlorophyll and nitrogen content may be erroneously overestimated. Since the TS stage plants are small, the reflectivity is actually a mixed characteristic of the plants and the paddy field water body. This value is overestimated because NDRE is calculated from the reflectance characteristics at 720nm and 800nm, and the water surface not covered by the plants increases the reflectance. Based on this inference, we cull TS-phase data, re-establishLinear regression model, NDRE has better correlation R with chlorophyll²NDRE also has a better linear relationship R with N ═ 0.8127²Above 0.60 (fig. 4B and D). Since N% is measured by the quantitative elemental analysis (EQA) method, we consider the regression model (y ═ 5.754 x)²+8.167x +0.5752) was a predictive model based on actual measurements as model i for further analysis below.

In summary, NDRE has a better correlation with chlorophyll and nitrogen content using only data from the fertile phase (JS phase and beyond) after the canopy has adequately covered the surface of the water for data processing. R of the established NDRE and nitrogen content model²The improvement is remarkable.

2.7, length of growth cycle affects NDRE accuracy in estimating chlorophyll and nitrogen

Due to the difference in the length of the growing cycle of 51 rice varieties, the standard for distinguishing the middle rice from the early rice and the late rice is adopted in the longer growing cycle. The growing cycle of medium rice is generally longer than 100 days from sowing to seed maturity, while early and late rice is generally shorter than 90 days. Therefore, the interval from sowing to maturity was set at 100 days as a cutoff value, and 51 rice varieties were divided into an early maturity group (EM) and a late maturity group (LM). To determine whether the length of the reproductive cycle affects the accuracy of chlorophyll and nitrogen content estimation, a linear regression analysis was performed on NDRE during FHS (fig. 5). The Regression Coefficient (RC) of each group after the grouping is raised. The RC between NDRE and chlorophyll rose from 0.6557 (mixed) to 0.7796(EM) and 0.7301 (LM). The RC between NDRE and nitrogen content ranges from 0.4919 (mixed) to 0.6152(EM) and 0.6282 (LM). This indicates that the maturation time affects the accuracy of the estimation of nitrogen content and chlorophyll in the reflectance profile of the meal.

From an agricultural perspective, the length of the reproductive cycle can create differences in biomass accumulation, leaf color, nitrogen flow and canopy structure, and thus can result in differences in reflectivity, which reduces the correlation between nitrogen content and NDRE (less than 0.5). By grouping varieties with similar growth cycle lengths together for group analysis, the correlation can be improved (above 0.6). From an agricultural perspective and phenomics research, taking into account the length of the growth cycle makes a better analysis when it is necessary to simultaneously estimate the nitrogen accumulation of hundreds or thousands of rice varieties.

10. The influence of LAI on the dependence of NDRE on nitrogen content

To determine whether canopy architecture is a key factor affecting the correlation of nitrogen content with NDRE, a training dataset (2017, table 5) of 42 rice varieties was used for analysis. Leaf Area Index (LAI) and nitrogen content (EQA) were measured and NDRE calculated from UAV data. N% LAI was used instead of N% as a parameter for correlation analysis. The nonlinear model y is obtained as 1.05571e^4.5666x(modelⅡ)，R²Is 0.86 (FIG. 6). The correlation of the Model is better than that of Model I. Thus, the above experiments demonstrate that NDRE appears to be strongly correlated with nitrogen content when canopy structures such as LAI are taken into account.

Our experiments also show that obtaining the height of reflectivity (50-200m) does not affect the estimation of nitrogen content, only needs the cell to cover at least 40 × 40 pixels. Moreover, in either ModelI or ModelII, we do not differentiate between the leaves and the ear for nitrogen content estimation, and only use the total reflectance characteristics of each variety for a blended estimation of nitrogen content.

It should be noted that although we try to fit the data of modelI and ModelII to the EQA measurement, we should point out that the EQA measurement is a sampling measurement with systematic error, and modelI and II are macroscopic data estimation results, we cannot in fact completely determine whether the error between the modelI and II estimation results and the true nitrogen content is large or the error between the EQA measurement results and the true nitrogen content is large. This systematic error may be the reason why the EQA method fails to distinguish LY9348 of high NUE from the population.

According to the characteristics of an image acquisition system, the size, the cell distribution and the shooting height of a test field are planned, the reflectivity characteristic information of each cell of the test field is obtained through the setting, an NDRE value is obtained through calculation, finally, two estimation models are used for successfully estimating the nitrogen content of the rice respectively, and the two estimation models can distinguish a NUE rice variety LY9348 from a higher NUE rice variety group. The above example demonstrates the inference of shot height, trial field size, cell plan herein in section 1 of the detailed description.

The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the present invention. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. An image acquisition system for screening rice varieties based on a remote sensing technology is characterized by comprising a test field, a flying carrier and an image acquisition tool;

and an interval is arranged between every two cells.

2. The image acquisition system according to claim 1, wherein the relationship among the flying height H of the flying vehicle when acquiring the test field image, the longitudinal maximum length a of the test field, and the transverse maximum length B of the test field satisfies the following inequality:

A/H is less than or equal to X/H, and B/H is less than or equal to Y/H,

3. The image capturing system of claim 2, wherein the flying vehicle has a flying height of 50-200m when capturing the image of the test field.

4. The image capturing system of claim 2, wherein each of the cells covers at least a minimum square having a side length 40 times a spatial resolution of the image captured by the image capturing tool at height H.

5. The image acquisition system of claim 1 wherein the spacing between each two of the cells is no less than 40 cm.

6. The image acquisition system according to any one of claims 1 to 6, wherein the image acquisition tool is equipped with one or more cameras with filters.

7. The image acquisition system of claim 6 wherein the filter has a center wavelength of 490,520,550,570,670,680,700,720,800,850,900 or 950 nm.

8. The image acquisition system according to any one of claims 1-6, wherein a ground calibration target is further provided in the test field to provide relatively stable reflectivity of visible to near infrared wavelengths.

9. The image acquisition system of any one of claims 1-6, wherein a chromatic aberration cell is further provided in the test field.