CN113607734B - Visualization method for nondestructively estimating chlorophyll content and chlorophyll distribution of plant - Google Patents

Abstract

The invention discloses a visual method for nondestructively estimating chlorophyll content and distribution of plants, which comprises the steps of acquiring plant images by using a visible light camera, identifying all branches of the plants by using a target detection algorithm, selecting a target part by using a rectangular frame, calculating a rectangular frame with the largest height as a main branch area of the plants, and dividing the main branch area; extracting layered color factors in different color spaces, and carrying out inversion modeling on the combination of the multi-color factors and the SPAD value measured by the chlorophyll content meter to obtain a color factor combination model with highest fitting degree; the model is applied to plant canopy leaves to represent the distribution of SPAD values, visual display of chlorophyll content on the whole plant plane is realized, photosynthesis of plants is judged through the chlorophyll content, and change of the chlorophyll content of the plants which are difficult to identify by naked eyes can be observed. The method does not need to destroy plants and has objective and accurate estimation results.

Description

Visualization method for non-destructive estimation of plant chlorophyll content and distribution

Technical Field

The present invention relates to the field of image analysis and processing, and in particular to a visualization method for non-destructively estimating plant chlorophyll content and distribution.

Background Art

Chlorophyll is a pigment in higher plants and an important basis for plant physiology research. The content of chlorophyll indicates the growth and health of plants, plays an important role in plant photosynthesis, and provides energy for plant growth and development. Nitrogen is a component element of chlorophyll. During plant growth, the chlorophyll content is closely related to the nitrogen content of the plant. Too high or too low nitrogen content has a certain impact on the growth process of the plant. Chlorophyll can indirectly reflect the nitrogen content level of the plant. By using the rapid visualization of chlorophyll content, early diagnosis of nitrogen deficiency and nitrogen excess can be carried out before the visible stress symptoms of the plant appear, and the growth monitoring and growth judgment of the plant can be realized, thereby providing technical guidance for determining and adjusting cultivation and management measures. Therefore, it is of great significance to quickly estimate and intuitively display the chlorophyll content.

During plant growth, when stress occurs, the physiological and biochemical phenotypic parameters of the plant have changed, but due to the duration of stress, the degree of stress and plant resistance, there is no obvious difference or change in the appearance of the plant, so it is difficult to observe the changes in the plant due to stress with the naked eye. Chlorophyll is a probe of plant photosynthesis. Determining the chlorophyll content of plants plays an important guiding role in understanding whether the plant is growing normally and judging whether it is under the influence of stress. With the demand for large-scale, high-precision and rapid monitoring of plant growth conditions, many methods for determining the chlorophyll content of plants have emerged.

The traditional methods for determining chlorophyll mainly include spectrophotometry and chlorophyll meter method.

When using the spectrophotometric method, take fresh (or dried) plant leaves, cut them into pieces (remove the midrib), mix them and put them in a mortar, add 80% acetone, calcium carbonate and quartz sand, grind them into a homogenate, add 80% acetone, transfer the homogenate into a centrifuge tube for centrifugation, take the extract, use a visible light spectrophotometer for colorimetry, measure the absorbance values at wavelengths of 663nm, 645nm, and 652nm, and use 80% acetone as a reference. Calculate chlorophyll a and chlorophyll b according to the formula, and calculate the total chlorophyll content (a+b). However, when using the spectrophotometric method to determine the chlorophyll content, there are problems such as destroying plant growth, multiple detection processes, and long cycles.

The chlorophyll meter determines the relative amount of chlorophyll in the leaves by measuring the optical concentration difference of the leaves at two wavelengths: 650nm and 940nm. When using the chlorophyll meter method, use a chlorophyll content meter such as SPAD-502 to clamp a point on the plant leaf, measure the SPAD value of the point, and use this value to represent the current chlorophyll content in the leaf. When using the chlorophyll meter to measure the chlorophyll content, the chlorophyll meter head must ensure that the sampled leaf completely covers the receiving window, and a certain sampling area (2mm*3mm) is clamped, and the leaf is not thick (maximum thickness 1.2mm) before measurement can be performed. Therefore, the chlorophyll meter method has poor universality in application, and it is difficult to measure the chlorophyll content of some small leaves. It is also more likely to have problems such as large errors caused by improper manual operation.

The chlorophyll fluorescence meter is an analytical instrument used in the field of biology. It uses the fluorescence phenomenon of chlorophyll, adopts a sensor with extremely high sensitivity and reaction speed, and is supplemented by a photoelectric pulse design to capture the fluorescence data at each time period during the electron transfer process, and then the chlorophyll content can be determined. However, the chlorophyll fluorescence meter is expensive and not widely used. In the process of using the chlorophyll fluorescence meter to determine the chlorophyll content, it is necessary to pick leaves from the growing plants, and the leaves need to be dark-treated for a long time before the measurement can be performed. Therefore, the chlorophyll fluorescence meter still has problems such as destroying plant growth, cumbersome detection process, and long detection cycle.

In summary, the traditional method of measuring chlorophyll content can only be used to separate leaves from plants by breaking, removing, cutting, etc. for destructive measurement, and it is impossible to measure the same plant continuously. Plant growth is a complex, continuous and dynamic process, which is regulated by genes and the environment. This requires the analysis of chlorophyll content at each growth stage of the whole plant to analyze the temporal variation pattern of genetic control in the formation of phenotypic traits.

In recent years, digital imaging technology has provided new directions and means for detecting the chlorophyll content in plants. Currently, there are also methods that use visible light cameras to collect plant leaf images and then analyze the chlorophyll content through image analysis. However, most of the current work of analyzing chlorophyll content by collecting plant leaf images requires removing leaves from growing plants, placing a single leaf under the visible light camera lens to collect the image, and then analyzing the chlorophyll content by analyzing and processing the image. This method will inevitably cause irreversible damage to the growth or leaves of the plant. It is impossible to estimate whether the chlorophyll content will change due to factors such as in vitro, time, and environment after the leaves are removed, and it is also impossible to achieve a visual distribution of the chlorophyll content canopy of the entire plant.

Therefore, the demand for non-destructive, objective, accurate, high-resolution, automatic and efficient in vivo chlorophyll measurement technology is becoming increasingly obvious. The whole plant images collected by visible light cameras can intuitively reflect the changes in chlorophyll content and thus monitor the changes in chlorophyll content. It is of great significance to study the estimation and analysis of chlorophyll content through the color parameters of the whole plant images.

Summary of the invention

The technical problem to be solved by the present invention is to provide a non-destructive, objective, accurate, high-resolution, automatic and efficient non-destructive visualization method for estimating the chlorophyll content and distribution of plants in order to address the problems of the traditional measurement of the chlorophyll content of plants, such as cumbersome process, small processing scale, only targeting a single leaf, strong destructiveness and non-intuitive results. The visualization method for non-destructive estimation of the chlorophyll content and distribution of plants does not require destruction of the plants. By analyzing and processing the plant images collected by the visible light camera, the chlorophyll content is visualized, which makes it easy to observe the changes in the chlorophyll content of the plants during their growth.

In order to achieve the above technical objectives, the technical solution adopted by the present invention is:

A visualization method for non-destructively estimating plant chlorophyll content and distribution comprises the following steps:

(1) Use a visible light camera to photograph the plant and collect a complete plant image;

(2) Extracting the image of the pure plant part in the main branch area from the complete plant image;

(3) Layering the image extracted in step (2);

(4) Use a chlorophyll meter to measure the SPAD (Soil and plantanalyzer development) values of all leaves in each layer, and calculate the SPAD average value of all leaves in each layer;

(5) The optimal regression model of chlorophyll content was established by using the color analysis method and combining the SPAD average value of all leaves in each layer;

(6) Use the optimal regression model of chlorophyll content to estimate the chlorophyll content and visualize the chlorophyll content.

As a further improved technical solution of the present invention, the step (2) specifically includes:

(2.1) Use the target detection algorithm to identify the complete plant image and then identify all the branches of the plant. Use a rectangular frame to select the target part, calculate the height of each rectangular frame, and take the rectangular frame with the largest height as the main branch area of the plant to segment the main branch area;

(2.2) The target detection algorithm is used to extract the region of interest from the pure plant part of the main branch area using the threshold of the G channel, and the maximum connected domain method is used to generate a mask of the pure plant part of the main branch area to remove the influence of the environmental background on the extracted color factor.

As a further improved technical solution of the present invention, the step (3) specifically includes:

The image extracted in step (2) is divided into an upper layer, a middle layer and a lower layer, and the division ratio of the upper layer, the middle layer and the lower layer is determined according to the height of the main branch area of the plant.

As a further improved technical solution of the present invention, the step (4) specifically includes:

The SPAD values of all leaves in the upper layer, middle layer and lower layer were measured by a chlorophyll meter, and the average SPAD value of all leaves in the upper layer, the average SPAD value of all leaves in the middle layer and the average SPAD value of all leaves in the lower layer were calculated.

As a further improved technical solution of the present invention, the step (5) specifically includes:

H、S、V、L、a、b；(5.1) Convert the image extracted in step (2) into color space RGB (Red, Green, Blue), HSV (Hue, Saturation, Value) and La*b* (Lightness, a-star, b-star), respectively, and calculate the parameter value of the color factor of each pixel in the upper, middle and lower images, respectively. The color factors include R, G, B, G*G,

H, S, V, L, a, b;

(5.2) Calculate the average parameter value of each color factor of all pixels in the upper layer image; calculate the average parameter value of each color factor of all pixels in the middle layer image; calculate the average parameter value of each color factor of all pixels in the lower layer image;

(5.3) Randomly combine multiple color factors to establish multiple color factor combination models. Use the parameter average of the color factors in each layer of the image and the SPAD average of all leaves in each layer as training data sets to train multiple color factor combination models respectively, and obtain multiple sets of trained color factor combination models, i.e., multiple sets of chlorophyll content regression models.

(5.4) Using the root mean square error RMSE and the determination coefficient ^R2 as indicators, the fitting performance of multiple groups of trained chlorophyll content regression models was evaluated and the regression model of chlorophyll content with the best fitting performance, that is, the optimal regression model of chlorophyll content, was determined.

As a further improved technical solution of the present invention, the optimal regression model of chlorophyll content in step (5) is:

Y＝-8.51*lg(G)+11.68*R-26.48*G+18.30*B+2.81*G/R+3.85*G/B+40;

Where Y is the chlorophyll content estimated by the best regression model of chlorophyll content.

As a further improved technical solution of the present invention, the step (6) comprises:

(6.1) collecting and processing the image of the plant to be tested according to the method of step (1) and step (2);

与

对所有像素点做标准化，计算标准化后的多个色彩因子参数值，代入叶绿素含量的最佳回归模型中，得到一张代表SPAD拟合值的灰度图，将SPAD的拟合值在像素点区间放大并转换为COLORMAP_JET色度的伪彩色图像，进而实现叶绿素含量的可视化。在整个伪彩色图像中，包含蓝-绿-黄-红等不同颜色且由深到浅的渐变范围。其中红色代表叶绿素含量高的区域、浅绿与浅黄色代表叶绿素含量中等的区域、蓝色代表叶绿素含量低的区域。最终做到把叶绿素含量的数据转换成直观的图形或图像在屏幕上显示出来，实现叶绿素含量在整个植物平面上的可视化。(6.2) Split the processed image into three channels: red, green, and blue, obtain the R, G, and B values of each pixel, and calculate lg(G),

and

All pixels are standardized, and the standardized color factor parameter values are calculated and substituted into the optimal regression model of chlorophyll content to obtain a grayscale image representing the SPAD fitting value. The SPAD fitting value is enlarged in the pixel interval and converted into a pseudo-color image of COLORMAP_JET chromaticity, thereby realizing the visualization of chlorophyll content. The entire pseudo-color image contains different colors such as blue-green-yellow-red and a gradient range from dark to light. Red represents areas with high chlorophyll content, light green and light yellow represent areas with medium chlorophyll content, and blue represents areas with low chlorophyll content. Finally, the chlorophyll content data is converted into intuitive graphics or images and displayed on the screen, realizing the visualization of chlorophyll content on the entire plant plane.

The present invention uses a visible light camera to collect plant images, uses an image processing algorithm to study the distribution of chlorophyll content in the whole plant, establishes an optimal estimation model for chlorophyll content and visualizes it, obtains an image that intuitively displays the distribution of chlorophyll content on the plant, and judges the photosynthesis of the plant through the chlorophyll content, and can observe changes in the chlorophyll content of the plant that are difficult to identify with the naked eye. In addition, during the growth of plants, the chlorophyll content is closely related to the nitrogen content of the plant. If the nitrogen content is too high or too low, it will have a certain impact on the growth process of the plant. Chlorophyll can indirectly display the nitrogen content level of the plant. By using the rapid visualization of the chlorophyll content, early nutrition diagnosis of nitrogen deficiency and nitrogen excess can be performed before the stress symptoms visible to the naked eye appear in the plant, so as to achieve plant growth monitoring and growth judgment, and then provide technical guidance for determining and adjusting cultivation management measures.

The beneficial effects of the present invention are:

Visible light cameras are low-cost and widely used. The image acquisition process is simple and does not require destructive sampling of plants. Computer graphics and image processing technology are used to convert the complex and redundant chlorophyll content data into intuitive graphics or images displayed on the screen, which is more intuitive and realizes the visualization of the distribution of chlorophyll content in the whole plant, providing a non-destructive visual analysis method for estimating the chlorophyll content and distribution of plants during their growth.

BRIEF DESCRIPTION OF THE DRAWINGS

FIG1 is a flow chart of the method of the present invention.

Figure 2 is the original plant image captured by the visible light camera, from left to right are the control group, low fertilizer group, high fertilizer group, and over-fertilization group.

FIG3 is a grayscale image of FIG2 .

Figure 4 is a schematic diagram of the recognition effect of branch and trunk morphological structure phenotypic parameters.

FIG5 is a grayscale image of FIG4.

FIG6 (a) is a schematic diagram of a plant height image layering ratio.

FIG6( b ) is another schematic diagram of the layered ratio of plant images according to plant height.

FIG. 7 (a) is a grayscale image of FIG. 6 (a).

FIG7( b ) is a grayscale image of FIG6( b ).

Figure 8 is a comparison of the original plant image captured by the visible light camera and the visualization of the distribution of chlorophyll content in the entire plant plane. From left to right are the control group, low fertilizer group, high fertilizer group, and over-fertilization group.

FIG9 is a grayscale image of FIG8 .

FIG. 10 is a schematic diagram of an implementation of a visualization program of the present invention.

FIG11 is a grayscale image of FIG10 .

DETAILED DESCRIPTION

When considered in conjunction with the accompanying drawings and schedules, the present invention can be more completely and better understood and many of its attendant shortcomings can be easily known by referring to the detailed description below, but the drawings and schedules described herein enhance further understanding of the present invention and constitute a part of the present invention.

In order to make the above-mentioned objects, features and advantages of the present invention more clearly understandable, the present invention is further described in detail below in conjunction with the accompanying drawings, attached tables and specific implementation methods.

Figures 1 to 11 are an application example of using the present invention to study the distribution of chlorophyll content in Willow of the Corydalis. Using the present invention, a non-destructive visual analysis method for estimating the chlorophyll content and distribution of plants can be used to obtain the best estimation model of the chlorophyll content in Willow of the Corydalis used in the experiment, and obtain the visualization result of the chlorophyll content in the Willow of the Corydalis plant.

In this application example, considering that the growth of leaves and branches of Willow of Siberia oxyphylla is random and the deep learning algorithm requires a large amount of data, the present invention avoids images with similar shooting angles and selects 2,000 images containing the growth conditions of various branches of Willow of Siberia oxyphylla as the data set to increase the diversity of the data set. 90% of the random samples are used for training and 10% for testing.

Step 1: Image acquisition stage:

As shown in Figures 2 and 3, the plants were photographed using a visible light camera to capture complete plant images.

Step 2, chlorophyll visualization stage:

The target detection algorithm was used to extract the branch part of the willow as the region of interest to extract the morphological and structural phenotypic parameters of the branches and trunks. The target part was selected with a rectangular frame, and the rectangular frame with the largest height was calculated as the main branch area of the plant, which was segmented, as shown in Figures 4 and 5. The ROI (Region Of Interest) was extracted using the threshold (25, 255) of the G (Green) channel and the maximum connected domain was obtained to generate a mask of the pure plant part of the main branch area.

In the process of applying the present invention to the actual test of Willow of Ilex fasciatus, the test samples of Willow of Ilex fasciatus are divided into three layers: upper, middle and lower (the stratification is to divide the image of the pure plant part of the segmented main branch area): the correlation between the SPAD value and the color factor is compared according to different stratification ratios of multiple previous tests, and the best stratification ratio in the growth stage of this plant is obtained: when the plant height in the seedling stage (the plant height in the seedling stage is the height of the main branch area) does not exceed 35 cm, the ratio of the upper, middle and lower layers is 3:4:3, as shown in (a) of Figure 6 and (a) of Figure 7; when the plant height in the seedling stage exceeds 35 cm, the ratio of the upper, middle and lower layers is 2:6:2, as shown in (b) of Figure 6 and (b) of Figure 7.

The SPAD-502 chlorophyll content meter was used to measure the SPAD values of all leaves in the upper, middle and lower layers. Each leaf was measured three times at different positions, and the average value was taken as the result. Finally, the SPAD values of all leaves in each layer were averaged as the SPAD value of the leaves in that layer.

By converting the image (the image refers to the image of the pure plant part of the main branch area) to different color space analysis, the characteristics of different color spaces can be effectively used to capture the feature area of interest, and more accurately obtain the characterization factors that are more relevant to the target value. In the process of applying the present invention to the example test of the willow of the dustpan, three color spaces RGB, HSV, La*b* and four common conversion factors based on RGB space are used for correlation analysis with SPAD values. Similarly, the combination of different color features that can be converted by the conversion algorithm is sometimes related to the chlorophyll of the leaves, so the three color factors R, G, and B are combined and calculated, and the color components that can reflect the degree of green are considered and converted into another 4 common color factors. The names of the color factors extracted from different color spaces are shown in Table 1. Then, the average value of each color factor extracted from each layer is taken as a data set for regression analysis with the SPAD value of each layer of leaves measured before. Some multi-color factors and SPAD values are shown in Table 2.

Table 1:

Table 2:

Considering that the sizes of different color factors may vary greatly, larger values are likely to dominate the target results, making it impossible for the regression algorithm to learn other features. Therefore, we first use the data standardization algorithm to make the numerical data dimensionless, and then transform the original data through formula (1-1) to transform the data into a range with a mean of 0 and a standard deviation of 1.

Wherein, X is the color factor parameter value directly extracted, dimensionless unit; X _mean is the color factor parameter mean value, dimensionless unit; X _std is the color factor parameter standard deviation value, dimensionless unit.

In the process of applying the present invention to the example test of Willow, a total of 580 groups of multi-color factors and SPAD values were obtained as data sets, of which 90% were used for training and 10% were used for testing. The best model was obtained by training when the model variance regularization strength alpha was set to 3.2e-6 and the maximum number of iterations was 10,000.

The root mean square error RMSE and determination coefficient ^R2 were used as evaluation indicators to evaluate the fit between different color factor combination models and the SPAD index to test the fitting performance of the regression model. The results are shown in Table 3. RMSE represents the average prediction error compared with the measured value. The lower the value, the higher the accuracy. ^R2 represents the model fitting effect, which is the percentage of the measurement variance explained by the algorithm model. The value range is [0,1]. The larger the ^R2 , the better the model fitting effect. The statistical calculation formulas are as follows (1-2) and (1-3):

Wherein, y _real is the SPAD value of the leaf clamped by the handheld chlorophyll meter, dimensionless unit; y _pred is the SPAD value predicted by the multi-color factor correlation model, dimensionless unit; m is the number of data, unit is group; y _mean is the average value of all SPAD predicted values, dimensionless unit.

Table 3:

Comparing the fitting degree of the linear regression model of color factors under different color spaces, as shown in Table 3, the first three groups directly use the color factors extracted from the three channels of the color space to fit the SPAD value to obtain the model fitting degree is not high, and ^R2 is mostly 0.58. In the process of using the present invention to apply the example test of the willow of the dustpan, the fitting performance of the H, S, and V color factors is the best, with an error RMSE of only 2.62 and ^R2 reaching 0.60, which is significantly better than other color spaces, but weaker than the combination transformation of the latter group of RGB color spaces. Model 4 adds two color factors, G/R and G/B, to the model based on the RGB space color factor, which represents the ratio of the green component to the red component and the blue component, respectively, and can better reflect the green degree of plant leaves. Compared with model 5 that adds a single color factor G/(R+B), it achieves the improvement of the determination coefficient and the reduction of the error on the basis of the original RGB color space model 1.

Table 4:

Table 4 shows the nonlinear polynomial regression models constructed with different color factors and the estimation errors. To reduce the complexity of the model, the variance selection method was used to filter low-variance variables for multiple independent variables in the model, and only the variables with the highest correlation were retained as quadratic terms and logarithmic terms. It can be seen that the regression model 14 constructed by the color factors lg(G), R, G, B, G/R, G/B and the regression model 15 constructed by the color factors lg(G), R, G, B, G/(R+B) have a determination coefficient R ² of up to 0.73, but the model 14 with the addition of G/R and G/B color factors has a smaller error (RMSE=2.16<2.21). Therefore, the regression model 14: Y=-8.51*lg(G)+11.68*R-26.48*G+18.30*B+2.81*G/R+3.85*G/B+40 has the best fit with the SPAD value, showing the most significant regression, and is the best chlorophyll content fitting model. In the linear regression model, it has been proven that G/R and G/B can better reflect the greenness of plant leaves than the G/(R+B) color factor. The results of the quadratic regression model and the logarithmic regression model also confirm that G/R and G/B can be adjusted separately as two parameters, thereby better fitting the relative greenness of plant leaves.

By visualizing the chlorophyll content in the images of the Willow of Ilex fasciatus samples, the changes in the chlorophyll content of the Willow of Ilex fasciatus caused by the application of nitrogen can be made more vivid and intuitive, and the effect of nitrogen on the chlorophyll content of the plant can be quickly reflected. In the process of applying the present invention to the example test of the willow of the dustpan, the image is split into three channels of red, green and blue, and the R, G and B values of each pixel are obtained. The lg(G), G/B and G/R of each non-zero pixel are calculated, and all the pixels are standardized to obtain the standardized color factor lg(G), R, G, B, G/R, and G/B parameter values, which are substituted into the regression model 14 with the highest correlation fitted in the previous section: Y = -8.51*lg(G)+11.68*R-26.48*G+18.30*B+2.81*G/R+3.85*G/B+40, and a grayscale image representing the SPAD fitting value is obtained. The SPAD fitting value is in the interval [30,50]. The SPAD interval is enlarged to [0,255] and converted into a pseudo-color image of COLORMAP_JET chromaticity. In the entire pseudo-color image, different colors such as blue-green-yellow-red and a gradient range from dark to light are included. Red represents areas with high chlorophyll content, light green and light yellow represent areas with medium chlorophyll content, and blue represents areas with low chlorophyll content, as shown in Figure 5. By visualizing the entire plant, the data related to chlorophyll content is presented clearly and intuitively, no longer limited to numbers.

Step 3, analysis phase:

It can be seen from the visualization effect diagrams of the chlorophyll content distributed throughout the plant plane in Figures 8 and 9 that for a single plant of Willow of the Asphodeloides, the SPAD value of the upper layer is usually the lowest, often not exceeding 40, because the upper leaves are the youngest, the internal structure is imperfect, the respiration is vigorous and the chlorophyll content is low; the SPAD value gradually increases in the middle and lower layers, because as the leaves grow, the chlorophyll content continues to increase, while the lower leaves are older, the internal structure of the structure begins to be damaged, and the photosynthetic rate is also low. It can be seen from the application of the present invention to the Willow of the Asphodeloides example test process that the SPAD value data table does not fluctuate significantly and the results are not intuitive. The SPAD value automatically calculated from the plant image is consistent with the SPAD value measured by the handheld chlorophyll meter, but the expression is more vivid and intuitive.

From the visualization effect diagram of the chlorophyll content distribution in the entire plant plane in Figures 8 and 9, it can be seen that the chlorophyll content of the willow of Example a is relatively low overall, and the chlorophyll content of the upper and lower layers is significantly different. The chlorophyll content of the lower leaves is high but the range is small. This also indirectly reflects that the nitrogen content applied or absorbed by the willow of Example a is relatively low, so nitrogen fertilizer can be appropriately increased in the subsequent growth process. The chlorophyll content of the willow of Example d is relatively high overall, and there is no obvious difference in chlorophyll content in the upper, middle and lower layers. By observing the plant image of the willow of Example d collected by the visible light camera, it can be seen that most of the leaves are drooping and wilting, which indirectly reflects that the nitrogen content applied or absorbed by the willow of Example d is relatively high and even has symptoms of seedling burn, so nitrogen fertilizer should be appropriately reduced in the subsequent growth process.

By comparing the original plant image captured by the visible light camera in Figure 8 or Figure 9 and the visualization effect diagram of the chlorophyll content distribution in the entire plant plane, it can be seen that the distribution of chlorophyll content and the changes produced during the growth of Willow of the Crassula ovata become more vivid and intuitive. It can also be seen that the chlorophyll visualization method can quickly and sensitively detect changes qualitatively and quantitatively before "the changes are seen by the naked eye", and conduct quantitative analysis, which plays an important role in monitoring plant growth and evaluating plant growth. Usually, the stress symptoms of nitrogen deficiency and nitrogen excess are very delayed. In the original image obtained under naked eye observation in Figure 8, except for the over-fertilized group, there is no significant difference in the application of different nitrogen fertilizers to plants. However, in the visualization effect diagram of chlorophyll content distribution, it can be found that there is a clear difference. The red area (i.e., the area with high chlorophyll content) in the fertilized group is larger than that in the control group, and the red area expands with the increase of nitrogen fertilizer. Through the rapid visualization of chlorophyll content, early diagnosis of nitrogen deficiency and nitrogen excess nutrition can be carried out before the visible stress symptoms of the plant appear, and the growth monitoring and growth judgment of the plant can be realized, thereby providing technical guidance for determining and adjusting cultivation management measures.

The above is only an embodiment of the present invention and is not intended to limit the scope of implementation of the present invention. In other words, any equivalent changes and modifications made in accordance with the scope of the claims of the present invention, as long as they do not substantially deviate from the inventive point and effect of the present invention, can be varied in many ways, which is obvious to those skilled in the art. Therefore, such variations are also all included in the protection scope of the present invention.

The above is a detailed introduction to a visual analysis method for non-destructively estimating plant chlorophyll content and distribution provided by the present invention, and the above is a description of an exemplary embodiment of the present application with reference to the accompanying drawings. It should be understood by those skilled in the art that the above embodiments are merely examples given for illustrative purposes and are not intended to be limiting. Any modifications, equivalent substitutions, etc. made under the teachings of this application and the scope of protection of the claims shall be included in the scope of protection claimed in this application.

Claims (1)

1. A method for non-destructively estimating plant chlorophyll content and distribution, characterized in that it comprises the following steps: (1) Use a visible light camera to photograph the plant and collect a complete plant image; (2) Extracting the image of the pure plant part in the main branch area from the complete plant image; (3) Layering the image extracted in step (2); (4) Use a chlorophyll meter to measure the SPAD values of all leaves in each layer, and calculate the average SPAD value of all leaves in each layer; (5) The optimal regression model of chlorophyll content was established by using the color analysis method and combining the SPAD average value of all leaves in each layer; (6) Using the optimal regression model of chlorophyll content, the chlorophyll content is estimated and the chlorophyll content is visualized; The step (2) specifically includes: (2.1) Use the target detection algorithm to identify the complete plant image and then identify all the branches of the plant. Use a rectangular frame to select the target part, calculate the height of each rectangular frame, and take the rectangular frame with the largest height as the main branch area of the plant to segment the main branch area; (2.2) extracting the region of interest from the pure plant part of the main branch area by using the target detection algorithm and the threshold of the G channel, and generating a mask of the pure plant part of the main branch area by using the maximum connected domain method; The step (3) specifically includes: Dividing the image extracted in step (2) into an upper layer, a middle layer and a lower layer, and determining the division ratio of the upper layer, the middle layer and the lower layer according to the height of the main branch area of the plant; The step (4) specifically includes: The SPAD values of all leaves in the upper layer, middle layer and lower layer were measured by using a chlorophyll meter, and the average SPAD value of all leaves in the upper layer, the average SPAD value of all leaves in the middle layer and the average SPAD value of all leaves in the lower layer were calculated. The step (5) specifically includes: (5.1) Convert the image extracted in step (2) to the color space RGB, HSV and La*b* respectively, and calculate the parameter value of the color factor of each pixel in the upper, middle and lower images respectively, wherein the color factor includes R, G, B, G*G,

H, S, V, L, a, b; (5.2) Calculate the average parameter value of each color factor of all pixels in the upper layer image; calculate the average parameter value of each color factor of all pixels in the middle layer image; calculate the average parameter value of each color factor of all pixels in the lower layer image; (5.3) Randomly combine multiple color factors to establish multiple color factor combination models. Use the parameter average of the color factors in each layer of the image and the SPAD average of all leaves in each layer as training data sets to train multiple color factor combination models respectively, and obtain multiple sets of trained color factor combination models, i.e., multiple sets of chlorophyll content regression models. (5.4) Using the root mean square error (RMSE) and the coefficient of determination (R2 ⁾ as indicators, the fitting performance of multiple sets of trained regression models of chlorophyll content was evaluated and the regression model of chlorophyll content with the best fitting performance was determined, i.e., the optimal regression model of chlorophyll content; The optimal regression model for the chlorophyll content in step (5) is: Y＝-8.51*lg(G)+11.68*R-26.48*G+18.30*B+2.81*G/R+3.85*G/B+40; Where Y is the chlorophyll content estimated by the best regression model of chlorophyll content; The step (6) comprises: (6.1) collecting and processing the image of the plant to be tested according to the method of step (1) and step (2); (6.2) Split the processed image into three channels: red, green, and blue, obtain the R, G, and B values of each pixel, and calculate lg(G),

and

All pixels are standardized, and multiple color factor parameter values after standardization are calculated and substituted into the optimal regression model of chlorophyll content to obtain a grayscale image representing the SPAD fitting value. The SPAD fitting value is enlarged in the pixel interval and converted into a pseudo-color image of COLORMAP_JET chromaticity, thereby realizing the visualization of chlorophyll content.

Images