Sorghum intelligent identification method based on working object structure vector

Abstract

The invention relates to a sorghum intelligent identification method based on a work object structure vector. Comprising the following steps: firstly, establishing an optical and radar time sequence data set of a research area, acquiring key phenological periods of sorghum based on the time sequence data set, and combining crop attributes such as plant structures, leaf wax and tannin compound contents of sorghum in each phenological period; the method comprises the following steps: respectively establishing a saddle-shaped radar index of sparse-stalk broad-leaf crops in a sorghum growth period, an index of high-wax crops in a leaf growth period and an index of red tassel cap spikes in a mature period, constructing each index for representing a plant structure and biochemical substance content, and constructing a crop structure vector based on the indexes; and finally, establishing an intelligent sorghum identification method as an object structure vector, and obtaining a sorghum planting spatial distribution diagram in the research area. According to the constructed method, massive training samples or empirical parameter setting is not needed, the high-precision sorghum automatic drawing capability is achieved, and the large-scale multi-year low-cost application and popularization capability is achieved.

Description

Technical Field

[0001] The invention relates to the field of agricultural remote sensing, and in particular to a sorghum intelligent recognition method based on crop structure vectors. Background Art

[0002] Sorghum is an annual tall herb with a layer of white wax on the surface of its stems and leaves, which can reduce water evaporation during drought and has strong environmental adaptability. In terms of agricultural production, it has the advantages of easy management and strong drought resistance. Sorghum grains contain mixed polyphenols, such as tannins. Sorghum has a wide range of uses. It can be used directly as a food crop, as a high-quality feed crop, and as a biological and chemical energy crop. In my country, sorghum is mostly used for feed, brewing and biofuel, except for a small part used for food. my country is a major importer of sorghum in the world, and its total import volume ranks first in the world. Sorghum is the main raw material for brewing in my country. In recent years, with the development of China's wine industry, the demand for sorghum has been strong, and the sorghum planting area has increased rapidly, and it is expected to show a continuous upward trend in the future. Timely and accurate grasp of sorghum area distribution information is of great significance for scientifically and orderly guiding the development of the sorghum planting industry and implementing precise control of arable land use.

[0003] Traditional field survey methods require a lot of manpower, material and financial resources, and the implementation cycle of large-scale surveys is long. In addition, due to the wide range of sorghum planting and the scattered plots, it is difficult to obtain large-scale sorghum planting areas in real time and accurately. With the rapid development of remote sensing technology and the continuous enrichment of remote sensing time-series image data, there are great opportunities for the development and in-depth application of agricultural remote sensing technology. The attention paid to sorghum mapping research is far lower than that of major crops such as rice, winter wheat, and corn. In the existing crop identification technology and its application research, small crops such as sorghum are rarely covered. In the few crop identification studies involving sorghum, the classification accuracy of small crops such as sorghum is very low, far lower than that of major grain crops such as rice and wheat. The main reasons are as follows: 1) As a minor crop, sorghum has less training sample data, which seriously affects the classification accuracy based on data-driven models; 2) There is a lack of systematic research on the iconic characteristics of sorghum, and the existing training models have weak transferability, making it difficult to achieve large-scale automatic mapping.

[0004] Sorghum is characterized by tall plants, broad leaves, and sparse spike stalks, and the wax content in its leaves and the tannin content in its spikes are significantly higher than those of other broad-leaved crops with sparse spike stalks. From aspects such as the crop plant structure, spike stalk density, and wax and tannin contents, this invention designs a series of crop indices respectively, further constructs a crop body structure vector, and designs a technical process method for intelligent sorghum recognition. This invention comprehensively utilizes the Sentinel-1 SAR synthetic aperture radar scattering coefficient and the Sentinel-2 MSI optical hyperspectral time series dataset to lay a foundation for comprehensively and synthetically describing the body structure of crops, thereby effectively designing crop indices to construct a crop intelligent recognition method. In terms of the plant structure, the canopy and vertical structure of sorghum change during the growth process, and the radar backscattering time series curve presents a saddle shape; in addition, due to the broad leaves and sparse spike stalks of sorghum, compared with wheat, rice, etc. with narrow leaves and dense spike stalks, the overall radar backscattering coefficient of sorghum is higher during the leaf growth period, and the change amplitude of the saddle of radar backscattering during the booting stage is smaller; in terms of the content of biochemical substances, compared with low-wax crops such as corn, the existence of high wax makes the optical chemical index value of sorghum lower during the leaf growth period; secondly, a large amount of tannin compounds are accumulated in the sorghum spikes during the maturity period, so the content of its tannin index is much higher than that of corn. This invention makes full use of the changes in the spectral and radar backscattering time series curves of sorghum during different key phenological periods, integrates the radar index, wax index, and tannin index of sparse spike stalk broad-leaved crops, constructs a sorghum body structure vector applicable to automatic mapping of large-scale sorghum spatial distribution, and realizes intelligent sorghum recognition. Summary of the Invention

[0005] The purpose of this invention is to provide a method for intelligent sorghum recognition based on a crop body structure vector. The constructed method does not require a large number of training samples or empirical parameter settings, has the ability of high-precision automatic mapping of sorghum, and has the ability of low-cost application and promotion on a large scale for multiple years.

[0006] To achieve the above purpose, the technical solution of this invention is: a method for intelligent sorghum recognition based on a crop body structure vector, including the following steps:

[0007] Step S01: Establish an optical and radar time series dataset for the study area;

[0008] Step S02: Obtain the key phenological periods of sorghum based on the optical and radar time series dataset;

[0009] Step S03: Design a vertical polarization radar saddle-shaped detection index for spike stalk crops;

[0010] Step S04: Design a radar index for a sparse spike stalk canopy;

[0011] Step S05: Design an index for sparse spike stalk broad-leaved crops;

[0012] Step S06: Design the high-waxy crop index;

[0013] Step S07: Design the red tassel spike index at maturity;

[0014] Step S08: Comprehensively construct the crop structure vector index;

[0015] Step S09: Construct a sorghum intelligent recognition technology process method based on the crop structure vector;

[0016] Step S10: Obtain the sorghum planting spatial distribution map of the study area.

[0017] In an embodiment of the present invention, in the step S01, based on the Sentinel-1 SAR time series dataset, data preprocessing is performed, including radar data denoising based on Lee filtering, image mosaicking, and time series data smoothing; at the same time, the Sentinel-2 MSI time series dataset is obtained, and preprocessing operations are performed, including cloud recognition and cloud removal, image mosaicking, remote sensing index calculation, linear interpolation, and time series data smoothing, to construct a smoothed optical and radar time series dataset of the study area.

[0018] In the step S01, based on the Sentinel-1 SAR time series dataset, data preprocessing is performed. In the step S02, based on the optical and radar time series datasets, the key phenological periods of sorghum are determined; the dates corresponding to the first valley value, peak value of the vegetation index time series curve, and the last peak value of the tannin index time series curve during the crop growth period are respectively determined as the emergence stage d e , the vigorous growth stage d g , the maturity stage d m ; the forty days before the vigorous growth stage are determined as the visible flag leaf stage d v , and the thirty days before the maturity stage are determined as the hard grain stage d h ; further combined with the radar time series curve to determine the half-flowering stage d f , the booting stage d b , and the soft grain stage d s ; the calculation formula is:

[0019]

[0020] where t represents time, and VV t represents the radar scattering value, i.e., the VV value, at the corresponding annual accumulated day t; when F(t) = 1, the corresponding extreme values are summarized into a set I = {t1, t2, t3, t4...}; based on the elements in the set I, the key phenological periods of sorghum are further judged; the date corresponding to the minimum value closest to the vigorous growth stage is determined as the half-flowering stage; the dates corresponding to the maximum values closest to the front and back of the half-flowering stage are successively determined as the booting stage and the soft grain stage;

[0021] d f = d g - min{|d h - t1|, |d h - t2|, |d h - t3|...}, (VV t - VV t-1 ) < 0 (2)

[0022] d b = d f - min{(d f - t1), (d f - t2), (d f - t3)...}, d f > t n (3)

[0023] d s = d f - max{(d f - t1), (d f - t2), (d f - t3)...}, d f < t n (4)

[0024] Wherein, VV t - VV t-1 judges the extreme value direction at the corresponding moment, d e represents the emergence stage, d m represents the maturity stage, d f represents the semi-blooming stage, d g represents the vigorous growth stage, d h represents the hard grain stage, d b represents the booting stage, d s represents the soft grain stage.

[0025] In the step S01, based on the Sentinel-1 SAR time series dataset, data preprocessing is performed. In the step S03, during the leaf growth process of the ear-stalk crops, the number of leaves increases and gradually covers the land, and a strong scattering will be formed on the canopy surface. When the crops enter the booting stage, the ear-stalks and ears block part of the radar backscattering from the leaves, and this attenuation of the backscattering is relatively obvious. When the ears gradually become plump, the number of scattering particles from the ears increases, and the radar backscattering gradually strengthens. Until after entering the soft grain stage, the leaves gradually droop and wither from bottom to top, the radar backscattering energy attenuates, and the VV polarization continuously decreases; therefore, the VV polarization time series curve of the ear-stalk crops presents a saddle-shaped characteristic, and the saddle of the saddle-shaped is located in the vigorous growth stage of the crops; for the non-ear-stalk crops including soybeans, peanuts, and potatoes, the radar reflection time series curve only presents a shape of rising first and then falling:

[0026]

[0027] In the formula, in function M, 1 represents ear-stalk crops, and 0 represents non-ear-stalk crops; d b represents the booting stage, d g represents the full growth stage, d s represents the soft grain stage, card(I) represents the number of elements contained in set I, and when card(I) ≥ 3, it means there are three or more extreme values.

[0028] In the step S01, based on the Sentinel-1 SAR time series dataset, data preprocessing is performed. In the step S04, during the booting stage, the density of the ear stalks has an important impact on radar backscattering. The denser the ear stalks, the more obvious the vertical structure, and the more the radar scattering weakens. Therefore, compared with dense ear-stalk crops including wheat, the degree of decrease in radar backscattering of sparse ear-stalk crops is smaller and the gradient is gentler; during the half-flowering stage, the ear gradually becomes plump, and the gradient of radar backscattering of sparse ear-stalk crops still rises gently, and this rising process continues until the soft grain stage; therefore, the ear-stalk crops with a gentle gradient change in the saddle of the radar backscattering curve from the booting stage to the soft grain stage are determined as sparse ear-stalk crops SPC:

[0029]

[0030] In the formula, in function SPC, 1 represents sparse ear-stalk crops, 0 represents other types of crops, d s represents the soft grain stage, d b represents the booting stage, VV s represents the radar scattering value at the soft grain stage, VV b represents the radar scattering value at the booting stage, VV f represents the radar scattering value at the half-flowering stage, the M function represents whether it is an ear-stalk crop, θ1 takes a value of -5, and θ2 takes a value of -2.

[0031] In the step S01, based on the Sentinel-1 SAR time series dataset, data preprocessing is performed. In the step S05, the coverage of the plant canopy leaves has an important impact on radar backscattering. Broad-leaved crops including sorghum and corn have a large leaf coverage, and the radar backscattering is strong, while for narrow-leaved crops including rice and wheat, there are obvious gaps between the leaves, and some leaves grow upward with an obvious vertical structure, so the radar backscattering is weak; based on this feature, the sparse ear-stalk crops with relatively strong overall radar backscattering from the leaf growth stage to the booting stage are determined as sparse ear-stalk broad-leaved crops SPBC:

[0032]

[0033] In the formula, in the function SPBC, 1 represents sparse panicle-stem broad-leaved crops, 0 represents other crops, d b represents the booting stage, d v represents the visible flag leaf stage, j represents the date, and VV j represents the VV value corresponding to the date j, and θ3 takes the value of -11.

[0034] In the step S01, based on the Sentinel-1 SAR time series dataset, data preprocessing is performed. In the step S06, there is a layer of white wax on the epidermis of the stems and leaves of sorghum. Especially from the booting stage to the full growth stage, the content of white wax on the surface of sorghum leaves is very high; while for other sparse panicle-stem broad-leaved crops including corn, the wax content on the leaf epidermis is very low. Based on the time series data of the wax index, by measuring the wax content level from the booting stage to the full growth stage, the leaf epidermal wax index (Leaf epidermal wax index, abbreviated as LWI) of sorghum leaves is designed:

[0035]

[0036] In the formula, in the function LWI, 1 represents high-wax crops, 0 represents other crops, d b represents the booting stage, d g represents the full growth stage, d g -d b represents the time span from the booting stage to the full growth stage, k represents the accumulated temperature days, and WI k represents the value of the wax content spectral index WI corresponding to the accumulated temperature days k, θ4 takes the value of -0.1, and θ5 takes the value of 0.1.

[0037] In the step S01, based on the Sentinel-1 SAR time series dataset, data preprocessing is performed. In the step S07, from the hard grain stage to the maturity stage of sorghum, the tannin content in the sorghum panicle gradually increases, and the tannin content in the corn leaves and stamens is also significantly lower than that of sorghum; based on the time series data of the tannin index, by evaluating the tannin content level from the hard grain stage to the maturity stage, the red tassel spike index RTI of sorghum at maturity is designed:

[0038]

[0039] In the formula, in the function RTI, 1 represents high-tannin content crops, 0 represents other crops, d m represents the maturity stage, d h represents the hard grain stage, l represents the accumulated temperature days, and TI l represents the TI value corresponding to the accumulated temperature days l, and θ6 takes the value of -0.2.

[0040] In the step S01, based on the Sentinel-1 SAR time series dataset, data preprocessing is performed. In the step S08, the plant structure, wax content, and tannin content of sorghum are used as the coordinate axes of the multi-dimensional vector space, and each component of the sorghum body structure vector is calculated:

[0041] e SPBC =(1, 0, 0), e LWI =(0, 1, 0), e RTI =(0, 0, 1) (10)

[0042] V = e SPBC + e LWI + e RTI =(1, 1, 1) (11)

[0043] In the formula, e SPBC represents the plant structure sub-vector of the crop, e LWI represents the wax content sub-vector, e RTI represents the tannin content sub-vector, and V represents the body structure vector.

[0044] In the step S01, based on the Sentinel-1 SAR time series dataset, data preprocessing is performed. In the step S09, according to the crop body structure vector index constructed in the step S08, pixel-by-pixel discrimination is carried out, and the pixels with the body structure vector V being (1, 1, 1) are determined as sorghum.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] (1) The method is ingeniously conceived, making full use of the characteristics of the crop's plant structure, ear stalk density, wax content, and tannin content, etc., laying a foundation for realizing large-scale automatic mapping;

[0047] (2) It has strong interpretability. Sorghum has a series of characteristics such as tall plants, broad leaves, sparse ear stalks, high wax content in leaves, and high tannin content in ears. The present invention designs a series of crop indices from the intensity change of the radar scattering coefficient caused by the plant type and ear stalk density of the plant, and combines the numerical distribution of the wax and tannin index content in the key phenological periods of sorghum, comprehensively depicting the body structure characteristics of sorghum. The designed sorghum mapping method has interpretability and the ability of automatic migration and popularization and application;

[0048] (3) The method is simple and easy to implement, with strong robustness, high-precision automatic mapping ability, and low promotion and application cost. It makes full use of the publicly available and free Sentinel-1 SAR radar data and Sentinel-2 MSI multi-spectral time-series images, does not rely on training sample data for calibration in different regions or different years, has strong spatio-temporal migration and promotion application ability, and has the ability of large-scale multi-year low-cost or even zero-cost automatic application promotion. Brief Description of the Drawings

[0049] Figure 1 It is a flowchart for the implementation of the embodiment of the present invention.

[0050] Figure 2 It is a time-series signal diagram of radar scattering coefficients and vegetation indices of sorghum, wheat, soybean, and peanut.

[0051] Figure 3 It is a time-series signal diagram of vegetation indices and waxiness indices of sorghum and corn.

[0052] Figure 4 It is a time-series signal diagram of vegetation indices and tannin indices of sorghum and corn.

[0053] Figure 5 It is a spatial distribution map of sorghum in the study area. Detailed Embodiment

[0054] The technical solution of the present invention will be specifically described below in conjunction with the drawings.

[0055] Term Explanation:

[0056] The Sentinel-1 satellite belongs to an active microwave remote sensing satellite, consisting of two polar-orbiting satellites, Sentinel-1A and Sentinel-1B. The revisit period of the dual-satellite constellation observation is 6 days. The C-band sensor carried by it is not affected by natural conditions such as cloud cover and can achieve all-weather and all-time earth observation. Sentinel-1 has 4 unique scanning imaging modes. Among them, the IW (Interferometric Wide Swath mode) has a wide coverage range and high spatial resolution at the same time. This patent uses the level-1 GRD (Ground Range Detected) product in this mode. In addition, Sentinel-1 provides 4 polarization modes, namely VV, VH, HH, and HV. Among them, the microwave beam emitted by VV is vertically polarized, and the reflected energy received is also vertically polarized.

[0057] Sentinel-2 is the only optical satellite that has three red-edge bands closely related to vegetation physical and chemical parameters. It consists of two high-resolution multispectral imaging satellites, namely 2A and 2B. The revisit period of the dual-satellite network observation is 5 days. It covers a total of 13 bands from visible light, near-infrared to short-wave infrared. Among them, the spatial resolutions of the blue, green, red, and near-infrared bands are 10m, and the red-edge and short-wave infrared bands are 20m, which is very effective for monitoring vegetation growth and health status.

[0058] Optical Vegetation Index: The vegetation index is a factor that characterizes the growth status and spatial distribution density of vegetation. The commonly referred to vegetation index is the optical vegetation index. Common optical vegetation indices include NDVI and EVI2. NDVI is the Normalized Difference Vegetation Index, and EVI2 is the Enhanced Vegetation Index. The calculation formula for the EVI2 index is: where ρ RED , ρ NIR are the reflectances of the red and near-infrared bands of the Sentinel image, respectively.

[0059] Wax Index: The Wax Index (WI for short) is derived from the narrowband reflectances of blue and red light. It can be used as a representative of short-term changes in photosynthetic activity, stress conditions, and pigment absorption. It is also significantly correlated with net CO2 absorption and radiation utilization, and will change rapidly with irradiance and leaf physiological status. It has been explored as an index representing the wax content of crops. The calculation formula for the wax content spectral index (Wax Index, WI) is:

[0060]

[0061] Tanning Index: The tannin compounds contained in vegetation form a strong reflection peak in the short-wave infrared band. Therefore, based on the reflectance data of the green light and the second short-wave infrared band of the Sentinel-2 MSI image, the tanning index is designed. This index is very effective for detecting tannin and other compounds. The calculation formula for the tanning index (Tanning Index, TI for short) is:

[0062]

[0063] The present invention provides a method for intelligent identification of sorghum based on the crop structure vector, including the following steps (see Figure 1 ):

[0064] Step S01, establish an optical and radar time-series dataset for the study area;

[0065] Step S02: Obtain the key phenological periods of sorghum based on the optical and radar time series datasets;

[0066] Step S03: Design the saddle-shaped detection index of the vertical polarization radar for ear-stem crops;

[0067] Step S04: Design the sparse ear-stem canopy radar index;

[0068] Step S05: Design the sparse ear-stem broad-leaved crop index;

[0069] Step S06: Design the high-wax crop index;

[0070] Step S07: Design the red tassel spike index at the maturity stage;

[0071] Step S08: Comprehensively construct the crop structure vector index;

[0072] Step S09: Construct the sorghum intelligent recognition technology process method based on the crop structure vector;

[0073] Step S10: Obtain the spatial distribution map of sorghum planting in the study area.

[0074] The following are the specific embodiments of the present invention.

[0075] Step S01: Establish the optical and radar time series datasets of the study area

[0076] Based on the Sentinel-1 SAR time series dataset, perform data preprocessing on the original data, including radar data denoising based on Lee filtering, image mosaicking, and time series data smoothing based on WS; obtain the Sentinel-2 MSI time series dataset and perform preprocessing operations: cloud recognition and cloud removal, image mosaicking, remote sensing index calculation, linear interpolation, and WS time series smoothing. Construct the optical and radar index time series sets.

[0077] Step S02: Obtain the booting stage, half-flowering stage, soft grain stage, and maturity stage of sorghum

[0078] Based on the optical and radar time series datasets of sorghum, determine the key phenological periods of sorghum. The dates corresponding to the first valley value, peak value of the vegetation index time series curve, and the last peak value of the tannin index time series curve during the crop growth period are respectively determined as the emergence stage d e 、the vigorous growth stage d g 、the maturity stage d m ; determine the visible flag leaf stage d v forty days before the vigorous growth stage, and determine the hard grain stage d h thirty days before the maturity stage; further combine the radar time series curve to determine the half-flowering stage d f 、the booting stage d b 、the soft grain stage ds 。 Its calculation formula is:

[0079]

[0080] In the above formula, t represents time, and VV t represents the VV value at the time of the corresponding day of the year t; when F(t) = 1, the corresponding extreme values are summarized into the set I = {t1, t2, t3, t4...}. Based on the elements in the set I, the key phenological periods of sorghum are further judged; the date of the minimum value closest to the growth peak period is determined as the half-flowering period; the dates of the maximum values closest to the front and back of the half-flowering period are determined as the booting period and the soft grain period in sequence.

[0081] d f = d g - min{|d h - t1|, |d h - t2|, |d h - t3|...}, (VV t - VV t-1 ) < 0 (2)

[0082] d b = d f - min{(d f - t1), (d f - t2), (d f - t3)...}, d f > t n (3)

[0083] d s = d f - max{(d f - t1), (d f - t2), (d f - t3)...}, d f < t n (4)

[0084] In the above formula, d e represents the emergence period, d m represents the maturity period, d f represents the half-flowering period, d g represents the growth peak period, d h represents the hard grain period, d b represents the booting period, d s represents the soft grain period.

[0085] Step S03: Design the vertical polarization radar saddle-shaped detection index for ear-stem crops

[0086] During the leaf growth process of panicle-stem crops, the number of leaves increases and gradually covers the land. At this time, the surface reflection of the radar from the leaves gradually becomes dominant, and strong scattering will form on the canopy surface. When the panicle-stem begins to grow, due to the obvious vertical structure of the panicle-stem, the radar backscattering attenuation is more obvious, causing the radar backscattering coefficient to decrease, reaching a local minimum value at the half-flowering stage of the panicle-stem crops; as the panicle-stem crops further grow and develop, the panicles of the panicle-stem crops gradually become plump, and the radar backscattering coefficient gradually increases, reaching a local peak at the soft grain stage; from the soft grain stage to the mature stage, the leaves of the panicle-stem crops gradually wither and droop, the leaf density and dielectric constant of the canopy gradually decrease, and the vertical structure of the panicle-stem crop plants is very obvious, directly leading to a gradual decrease in the radar backscattering coefficient. Therefore, the radar scattering coefficient of panicle-stem crops shows a saddle shape during the growth period. For non-panicle crops such as soybeans, peanuts, and potatoes, during their growth period, the radar reflection signal mainly goes through two stages, that is, as the leaves grow, the radar backscattering increases, and as the leaves wither, the radar backscattering decreases, only showing a shape of first rising and then falling on the time series trajectory, without the saddle shape characteristic.

[0087]

[0088] In the above formula, 1 in the function M represents panicle-stem crops, and 0 represents non-panicle-stem crops; d b represents the booting stage, d g represents the full growth stage, d s represents the soft grain stage, card(I) represents the number of elements contained in the set I. When card(I) ≥ 3, it means that there are three or more extreme points, representing that the spectral time series curve has a saddle shape, and it is determined whether the saddle shape is located in the full growth stage through d b <d g <d s

[0089] Step S04, design a sparse panicle-stem canopy radar index

[0090] During the booting stage, the density of the panicle-stem has an important impact on the radar backscattering. The denser the panicle-stem, the more obvious the vertical structure, and the more the radar scattering weakens. Therefore, compared with dense panicle-stem crops such as wheat, the degree of decrease in the radar backscattering of sparse panicle-stem crops is smaller and the gradient is gentler; during the half-flowering stage, the panicles gradually become plump, and the rising gradient of the radar backscattering of sparse panicle-stem crops remains gentle, and this rising process continues until the soft grain stage. Therefore, the panicle-stem crops with a gentle gradient change at the saddle part of the radar backscattering curve from the booting stage to the soft grain stage are determined as sparse panicle-stem crops (SPC).

[0091]

[0092] ​In the above formula, in the function SPC, 1 represents sparse panicle-stem crops, 0 represents other types of crops, d s represents the soft grain stage, d b represents the booting stage, VV s represents the radar scattering value at the soft grain stage, VV b represents the radar scattering value at the booting stage, VV f represents the radar scattering value at the half-flowering stage, the M function represents whether it is a panicle-stem crop, and it is recommended that θ1 take the value of -5 and θ2 take the value of -2.

[0093] Step S05: Design the sparse panicle-stem broadleaf crop index

[0094] The coverage of the plant canopy leaves has an important impact on the radar backscattering. The leaves of broadleaf crops such as sorghum and corn have a large coverage, and the radar backscattering is strong. However, there are obvious gaps between the leaves of narrow-leaf crops such as rice and wheat, and some leaves grow upward with an obvious vertical structure, so the radar backscattering is weak. Based on this characteristic, the sparse panicle-stem crops with relatively strong overall radar backscattering from the leaf growth stage to the booting stage are determined as sparse panicle-stem broadleaf crops (SPBC).

[0095]

[0096] In the above formula, in the function SPBC, 1 represents sparse panicle-stem broadleaf crops, 0 represents other crops, d b represents the booting stage, d v represents the visible flag leaf stage, d b -d v represents the time span from the visible flag leaf stage to the booting stage, j represents the date, VV j represents the VV value corresponding to the date j, and it is recommended that θ3 take the value of -11.

[0097] Step S06: Design the high-wax crop index

[0098] Sorghum is an annual C4 plant of the Gramineae family, with the characteristics of drought tolerance and heat tolerance. The reason is not only that its root system is developed and it has a strong ability to absorb water from the soil, but also because there is a layer of white wax on the epidermis of the stem and leaves of sorghum, which can reduce sensitivity during drought. The presence of epidermal wax will affect the reflectivity of the leaf surface. The period from the visible flag leaf stage to the growth peak stage of sorghum is the period with the highest content of white wax on the leaf surface, while the content of epidermal wax on the leaves of crops such as corn is very low. Therefore, the overall wax index of sorghum leaves is relatively high compared to crops such as corn. Based on this characteristic, the leaf epidermal wax index (LWI) of sorghum leaves is designed.

[0099]

[0100] In the above formula, in the function LWI, 1 represents high-waxy crops, 0 represents other crops, d b represents the booting stage, d g represents the full growth stage, d g -d b represents the time span from the booting stage to the full growth stage, k represents the accumulated temperature days, and WI k represents the value of the waxy content spectral index WI at the corresponding accumulated temperature days k. It is recommended that θ4 takes the value of -0.1 and θ5 takes the value of 0.1.

[0101] Step S07: Design the red tassel spike index at maturity

[0102] All sorghums contain tannin substances. During the life cycle of sorghum, as sorghum gradually matures, the dry grain weight and dry matter of sorghum spikes increase significantly. At the hard grain stage, the biomass of the spike reaches 75% of its final dry weight, and nutrient absorption is almost complete. From the hard grain stage to the maturity stage of sorghum, the spike gradually dominates the spectral reflection of the canopy. The tannin index has an obvious effect on detecting the tannin content of sorghum spikes. The tassel of maize is located at the top of the plant, mainly composed of protein, lipid, etc., and the tassel has a small volume and a greenish-red color. Even at the mature stage, the tannin index of maize leaves and tassels is not high. Based on this characteristic, the red tassel spike index (RTI) at the maturity stage of sorghum is designed.

[0103]

[0104] In the above formula, in the function RTI, 1 represents high-tannin-content crops, 0 represents other crops, d m represents the maturity stage, d h represents the hard grain stage, l represents the accumulated temperature days, and TI l represents the value of TI at the corresponding accumulated temperature days l. It is recommended that θ6 takes the value of -0.2.

[0105] Step S08: Comprehensively construct the crop body structure vector index

[0106] Take the plant structure, waxiness, and tannin content of sorghum as the coordinate axes of a multi-dimensional vector space, and calculate each component of the sorghum body structure vector.

[0107] e SPBC =(1, 0, 0), e LWI =(0, 1, 0), e RTI =(0, 0, 1) (10)

[0108] V = e SPBC + e LWI + e RTI=(1, 1, 1) (11)

[0109] In the above formula, e SPBC represents the plant structure component of sorghum, e LWI represents the wax content component of sorghum, e RTI represents the tannin content component of sorghum. In formula (11), V represents the volume structure vector of sorghum.

[0110] Step S09: Construct a sorghum intelligent recognition technology process method based on the crop volume structure vector

[0111] According to the rules established in step S08, the optical and radar time-series curves of all candidate crops are discriminated pixel by pixel, and those with the volume structure vector V of (1, 1, 1) are determined as sorghum.

[0112] Step S10: Obtain the sorghum planting spatial distribution map of the study area

[0113] Taking a certain town in a certain city in a certain province as an example, according to the above step method provided in this embodiment, the sorghum spatial distribution map of the study area is obtained (see Figure 5 ).

[0114] The above is the preferred embodiment of the present invention. All changes made according to the technical solution of the present invention, when the functions and effects produced do not exceed the scope of the technical solution of the present invention, shall fall within the protection scope of the present invention.

Claims

1. A sorghum intelligent recognition method based on the crop structure vector, characterized in that It includes the following steps: Step S01: Establish the optical and radar time-series datasets for the study area; Step S02: Obtain the key phenological periods of sorghum based on the optical and radar time-series datasets; Step S03: Design the vertical polarization radar saddle-shaped detection index for ear-stem crops; Step S04: Design the sparse ear-stem canopy radar index; Step S05: Design the sparse ear-stem broad-leaved crop index; Step S06: Design the high-wax crop index; Step S07: Design the red tassel ear index at the maturity stage; Step S08: Comprehensively construct the crop body structure vector index; Step S09: Construct the sorghum intelligent recognition technology process method based on the crop body structure vector; Step S10: Obtain the sorghum planting spatial distribution map of the study area.

2. The sorghum intelligent recognition method based on the crop structure vector according to claim 1, characterized in that In the step S01, based on the Sentinel-1 SAR time-series dataset, data preprocessing is carried out, including radar data denoising based on Lee filtering, image mosaicking, and time-series data smoothing; at the same time, the Sentinel-2 MSI time-series dataset is obtained, and preprocessing operations are carried out, including cloud recognition and cloud removal, image mosaicking, remote sensing index calculation, linear interpolation, and time-series data smoothing, to construct the smooth optical and radar time-series datasets for the study area.

3. The sorghum intelligent recognition method based on the crop structure vector according to claim 1, characterized in that, In the step S02, based on the optical and radar time series datasets, the key phenological periods of sorghum are determined; the dates corresponding to the first valley value, the peak value of the vegetation index time series curve, and the last peak value of the tannin index time series curve during the crop growth period are respectively determined as the emergence stage d e , the full growth stage d g , and the maturity stage d m ; the first forty days before the full growth stage are determined as the visible flag leaf stage d v , and the thirty days before the maturity stage are determined as the hard grain stage d h ; further, the half-blooming stage d f , the booting stage d b , and the soft grain stage d s are determined in combination with the radar time series curve; the calculation formula is: where t represents time, and VV t represents the radar scattering value at the time of the corresponding day of the year t, i.e., the VV value; when F(t) = 1, the corresponding extreme values are summarized into a set I = {t1, t2, t3, t4...}; based on the elements in the set I, the key phenological periods of sorghum are further judged; the date of the minimum value closest to the growth peak period is determined as the semi-blooming period; the dates of the maximum values closest to the front and back of the semi-blooming period are determined as the booting period and the soft grain period in sequence; d f = d g - min{|d h - t1|, |d h - t2|, |d h - t3|...}, (VV t - VV t-1 ) < 0 (2) d b = d f - min{(d f - t1), (d f - t2), (d f - t3)...}, d f > t n (3) d s = d f - max{(d f - t1), (d f - t2), (d f - t3)...}, d f <t n (4) Where, VV t -VV t-1 Determine the extreme value direction at the corresponding moment, d e Represents the emergence stage, d m Represents the maturity stage, d f Represents the semi-blooming stage, d g Represents the vigorous growth stage, d h Represents the hard grain stage, d b Represents the booting stage, d s Represents the soft grain stage.

4. The sorghum intelligent recognition method based on the crop structure vector according to claim 3, wherein, In the step S03, during the leaf growth process of ear-stem crops, the number of leaves increases and gradually covers the land, and a strong scattering will be formed on the canopy surface. When the crop enters the booting stage, the ear-stem and ear block part of the radar backscattering from the leaves, and this attenuation of the backscattering is more obvious. When the ear gradually becomes plump, the number of scattering particles from the ear increases, and the radar backscattering gradually increases. Until after entering the soft grain stage, the leaves gradually droop and wither from bottom to top, the radar backscattering energy decays, and the VV polarization continuously decreases; therefore, the VV polarization time-series curve of ear-stem crops shows a saddle-shaped feature, and the saddle of the saddle shape is located in the peak growth period of the crop; for non-ear-stem crops including soybeans, peanuts, and potatoes, the radar reflection time-series curve only shows a shape that first rises and then falls: In the formula, 1 in the function M represents spike-straw crops, and 0 represents non-spike-straw crops; d b represents the booting stage, d g represents the full-growth stage, d s represents the soft-grain stage, card(I) represents the number of elements included in the set I, and when card(I) ≥ 3, it means that there are three or more extreme values.

5. The sorghum intelligent recognition method based on the crop structure vector according to claim 1, characterized in that In the step S04, during the booting stage, the density of the ear-stem has an important impact on the radar backscattering. The denser the ear-stem, the more obvious the vertical structure, and the more the radar scattering weakens. Therefore, compared with dense ear-stem crops including wheat, the degree of decrease in the radar backscattering of sparse ear-stem crops is smaller and the gradient is gentler; during the semi-flowering stage, the ear gradually becomes plump, and the radar backscattering gradient of sparse ear-stem crops still rises gently, and this rising process continues until the soft grain stage; therefore, the ear-stem crops with a gentle gradient change at the saddle of the radar backscattering curve from the booting stage to the soft grain stage are determined as sparse ear-stem crops SPC: In the formula, in the function SPC, 1 represents sparse spike-stalk crops and 0 represents other types of crops, d s represents the soft grain stage, d b represents the booting stage, VV s represents the radar scattering value at the soft grain stage, VV b represents the radar scattering value at the booting stage, VV f represents the radar scattering value at the semi-blooming stage, the M function represents whether it is a spike-stalk crop, θ1 takes the value of -5, and θ2 takes the value of -2.

6. The sorghum intelligent recognition method based on the crop structure vector according to claim 1, wherein, In the step S05, the coverage of the leaves in the plant canopy has an important impact on the radar backscattering. The leaves of broad-leaved crops including sorghum and corn have a large coverage, and the radar backscattering is strong, while the gaps between the leaves of narrow-leaved crops including rice and wheat are obvious, and some leaves grow upward, with an obvious vertical structure, so the radar backscattering is weak; based on this feature, the sparse ear-stem crops with relatively strong overall radar backscattering from the leaf growth period to the booting stage are determined as sparse ear-stem broad-leaved crops SPBC: In the formula, 1 in the function SPBC represents sparse panicle stalk broad-leaved crops, 0 represents other crops, d b represents the booting stage, d v represents the visible stage of the flag leaf, j represents the date, and VV j represents the VV value corresponding to the date j, and the value of θ3 is -11.

7. The sorghum intelligent recognition method based on the crop structure vector according to claim 1, characterized in that In the step S06, there is a layer of white wax on the epidermis of the stems and leaves of sorghum. Especially from the booting stage to the peak growth stage, the content of white wax on the surface of sorghum leaves is very high; while for other sparse ear-stemmed broad-leaved crops including corn, the wax content on the leaf epidermis is very low. Based on the time-series data of the wax index, by measuring the wax content level from the booting stage to the peak growth stage, the leaf epidermal wax index of sorghum, abbreviated as LWI, is designed: In the formula, 1 in the function LWI represents high-wax crops, 0 represents other crops, d b represents the booting stage, d g represents the full growth stage, d g -d b represents the time span from the booting stage to the full growth stage, k represents the accumulated temperature days of the year, WI k represents the value of the wax content spectral index WI at the accumulated temperature days k of the corresponding year, θ4 takes a value of -0.1, and θ5 takes a value of 0.

1.

8. The sorghum intelligent recognition method based on the crop body structure vector according to claim 1, characterized in that, In the step S07, from the hard grain stage to the mature stage of sorghum, the tannin content in the sorghum ear gradually increases, and the tannin content in the leaves and stamens of corn is also significantly lower than that of sorghum; based on the time-series data of the tannin index, by evaluating the tannin content level from the hard grain stage to the mature stage, the red tassel ear index RTI at the mature stage of sorghum is designed: In the formula, in the function RTI, 1 represents crops with high tannin content, 0 represents other crops, d m represents the mature stage, d h represents the hard grain stage, l represents the accumulated day of the year, TI l represents the TI value corresponding to the accumulated day of the year l, and the value of θ6 is -0.

2.

9. The sorghum intelligent recognition method based on the crop body structure vector according to claim 1, characterized in that, In the step S08, the plant structure, wax and tannin content of sorghum are used as the coordinate axes of the multi-dimensional vector space, and each component of the sorghum body structure vector is calculated: e SPBC =(1, 0, 0), e LWI =(0, 1, 0), e RTI =(0, 0, 1) (10) V = e SPBC + e LWI + e RTI = (1,1,1) (11) where, e SPBC represents the plant structure sub-vector of the crop, e LWI represents the wax content sub-vector, e RTI represents the tannin content sub-vector, and V represents the body structure vector.

10. The sorghum intelligent recognition method based on the crop structure vector according to claim 1, wherein In the step S09, according to the crop body structure vector index constructed in the step S08, pixel-by-pixel discrimination is carried out, and the pixel with the body structure vector V being (1, 1, 1) is determined as sorghum.

Images