A method for evaluating yield reduction due to heat stress during rice filling period using time-series images

Abstract

The present invention discloses a method for evaluating the yield reduction of rice during the grain filling period by using time series images, which relates to the field of digital monitoring and yield estimation of crop production. The method includes: constructing a time series curve of growth appearance characteristics; identifying the impact of heat stress on rice; fitting the coefficients of the yield estimation model; and estimating the yield loss to achieve the yield prediction of rice compared with normal years. The method solves the problem of quantitatively estimating the yield of rice growth monitoring time series images, and effectively realizes automatic identification of heading and flowering period, monitoring growth during the grain filling period, and evaluating the impact of heat stress and yield loss.

Description

Technical Field

[0001] The invention relates to the field of digital monitoring and yield estimation of crop production, and in particular to a method for evaluating yield reduction due to heat stress during rice filling period using time-series images. Background Art

[0002] As a staple food crop, the stability of rice yield is crucial to meet the needs of a growing population. However, with global warming, the frequency, duration and intensity of extreme high temperature weather have increased significantly, which has an adverse effect on the growth and development of rice, reducing the yield and quality of rice. In the growth cycle of rice, the heading and flowering stages and the grain filling stage are most sensitive to high temperature weather and are also the key periods for yield formation. High temperatures during the heading and flowering stages will hinder pollen maturation, anther cracking, pollen germination on the stigma, and pollen tube elongation, affecting the pollination and fertilization process and reducing the rice fruiting rate; high temperatures during the grain filling stage will weaken the activity of phosphorylase and starch in the grains, reduce the filling rate and photosynthesis efficiency, affect the synthesis and maintenance of chlorophyll, shorten the rice filling period, and damage dry matter accumulation and thousand-grain weight, directly reducing rice yield. By monitoring the growth status of rice, accurately identifying the heading, flowering and filling periods, judging the impact of heat stress, establishing a yield model and coefficient fitting, estimating the yield loss under heat stress, and realizing rice yield prediction compared to normal years, it provides an important reference for subsequent production management and yield forecasting.

[0003] Current rice production management practices mainly rely on manual field observations, and lack methods to automatically identify and judge rice growth conditions. Rice yield predictions are mainly based on crop models or combined with remote sensing (satellite or drone data), which requires a large number of model input parameters and does not fully consider the impact of heat stress on rice yield. Summary of the invention

[0004] The purpose of the present invention is to provide a method for evaluating the yield reduction due to heat stress in the filling period of rice by using time series images. The rice heading period is detected by a deep learning method, and the rice heading and filling periods are extracted by using a constructed growth epigenetic characteristic time series curve to distinguish heat stress. According to the rice yield formation mechanism, the yield estimation model coefficients and the yield model are fitted to realize the method for evaluating the yield reduction due to heat stress in the filling period of rice by monitoring and collecting time series images.

[0005] In order to achieve the above object, the present invention adopts the following technical solutions:

[0006] A method for evaluating heat stress reduction in rice during the grain filling period using time-series images comprises the following steps:

[0007] Construction of time series curves of growth appearance characteristics;

[0008] Identification of the effects of heat stress on rice;

[0009] Yield estimation model coefficient fitting;

[0010] Yield loss estimates enable rice production forecasts to be higher than normal.

[0011] Furthermore, the construction of the time series curve of the apparent characteristics of rice growth includes calculation of apparent color characteristic values ​​and construction of the time series curve; the calculation of the apparent color characteristic values ​​is to obtain the daily apparent color characteristic values ​​of the area of ​​interest by using the time series images reflecting the rice canopy, calculating and averaging the image bands; the construction of the time series curve includes filtering and smoothing the time series data of the apparent color characteristic values ​​and normalizing the time series data to obtain the green vegetation index time series data of consecutive days.

[0012] Furthermore, the apparent color characteristic values ​​are subjected to time series data filtering and smoothing, including: equal-interval generation of green vegetation index time series data and time series data filtering and smoothing, the equal-interval generation of green vegetation index time series data is to fill in missing values ​​through linear interpolation of the previous and next vegetation indices, and the time series data filtering and smoothing is to smooth the apparent green vegetation index time series data using the SG filtering algorithm.

[0013] Furthermore, the green vegetation index is the difference between the green light channel and the red light channel in the image, divided by the sum of the two, and is used to detect and quantify the trend of plant greenness decline.

[0014] Furthermore, the normalization process of the time series data is performed by dividing the difference between the green vegetation index of rice at a certain moment in the image and the minimum value of all green vegetation indices by the variation range of all green vegetation indices.

[0015] Furthermore, the identification of the impact of heat stress on rice includes identification of key points of growth dynamics and determination of the degree of high temperature stress.

[0016] Furthermore, the identification of key growth dynamic points includes: identifying the heading period D1, the filling period D2, the harvest period D3 and the filling period duration D3-2, using the color and shape characteristics of the rice ears in the image, extracting the heading period D1 through deep learning, and extracting the filling period D2 and the harvest period D3 by using the slope change of the apparent characteristic time series curve.

[0017] Furthermore, the high temperature stress degree judgment includes calculating the duration D3-2 from D3 to D2, and calculating whether the grouting time of the monitored object is shortened by comparing it with the normal grouting duration, and judging the high temperature stress degree according to the shortened duration.

[0018] Furthermore, the yield estimation model coefficient fitting includes establishing a canopy leaf senescence model based on the rice yield formation mechanism, extracting the green vegetation index during the filling period for model coefficient fitting, and calculating the cumulative value of photosynthetic radiation efficiency during the filling period as a yield reference value.

[0019] Furthermore, the yield model realizes the prediction of rice yield compared with normal years, that is, the cumulative biomass value of rice in the filling period is calculated by the established yield model and fitting coefficient, which is used as the yield prediction value of the current year; taking the normal year as a reference, the cumulative biomass value of rice after heat stress in the heading-filling period is calculated, and the value is compared with the yield prediction value of the normal year, which is the estimated value of yield loss.

[0020] The method disclosed in the application for evaluating the yield reduction of rice due to heat stress during the grain filling period using time-series images may bring about beneficial effects including but not limited to:

[0021] The final result is to quantitatively reflect the duration of the apparent color change of rice in the reproductive growth stage over the filling period, and to judge the impact of heat stress by comparing it with the normal filling period, and to evaluate the yield loss of the current rice after stress based on the yield model and normal years. The data processed by the present invention quantitatively describes the apparent color process of rice in the reproductive growth stage, and obtains more reasonable time series data of the apparent color characteristics of rice. After the grain-filling period suffers from heat stress, the harvest period and the duration of the grain-filling period can be estimated, the impact of heat stress in the grain-filling period can be qualitatively evaluated, and the yield loss can be quantitatively evaluated.

[0022] Compared with traditional manual observation and judgment, the present invention can effectively realize automatic identification of heading and flowering period, monitoring of growth in the grain filling period, and evaluation of heat stress impact and yield loss. The present invention is not limited to the impact of heat stress on rice yield reduction in the grain filling period, but is also applicable to the evaluation of rice yield reduction when abnormal field management occurs in the late grain filling period. The present invention has better adaptability in digitalized and informationized fields, and can achieve relatively stable prediction effects for rice growth image data taken regularly by automated image acquisition equipment under different environmental conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 This is a flow chart of a method for evaluating heat stress reduction in rice filling period using time series images according to an embodiment of the present invention;

[0024] Figure 2 This is a schematic diagram of the key point change recognition principle of rice apparent color characteristics according to an embodiment of the present invention;

[0025] Figure 3 This is a schematic diagram of time series data of rice growth appearance characteristics according to an embodiment of the present invention;

[0026] Figure 4 This is a schematic diagram of a time series curve of rice growth apparent characteristics after data smoothing and normalization according to an embodiment of the present invention;

[0027] Figure 5 A schematic diagram of a method for extracting key points of changes in rice growth appearance characteristics according to an embodiment of the present invention;

[0028] Figure 6 This is a schematic diagram of the high temperature stress recognition principle of rice during the grain filling period according to an embodiment of the present invention;

[0029] Figure 7 This is a graph showing the results of heat stress identification and yield prediction during the rice filling period according to an embodiment of the present invention. DETAILED DESCRIPTION

[0030] In order to make the purpose, technical solution and advantages of the present application more clear, the present application is further described in detail below in conjunction with the embodiments. It should be understood that the specific embodiments described here are only used to explain the present application and are not used to limit the present application.

[0031] On the contrary, the present application covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present application as defined by the claims. Further, in order to make the public have a better understanding of the present application, some specific details are described in detail in the detailed description of the present application below. Those skilled in the art can fully understand the present application without the description of these details.

[0032] The following is a detailed description of a method for evaluating heat stress reduction in rice filling period using time series images according to an embodiment of the present application. It should be noted that the following embodiments are only used to explain the present application and do not constitute a limitation on the present application.

[0033] like Figure 1 As shown, a method for evaluating heat stress reduction in rice filling period using time series images comprises the following steps:

[0034] Construction of time series curves of growth appearance characteristics;

[0035] Identification of the effects of heat stress on rice;

[0036] Yield estimation model coefficient fitting;

[0037] Yield loss estimates enable rice production forecasts to be above normal.

[0038] Step 1, constructing a growth apparent characteristic time series curve, including calculation of apparent color characteristic values ​​and construction of the time series curve.

[0039] The calculation of the apparent color characteristic value includes selecting a fixed area that can only reflect the rice canopy as the area of ​​interest, fixing and cropping an image with the best quality taken every day, referring to the remote sensing index calculation principle, and using the green vegetation index calculation formula (R and G are the green and blue channel values ​​of the image, respectively. b is the empirical coefficient) to obtain the apparent color feature value of the image, pixel i Corresponding to the time series of daily apparent color characteristics The color state value is , calculate the average of the apparent characteristic time series data of all pixels on the image of each day in the area of ​​interest, and obtain the apparent time series characteristic value of rice .

[0040] The time series curve construction includes smoothing the color appearance characteristic data of rice obtained every day by using the SG filtering algorithm, and normalizing it to the interval of 0-1 to obtain the normalized appearance color characteristic time series data. .

[0041] Step 2, identifying the effects of heat stress on rice, wherein the heat stress identification includes identification of key growth dynamic points and comparison of filling period time.

[0042] The key points of growth dynamics include: the time when rice ears begin to emerge D1, the time when grain filling begins D2, the time when rice matures D3 and the length of the grain filling period D 3-2 Through various rice ear image sample data sets, the color and shape features of the rice ears in the image are used to train the deep learning target detection model. The trained model is used to detect the images before and after the heading period in turn to extract the initial time D1 of the heading period.

[0043] Determine D2 and D3 according to the slope of the change curve of the apparent color characteristic time series data. The slope of the day is given by the formula Calculated and smoothed by SG filtering algorithm. The time corresponding to grouting and maturity is extracted according to the slope change after D1. When the slope returns from the minimum value to approximately 0, it corresponds to time D2. The second time the slope returns from a negative value to near 0, it corresponds to time D3.

[0044] Comparison of the duration of the grouting period, calculation of the duration D between the key points D3 and D2 3-2 ,By comparing with the length of the normal filling period, we can determine whether the shortened time exceeds the threshold.,When the filling duration is 4 days or more less than the normal time, it is judged that the filling period is affected by heat stress.

[0045] Step 3: fitting the yield estimation model coefficients, which includes model establishment and coefficient fitting. Considering the rice yield formation mechanism, based on the characteristics of the rice leaves gradually turning yellow and aging during the filling period, a yield estimation model is constructed. ,in is the photosynthetic radiation rate of leaves during the filling period, which is replaced by the daily green vegetation index value. a is the correction coefficient between the cumulative value of photosynthetic radiation rate and yield, f It is the yield coefficient relative to the normal growth of rice. The yield estimation model is established by extracting the time series data of green vegetation index during the rice filling period, fitting the polynomial model coefficients, and then building a yield estimation model.

[0046] Step 4, yield loss estimation and yield prediction, the yield loss estimation includes calculating the cumulative value of photosynthetic radiation rate in normal years and during the grain filling period under heat stress based on the fitted yield estimation model , , by setting f By substituting the value into the yield estimation model, we can get the degree of yield loss compared with normal years. ; Based on the production reference value of normal years , the rice yield prediction value can be predicted .

[0047] The following is an explanation using a specific embodiment as an example:

[0048] The rice growth images and real-time meteorological data are collected at fixed points at intervals of 2 hours from 10:00 to 16:00 during the day, and the heat stress impact identification and yield loss estimation are carried out according to the temperature at the fixed point being higher than 35 degrees; the optimal color synthesis method is used to synthesize a high-quality daily photo time series of one photo per day from all the photos taken every day, and the region of interest of only the rice canopy image in the image is cropped and saved in a fixed range frame as the data to be processed. In order to obtain the color appearance characteristics of rice, the monitoring image data includes at least red, green and blue three-channel values. In this embodiment, the green vegetation index is used to represent the color appearance characteristics of rice.

[0049] Step 1: Construction of growth epitaxial characteristic time series curve, based on Figure 2 The key change recognition principle of rice apparent color features is based on the green and red channel values ​​of the image. b Set it to 1.2, calculate the green vegetation index of rice every day and find the average value to get the apparent color characteristic value of rice every day (see Figure 3 ). Then the color appearance feature time series data is smoothed by SG filtering algorithm, and the time series data is normalized to the range of [0,1] to obtain the rice appearance color feature time series curve. Figure 4 .

[0050] Step 2: Identify the effects of heat stress on rice. Use the deep learning target detection model trained with a sample dataset of labeled rice ears to sequentially identify rice images acquired every day after rice transplanting, and take the time when the ear is first detected as the heading stage D1.

[0051] The slope of the change curve Slope is calculated based on the normalized rice apparent color feature time series data and SG smoothing is performed. The change points D2 and D3 are determined based on the change trend of the slope. The specific time points are extracted as follows: the time when the slope Slope returns from the minimum value to approximately 0 after D1 is D2, and the time when the slope is briefly greater than 0 and then becomes negative and returns to near 0 is D3. The schematic diagram of determining the key points of the dynamic growth of rice apparent color can be found in Figure 5 .

[0052] Calculate the duration D from rice filling to rice maturity based on D3 and D2 3-2 , see Figure 2 By comparing with the normal filling period, when the filling period lasts for less than 4 days or more than the normal period, it can be judged that the rice has been affected by high temperature stress during the filling period. The judgment principle can be found in Figure 6 .

[0053] Step 3: Fitting the yield model coefficients. Considering the rice yield formation mechanism, the yield model is constructed based on the characteristics of the gradual yellowing and aging of rice leaves during the filling period. The time series data of rice green vegetation index in the filling period were extracted, and the polynomial model coefficients were fitted. The order of the polynomial was adjusted to make the determination coefficient and mean square error after fitting optimal, and thus the yield model was established.

[0054] Step 4: Yield loss estimation and yield prediction. According to the fitted yield model, the yield loss in normal years and under heat stress are calculated. The cumulative values ​​are and , according to the set f The value was used to calculate the rice yield loss under heat stress during the filling period; according to the rice yield value in normal years, the a The yield loss degree of rice under heat stress during the grain filling period was calculated, and the yield prediction results under heat stress compared with normal years were obtained. Figure 7 .

[0055] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for evaluating heat stress reduction in rice during the filling period using time series images, characterized in that: The following steps are involved: Construction of time series curves of growth appearance characteristics; Identification of the effects of heat stress on rice; Yield estimation model coefficient fitting; Yield loss estimation realizes rice yield forecasts compared to normal years; The construction of the growth apparent characteristic time series curve includes the calculation of apparent color characteristic values ​​and the construction of the time series curve; the calculation of apparent color characteristic values ​​is to obtain the daily apparent color characteristic values ​​of the area of ​​interest by using the time series images reflecting the rice canopy through image band calculation and averaging; The time series curve construction includes filtering and smoothing the apparent color feature values ​​and normalizing the time series data to obtain the green vegetation index time series data for consecutive days; The apparent color feature value is subjected to time series data filtering and smoothing, including: equal spacing generation of green vegetation index time series data and filtering and smoothing of time series data, wherein the equal spacing generation of green vegetation index time series data is to fill missing values ​​through linear interpolation of the previous and next vegetation indexes, and the filtering and smoothing of time series data is to smooth the apparent green vegetation index time series data using the SG filtering algorithm; The identification of the impact of heat stress on rice includes identification of key points of growth dynamics and judgment of the degree of high temperature stress; the identification of key points of growth dynamics includes: identifying the heading period D1, the filling period D2, the harvest period D3 and the filling period duration D3-2, using the color and shape characteristics of the rice ears in the image, extracting the heading period D1 through deep learning, and extracting the filling period D2 and the harvest period D3 by using the slope change of the apparent characteristic time series curve; the judgment of the degree of high temperature stress includes calculating the duration D3-2 from D3 to D2, calculating whether the filling time of the monitored object is shortened by comparing it with the normal filling duration, and judging the degree of high temperature stress according to the shortened duration.

2. The method for evaluating heat stress reduction in rice filling period by using time series images according to claim 1, characterized in that: The green vegetation index is the difference between the green light channel and the red light channel in the image, divided by the sum of the two, and is used to detect and quantify the trend of plant greenness decline.

3. The method for evaluating heat stress reduction in rice filling period by using time series images according to claim 1, characterized in that: The normalization process of the time series data is performed by dividing the difference between the green vegetation index of rice at a certain moment in the image and the minimum value of all green vegetation indices by the variation range of all green vegetation indices.

4. The method for evaluating heat stress reduction in rice filling period by using time series images according to claim 1, characterized in that: The yield estimation model coefficient fitting includes establishing a canopy leaf senescence model based on the rice yield formation mechanism, extracting the green vegetation index during the filling period for model coefficient fitting, and calculating the cumulative value of photosynthetic radiation efficiency during the filling period as a yield reference value.

5. The method for evaluating heat stress reduction in rice filling period by using time series images according to claim 1, characterized in that: The yield loss estimation realizes the prediction of rice yield compared with normal years, that is, the cumulative biomass value of rice in the filling period is calculated by the established yield model and fitting coefficient, which is used as the yield prediction value of the current year; taking the normal year as a reference, the cumulative biomass value of rice after heat stress in the heading-filling period is calculated, and the value is compared with the yield prediction value of the normal year, which is the estimated value of yield loss.

Images