A rice early remote sensing identification method based on planting probability

Abstract

The application discloses a rice early remote sensing identification method based on planting probability, which comprises the following steps: step 1, selecting an image time window; step 2, image segmentation to obtain a land object; step 3, extracting an object-level time sequence characteristic parameter; step 4, distinguishing and marking a vegetation area and a non-vegetation area; step 5, calculating a rice planting probability under different time sequence characteristics; step 6, calculating a rice probability index TRPI based on the time sequence characteristics; and step 7, threshold classification to extract a rice planting distribution. The application extracts time sequence characteristics of rice transplanting to tillering by using SAR image data, constructs a formula to calculate the probability of planting rice under each time sequence characteristic index, simultaneously marks a non-rice planting area by using a vegetation index calculated by using multi-spectral data of a rice previous crop in a vigorous period, thereby integrally constructing a rice probability index TRPI based on the time sequence characteristics, and finally realizes early rice identification by using a threshold two-classification for the TRPI calculation result.

Description

Technical Field

[0001] The invention belongs to the technical field of remote sensing monitoring and identification, and in particular relates to an early rice remote sensing identification method based on planting probability. Background Art

[0002] Keeping abreast of rice planting distribution is crucial for agricultural applications such as verifying crop subsidy applications, monitoring crop growth, and assessing insurable crop areas by agricultural insurance institutions, improving coverage, and verifying insurance authenticity. With the rapid development of remote sensing technology in recent years, research on its applications in agriculture has continued to deepen. Currently, there is a growing body of research on crop identification and extraction using remote sensing, such as for wheat, rice, corn, and soybeans. Most related research utilizes remote sensing data features from the entire crop growth period or during peak growth phases. Research and inventions on crop identification during planting and early growth stages are relatively limited, especially for early rice identification, for which no research has been found. During the early rice planting period, weather conditions affect the quality of available optical images, resulting in limited data and low biomass. Therefore, identifying rice plants in this early stage is challenging.

[0003] Synthetic airborne radar (SAR) data is sensitive to water and unaffected by clouds and rain. Its high temporal resolution offers advantages for agricultural applications. Rice cultivation typically involves land preparation, field flooding, transplanting, and rice growth (tillering, jointing, heading, heading, and maturation). During rice planting and early growth stages, field water content and surface water visibility change, providing a basis for capturing rice cultivation conditions using multi-temporal SAR data. Therefore, time-series SAR imagery can serve as an effective data source for early rice extraction. Summary of the Invention

[0004] In view of the above-mentioned deficiencies in the prior art, the purpose of the present invention is to provide a method for early remote sensing identification of rice based on planting probability. The present invention proposes a rice probability index (TRPI) based on time series characteristics, forming an early identification method for rice. The present invention uses SAR image data to extract the time series characteristics of rice transplanting to tillering period, calculates the probability of planting rice under each time series characteristic index, and uses the vegetation index calculated by multispectral image data of the vigorous period (history) of the previous rice crop to mark obvious non-rice planting areas, thereby integrating and constructing TRPI, and finally realizing early rice identification by using threshold binary classification. It solves the problem of difficulty in remote sensing identification and extraction of rice in the early stage of rice planting and growth.

[0005] In order to achieve the above-mentioned object of the invention, the technical solution adopted by the present invention is:

[0006] A method for remote sensing identification of early rice based on planting probability is provided, the method comprising the following steps:

[0007] Step 1: Select the image time window

[0008] The rice planting and early growth period (rice transplanting to tillering, generally from early June to mid-July) was selected as the time window for early rice identification. Time-series SAR imagery (Sentinel-1) within this time window was selected as the data source for early rice identification, with the interval between adjacent images being 5-7 days. The time-series SAR images were preprocessed to obtain the backscatter coefficient in VH polarization mode for each SAR image of the target area.

[0009] Step 2: Image segmentation to obtain land objects

[0010] Using Sentinel-2 multispectral images of the previous rice crop during its peak period, we performed multi-scale segmentation to obtain plot objects, and set the segmentation scale and various factors based on the image resolution and area size.

[0011] Step 3: Extract object-level temporal feature parameters

[0012] The average backscatter coefficient of all pixels in each plot of SAR VH polarization image is calculated as the backscatter coefficient of the plot object. Then, the maximum value (VH) of the same plot object between time series images is extracted. Max ), minimum value (VH Min ), mean(VH Mean ), and finally generate the time series maximum value graph, time series minimum value graph, and time series mean value graph;

[0013] Step 4: Distinguish and mark vegetated and non-vegetated areas

[0014] Calculate the Normalized Difference Vegetation Index (NDVI) using multispectral images, set thresholds to distinguish and mark vegetation and non-vegetation areas;

[0015] Step 5: Calculate the probability of rice planting under different time series characteristics

[0016] Calculate the rice planting probability of each plot object in the time series SAR data under the time series maximum value, time series minimum value, and time series mean value characteristics respectively;

[0017] Step 6: Calculate the rice probability index TRPI based on time series characteristics

[0018] A rice probability index (TRPI) based on time series characteristics was constructed and used to calculate the probability of early rice planting.

[0019] Step 7: Threshold classification to extract rice planting distribution

[0020] The condition TRPI≥T1 is set. When the condition is true, it is rice, otherwise it is non-rice, thus realizing early rice identification.

[0021] Furthermore, in step 1, the time series SAR images are subjected to orbit correction, thermal noise removal, radiation calibration, filtering, terrain correction, decibel processing, registration, and cropping preprocessing to obtain the backscattering coefficient of each SAR image of the target area under the VH polarization mode.

[0022] Furthermore, the method of step 4 specifically includes the following sub-steps:

[0023] Step 4-1, according to the formula:

[0024]

[0025] Get NDVI; where ρ Red , ρ NIR are the reflectances in the red and near-infrared bands, respectively;

[0026] Step 4-2, according to the function:

[0027]

[0028] Set the condition NDVI ≥ T0. When the condition is true, it indicates a vegetation area, that is, an area where rice may be planted, and is marked as 1; when the condition is false, it indicates a non-vegetation area, that is, an area that is obviously not planted with rice, and its probability of rice planting is 0, and is marked as 0.

[0029] Furthermore, in step 5, according to the formula:

[0030]

[0031] Get the probability f(VH) that the object to be calculated is rice under different time series characteristic indicators i );in, is the mean of rice samples under the current time series characteristics, x i is the backscattering coefficient value of the object to be calculated in the current time series characteristic index graph, (VH i ) max 、(VH i ) min They are the maximum and minimum values ​​after removing the non-vegetation area from the current time series characteristic index map; It is used to reflect the difference between the object to be identified and rice. The smaller the value, the smaller the difference, that is, the greater the possibility that it is rice. Normalize it, Indicates the probability that the object to be calculated is rice. The larger the value, the greater the probability that it is rice. f(VH i) ranges from 0 to 1. When i takes Max, Min, or Mean, according to the formula:

[0032]

[0033]

[0034]

[0035] Obtain the probability of rice planting f(VH under the time series maximum, minimum and mean characteristic indicators Max )、f(VH Min )、f(VH Mean ).

[0036] Furthermore, in step 6, according to the formula:

[0037] TRPI=f()×f( Max )×f(VH Min )×f(VH Mean )

[0038] Get the rice probability index TRPI; where f(NDVI) is the vegetation and non-vegetation areas marked in step 4, with values ​​of 1 and 0, and f(VH Max )、f(VH Min )、f(VH Mean ) are the rice planting probabilities under the time series maximum, minimum, and mean characteristics calculated in step 5, respectively.

[0039] The beneficial effects of the present invention are:

[0040] 1. In the early stage of rice planting (tillering stage), the changes in rice growth and the changes in the water surface appearance of paddy fields during rice planting and early growth stages are used as breakthrough points. Synthetic aperture radar data, which is not affected by weather, is used to identify rice, which has certain advantages in the timeliness of rice identification.

[0041] 2. Created the rice probability index TRPI based on time series characteristics and constructed three characteristic indicators f(VH Max )、f(VH Min )、f(VH Mean ) to reflect the possibility of rice planting under different temporal characteristics, and the final integration can be used for early identification of rice. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 Flow chart of the method of the present invention;

[0043] Figure 2 A time series maximum value distribution diagram provided by an embodiment of the present invention;

[0044] Figure 3 A time series minimum value distribution diagram provided by an embodiment of the present invention;

[0045] Figure 4 A time series mean distribution diagram provided by an embodiment of the present invention;

[0046] Figure 5 A diagram showing the TRPI calculation results provided by an embodiment of the present invention;

[0047] Figure 6 This is a distribution diagram of rice extraction results provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0048] The specific embodiments of the present invention are described below to facilitate understanding of the present invention by those skilled in the art. However, it should be clear that the present invention is not limited to the scope of the specific embodiments. For those skilled in the art, as long as various changes are within the spirit and scope of the present invention as defined and determined by the appended claims, these changes are obvious, and all inventions and creations utilizing the concepts of the present invention are protected.

[0049] Example

[0050] In this embodiment of the present invention, part of Dafeng District, Yancheng City, Jiangsu Province was selected as the research area, Sentinel-1 images were used as the data source for early rice identification, and Sentinel-2 images were used as the data source for object-oriented segmentation and calculation of NDVI. Figure 1 , provides a method for remote sensing identification of early rice based on planting probability, the method comprising the following steps:

[0051] Step 1: Image time window selection

[0052] Rice transplanting or direct seeding generally begins in early June in the Dafeng area of ​​Yancheng. Most rice is planted by the end of June, with a small number completing the process in early July. Six Sentinel-1GRD imagery (SAR data) from June 11 to July 12, 2021, were selected for early rice identification. The images were preprocessed using orbit correction, thermal noise removal, radiometric calibration, filtering, terrain correction, decibel processing, registration, and cropping. The backscatter coefficients for each SAR image were obtained for the target area using VH polarization.

[0053] Step 2: Image segmentation to obtain land objects

[0054] Multi-scale segmentation was performed using the Sentinel-2 image on May 8, 2021, to obtain plot objects. The segmentation scale was set to 80, and the shape factor and complexity were set to 0.6 and 0.4, respectively.

[0055] Step 3: Object-level temporal feature parameter extraction

[0056] Calculate the object-level backscatter coefficient of each phase of the image, and extract the maximum value of the backscatter coefficient of the plot object between the time series images (VH Max ), minimum value (VH Min ), mean(VH Mean ), get the maximum value map, minimum value map, and mean value map, such as Figure 2-4 shown.

[0057] Step 4: Distinguish and mark vegetated and non-vegetated areas

[0058] NDVI is calculated using the ratio of the difference and sum between the near-infrared and red bands of the Sentinel-2 image. The condition NDVI ≥ 0.55 is set. When the condition is true, it is marked as 1, and when the condition is false, it is marked as 0. The values ​​under different conditions are expressed using f(NDVI). That is:

[0059]

[0060] Step 5: Calculate the probability of rice planting under each time series feature

[0061] According to the extraction results of step 3, calculate the f(VH Max )、f(VH Max )、f(VH Mean ), the calculation formula is as follows:

[0062]

[0063]

[0064]

[0065] In the above formula, They are -18.1448, -23.8121, and -21.0849 respectively; (VH Max ) max 、(VH Max ) min 8.8171, -26.3483 respectively; (VH Min ) max 、(VH Min ) min -10.8363, -55.0572 respectively; (VH Mean ) max, (VH Mean ) mix -6.2737, -32.3642, respectively.

[0066] Step 6: Calculate the rice probability index TRPI based on time sequence characteristics

[0067] According to the calculation results of step 4 and step 5, TRPI is calculated. The calculation formula is as follows, and the TRPI calculation result is shown in Figure 5 .

[0068] TRPI = f() x f( Max ) x f(VH Min ) x f(VH Mean )

[0069] Step 7: Threshold classification, extract rice planting distribution

[0070] Set the condition TRPI >= 0.8, when the condition is true, it is rice, and when the condition is false, it is non-rice. The early rice extraction result is shown in Figure 6 .

[0071] The present application is in the early stage of rice planting (tillering stage), taking the change characteristics of rice planting and growth in the early stage of rice growth and the change characteristics of water body on the surface of paddy field as the breakthrough, using the synthetic aperture radar (SAR) data which is not affected by the weather to identify the rice, and has certain advantages in the timeliness of rice identification. The rice probability index TRPI based on time sequence characteristics is created, and three feature indexes f(VH Max ), f(VH Min ) and f(VH Mean ) are constructed to reflect the possibility of rice planting under different time sequence characteristics, and finally the early identification of rice can be integrated.

[0072] It is obvious to those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting, and the scope of the present application is defined by the appended claims rather than the above description, and therefore all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be included in the present application.

[0073] In addition, it should be understood that although the present specification is described in terms of embodiments, each embodiment does not contain only one independent technical solution, and the description manner of the specification is only for clarity, and those skilled in the art should regard the specification as a whole, and the technical solutions in each embodiment can also be properly combined to form other embodiments which can be understood by those skilled in the art.

Claims

1. A method for early rice remote sensing identification based on planting probability, characterized in that: The following steps are involved: Step 1: Select the image time window The time window from rice transplanting to tillering was selected as the early rice identification time window. The time series SARSentinel-1 image data within this time window was selected as the data source for early rice identification. The time series SAR images were preprocessed to obtain the backscatter coefficient of each SAR image in the target area under the VH polarization mode. Step 2: Image segmentation to obtain land objects Using Sentinel-2 multispectral images of the previous rice crop during its peak period, we performed multi-scale segmentation to obtain plot objects, and set the segmentation scale and various factors based on the image resolution and area size. Step 3: Extract object-level temporal feature parameters Calculate the average backscatter coefficient of all pixels in each plot object of each SAR VH polarization image as the backscatter coefficient of the plot object, and then extract the maximum VH of the same plot object between time series images. Max , minimum value VH Min , mean VH Mean , and finally generate the time series maximum value graph, time series minimum value graph, and time series mean value graph; Step 4: Distinguish and mark vegetation and non-vegetation areas Calculate the Normalized Difference Vegetation Index (NDVI) using multispectral images, set thresholds to distinguish and mark vegetation and non-vegetation areas; Step 5: Calculate the probability of rice planting under different time series characteristics Calculate the rice planting probability of each block object in the time series SAR data under the time series maximum value, time series minimum value and time series mean value characteristics respectively; Step 6: Calculate the rice probability index TRPI based on time series characteristics Constructing a rice probability index based on time series characteristics , using this index to calculate the probability of early rice planting; Step 7: Threshold classification to extract rice planting distribution Setting conditions ≥T1, when the condition is true, it is rice, otherwise it is non-rice, achieving early rice identification; In step 5, the probability that the object to be calculated is rice under different time series characteristic indicators is The calculation expression is: ； in, is the mean of rice samples under the current time series characteristics, is the backscattering coefficient value of the object to be calculated in the current time series characteristic index graph, 、 are the maximum and minimum values ​​after removing the non-vegetation area from the current time series characteristic index map; when i takes Max, Min, and Mean, The probability of rice planting under the characteristic indicators corresponding to the maximum, minimum and mean values ​​of the time series respectively; In step 6, according to the formula: ； Calculating the rice probability index ;in, are the vegetation and non-vegetation areas marked in step 4, with values ​​of 1 and 0, are the rice planting probabilities under the maximum, minimum, and mean time series features calculated in step 5, respectively.

2. The method for early rice remote sensing identification based on planting probability according to claim 1, characterized in that: In step 1, the time series SAR images are pre-processed by orbit correction, thermal noise removal, radiation calibration, filtering, terrain correction, decibel processing, registration, and cropping to obtain the backscattering coefficient of each SAR image in the target area under the VH polarization mode.

3. The method for early rice remote sensing identification based on planting probability according to claim 1, characterized in that: The method of step 4 specifically includes the following sub-steps: Step 4-1, according to the formula: ； Get NDVI; where, 、 are the reflectances in the red and near-infrared bands, respectively; Step 4-2, according to the function: ； Set the condition NDVI ≥ T0. When the condition is true, it indicates a vegetation area, that is, an area where rice may be planted, and is marked as 1; when the condition is false, it indicates a non-vegetation area, that is, an area that is obviously not planted with rice, and its probability of rice planting is 0, and is marked as 0.

4. The method for early rice remote sensing identification based on planting probability according to claim 1, characterized in that: In step 5, when i takes Max, Min, or Mean, according to the formula: ； ； ； Obtain the probability of rice planting under the time series maximum, minimum, and mean characteristic indicators .

Images