Method for detecting chlorophyll content of plant leaves

Abstract

The invention discloses a plant leaf chlorophyll content detection method. The method comprises the following steps: collecting hyperspectral reflectance data of a plant leaf sample to be detected, wherein the hyperspectral reflectance data comprises n wavelength pairs and spectral reflectance corresponding to each wavelength in the wavelength pairs; calculating a spectral index corresponding to the plant leaf sample to be detected according to the hyperspectral reflectivity data; inputting the spectral index into a chlorophyll content prediction model, and outputting the chlorophyll content of the plant leaf sample to be detected by the chlorophyll content prediction model. The device can perform nondestructive detection on the plant leaves, improves the detection precision, can be suitable for different plant leaves in different growth environments, and is high in applicability.

Description

Technical Field

[0001] The present invention relates to the technical field of plant leaf detection, and in particular to a method for detecting chlorophyll content in plant leaves. Background Art

[0002] In the field of modern agriculture and ecological monitoring, accurate detection of plant chlorophyll content is of great significance for crop management, plant physiology research and environmental monitoring. Chlorophyll is a key element in plant photosynthesis, and its content directly reflects the health status, nutrient absorption and environmental pressure of the plant.

[0003] The traditional method for determining the chlorophyll content in plant leaves is chemical method, such as spectrophotometry and high performance liquid chromatography. These methods require grinding the samples and extracting them with chemical reagents such as ethanol or acetone. They have the advantage of high detection accuracy, but have the disadvantages of being cumbersome, time-consuming, labor-intensive, inefficient, and destructive to the leaves.

[0004] In recent years, spectral analysis technology has been widely used in the detection of chlorophyll content in plant leaves. The chlorophyll content detection method based on spectral analysis technology has the advantages of simple operation, fast detection, and non-destructiveness to leaves, but the detection accuracy is not as good as the chemical method. This method is to construct a predictive relationship model between the chlorophyll content and spectral index of plant leaves, input the spectral index of the plant leaves to be tested into the predictive relationship model, and obtain the chlorophyll content of the plant leaves to be tested. The existing method for calculating the spectral index is to collect the spectral reflectance NIR of the plant leaves in the near-infrared band and the spectral reflectance R in the red light band, and calculate the spectral index NDVI according to the formula NDVI=(NIR-R) / (NIR+R). The method for calculating the spectral index has the following disadvantages: (1) These fixed wavelength combinations may not perform optimally for different plant leaf types and plant leaf growth stages; (2) The parameters of the spectral index calculation formula were not adjusted according to the dynamic changes of spectral data, resulting in poor adaptability under different environmental and plant leaf conditions; (3) It is easily affected by noise and changes in spectral reflectance, making it difficult to ensure its universal applicability and accuracy; (4) The complex nonlinear relationship between chlorophyll and spectral reflectance is not fully considered. In particular, the response relationship between chlorophyll and spectral reflectance under different wavelength combinations is not a simple linear relationship.

[0005] In summary, the existing chlorophyll content detection methods based on spectral analysis technology rely on the linear or simple nonlinear relationship between the sensitive wavelength in the spectral image and the chlorophyll content. They often have limitations when faced with complex wavelength combinations and nonlinear characteristics, which affects the detection accuracy and has poor applicability. Summary of the invention

[0006] In order to solve the above technical problems, the present invention provides a method for detecting chlorophyll content in plant leaves, which can perform non-destructive detection on plant leaves, improves detection accuracy, and can be applied to different plant leaves in different growth environments, with strong applicability.

[0007] In order to solve the above problems, the present invention adopts the following technical solutions: A method for detecting chlorophyll content in plant leaves of the present invention comprises the following steps: S1: Collecting hyperspectral reflectance data of a plant leaf sample to be tested, wherein the hyperspectral reflectance data includes n wavelength pairs and the spectral reflectance corresponding to each wavelength in the wavelength pair, where n>1; S2: Calculate the spectral index corresponding to the plant leaf sample to be tested based on the hyperspectral reflectance data; S3: The spectral index is input into a chlorophyll content prediction model, and the chlorophyll content prediction model outputs the chlorophyll content of the plant leaf sample to be tested.

[0008] Preferably, the chlorophyll content prediction model is obtained by the following method: N1: prepare g plant leaf samples of the same type as the plant leaf samples to be tested as training samples; N2: Collect the hyperspectral reflectance data of each training sample and use the chemical method to detect the chlorophyll content of each training sample; N3: Calculate the spectral index corresponding to each training sample based on the hyperspectral reflectance data; N4: The chlorophyll content and spectral index of each training sample constitute the sample data of the training sample. All sample data are input into the machine learning model for training, and the trained machine learning model is used as the chlorophyll content prediction model.

[0009] Preferably, the method for collecting hyperspectral reflectance data of plant leaf samples is as follows: M1: Collect spectral data of plant leaf samples; M2: n1 wavelength pairs are selected in the wavelength interval [550nm, 680nm), n2 wavelength pairs are selected in the wavelength interval [680nm, 750nm), and n3 wavelength pairs are selected in the wavelength interval [750nm, 850nm], where n=n1+n2+n3, and the wavelength pairs include two wavelengths of different sizes; M3: Calculate the spectral reflectance of the plant leaf sample at each wavelength in each wavelength pair based on the spectral data.

[0010] Preferably, the formula for calculating the spectral index corresponding to the plant leaf sample based on the hyperspectral reflectance data is as follows: , , Where CI is the spectral index, (λ i1 ,λ i2 ) is the i-th wavelength pair, λ i1 is the larger wavelength in the i-th wavelength pair, λ i2 is the smaller wavelength in the i-th wavelength pair, λ i1 ≠λ i2 , w i is the weight corresponding to the i-th wavelength pair, a is a positive number, is the plant leaf sample at wavelength λ i1 The spectral reflectance at is the plant leaf sample at wavelength λ i2 The spectral reflectance at is the segment index corresponding to the i-th wavelength pair, 1≤i≤n.

[0011] The characteristics and changing trends of spectral data in different bands (visible light region, red edge region, near infrared region) are different. The adaptive segmentation index k i The index can be flexibly adjusted according to the band characteristics, so that the spectral index calculation formula can apply different degrees of nonlinearity in high reflectance and low reflectance bands, respectively, to enhance the adaptability of the chlorophyll content detection method.

[0012] Preferably, the absolute value of the difference between the two wavelengths in the wavelength pair is greater than or equal to 10 nm.

[0013] Preferably, n2>n1=n3.

[0014] Preferably, the wavelengths in any two wavelength pairs are different.

[0015] Preferably, the weight w corresponding to the i-th wavelength pair is i The calculation formula is as follows: , , , Among them, H i1 is the wavelength λ i1 The corresponding information entropy, H i2 is the wavelength λ i2 The corresponding information entropy, p k (i1) is the wavelength λ i1 The spectral reflectance at is R i1 The probability that the chlorophyll content of the plant leaf sample is in the kth chlorophyll content interval, pk (i2) is the wavelength λ i2 The spectral reflectance at is R i2 The probability that the chlorophyll content of the plant leaf sample is in the kth chlorophyll content interval, m is the number of chlorophyll content intervals, 1≤k≤m.

[0016] Information entropy reflects the uncertainty or amount of information in the data. In spectral data, spectral reflectances of different wavelengths often carry different information, and their contributions can be measured by information entropy. The weight is calculated by the inverse of the information entropy of the spectral reflectance distribution. The smaller the information entropy, the larger the weight, indicating that the corresponding wavelength pair is more important; the larger the information entropy, the smaller the weight, indicating that the corresponding wavelength pair is less important.

[0017] Preferably, the machine learning model is a partial least squares regression model, a random forest model or a gradient boosting decision tree model.

[0018] Preferably, in step N2, the chlorophyll content of each training sample is detected by spectrophotometry or high performance liquid chromatography.

[0019] The beneficial effects of the present invention are as follows: (1) It is capable of performing non-destructive testing on plant leaves, selecting multiple wavelength pairs in the visible light region (550nm-680nm), the red edge region (680nm-750nm), and the near infrared region (750nm-850nm) and extracting the spectral reflectance corresponding to each wavelength in the wavelength pair, and calculating the spectral index in combination with the spectral index calculation formula. The spectral index calculation formula introduces an adaptive segmented index, thereby improving the detection accuracy, and can be applied to different plant leaves in different growth environments, with strong applicability and strong anti-noise ability. (2) The spectral index calculation formula introduces an adaptive weight, which is determined by the information entropy corresponding to the wavelength, further improving the detection accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 is a flow chart of Example 1; Figure 2 is a relationship diagram of the detection values ​​obtained by using the existing chlorophyll content detection method based on spectral analysis technology and the chemical method to detect the chlorophyll content of Perilla leaves in Example 2; Figure 3 This is a relationship diagram of the detection values ​​obtained by detecting the chlorophyll content of Perilla leaves using the method of this embodiment and the chemical method in Example 2. DETAILED DESCRIPTION

[0021] The technical solution of the present invention is further specifically described below through embodiments and in conjunction with the accompanying drawings.

[0022] Embodiment 1: A method for detecting chlorophyll content in plant leaves according to this embodiment, such as Figure 1 As shown, the following steps are included: S1: Collecting hyperspectral reflectance data of the plant leaf sample to be tested, the hyperspectral reflectance data includes n wavelength pairs and the spectral reflectance corresponding to each wavelength in the wavelength pair, and the spectral reflectance corresponding to a wavelength is the spectral reflectance of the plant leaf sample at the wavelength; S2: Calculate the spectral index corresponding to the plant leaf sample to be tested based on the hyperspectral reflectance data; S3: The spectral index is input into a chlorophyll content prediction model, and the chlorophyll content prediction model outputs the chlorophyll content of the plant leaf sample to be tested.

[0023] The chlorophyll content prediction model was obtained by the following method: N1: prepare g plant leaf samples of the same type as the plant leaf samples to be tested as training samples; N2: Collect the hyperspectral reflectance data of each training sample and use chemical methods (such as spectrophotometry, high performance liquid chromatography, etc.) to detect the chlorophyll content of each training sample; N3: Calculate the spectral index corresponding to each training sample based on the hyperspectral reflectance data; N4: The chlorophyll content and spectral index of each training sample constitute the sample data of the training sample. All sample data are input into the machine learning model for training. The trained machine learning model is used as the chlorophyll content prediction model. The machine learning model is a partial least squares regression model, a random forest model or a gradient boosting decision tree model.

[0024] The method for collecting hyperspectral reflectance data of plant leaf samples is as follows: M1: Collect spectral data of plant leaf samples.

[0025] A halogen lamp is used as a light source to illuminate the plant leaf sample, and a spectrometer is used to detect the spectrum data of the plant leaf sample. The halogen lamp is a 150W ring halogen lamp.

[0026] M2: n1 wavelength pairs are selected in the wavelength range [550nm, 680nm), n2 wavelength pairs are selected in the wavelength range [680nm, 750nm), and n3 wavelength pairs are selected in the wavelength range [750nm, 850nm], where n=n1+n2+n3, and the wavelength pairs include two wavelengths of different sizes.

[0027] The absolute value of the difference between the two wavelengths in the wavelength pair is greater than or equal to 10nm, ensuring that the spectral difference is significant, while avoiding the selection of too close wavelengths to cause redundant information or noise amplification. n2>n1=n3, the red edge region of the wavelength interval [680nm,750nm) is most sensitive to chlorophyll, and its changes directly reflect the photosynthesis characteristics of plant leaves, so the number of wavelength pairs in this interval is increased. The wavelengths in any two wavelength pairs are different, that is, the wavelengths in all wavelength pairs are extracted to form a wavelength set, and there are no identical wavelengths in the wavelength set.

[0028] For example: in the wavelength range [550nm, 680nm), select 2 wavelength pairs (580, 590) and (630, 640); in the wavelength range [680nm, 750nm), select 4 wavelength pairs (680, 690), (695, 705), (710, 720), (725, 735); in the wavelength range [750nm, 850nm], select 2 wavelength pairs (770, 780) and (800, 810).

[0029] M3: Calculate the spectral reflectance of the plant leaf sample at each wavelength in each wavelength pair based on the spectral data.

[0030] The formula for calculating the spectral index corresponding to the plant leaf sample based on the hyperspectral reflectance data is as follows: , , Where CI is the spectral index, (λ i1 ,λ i2 ) is the i-th wavelength pair, λ i1 is the larger wavelength in the i-th wavelength pair, λ i2 is the smaller wavelength in the i-th wavelength pair, λ i1 ≠λ i2 , w i is the weight corresponding to the i-th wavelength pair, a is a positive number, is the plant leaf sample at wavelength λ i1 The spectral reflectance at is the plant leaf sample at wavelength λ i2 The spectral reflectance at is the segment index corresponding to the i-th wavelength pair, 1≤i≤n.

[0031] A corresponding value is calculated based on a wavelength pair and added in a polynomial form to enhance the expression of spectral features. The characteristics and change trends of spectral data in different bands (visible light region, red edge region, near infrared region) are different. The adaptive segmentation index k iThe index can be flexibly adjusted according to the band characteristics, so that the spectral index calculation formula can apply different degrees of nonlinearity in high reflectance and low reflectance bands, respectively, to enhance the adaptability of the chlorophyll content detection method.

[0032] Since different bands have different sensitivities to chlorophyll, the adaptive segmentation index k i , the response can be adjusted specifically in different bands, so as to more accurately capture the differences between bands and ultimately improve the prediction effect.

[0033] The wavelength range (550nm-680nm) is the visible light region: in the visible light band, chlorophyll absorbs light strongly, and the difference between the reflectance bands can reflect the health of plant leaves. The use of exponential attenuation can effectively control the change in reflectance under different plant leaf conditions and avoid spectral index instability caused by excessive differences.

[0034] The wavelength range (680nm-750nm) is the red edge region: the red edge band is highly sensitive to chlorophyll and plant leaf health. The introduction of an adjustment coefficient of 1.5 amplifies the contribution of the red edge band, so that changes in this band can more significantly affect the spectral index, thereby increasing sensitivity to changes in plant leaf status.

[0035] The wavelength range (750nm-850nm) is the near-infrared region: the near-infrared band reflects the structure and health of plant leaves. Healthy plant leaves usually have strong reflection in this band. The use of logarithmic functions can smooth the spectral reflectance differences and avoid the higher spectral reflectance in the near-infrared band from affecting the index too drastically, thereby maintaining smooth changes in the model.

[0036] This adaptive segment index k i The spectral index calculation formula is allowed to apply different degrees of nonlinearity under different spectral reflectance conditions, so that the spectral index calculation formula has better adaptability to high reflectance areas (such as near-infrared band) and low reflectance areas (such as visible light band), thereby improving the prediction accuracy of chlorophyll.

[0037] This scheme can perform non-destructive testing on plant leaves. It selects multiple wavelength pairs in the three wavelength ranges of visible light, red edge, and near-infrared, respectively, and extracts the spectral reflectance corresponding to each wavelength in the wavelength pair. The spectral index is calculated by combining the spectral index calculation formula. The spectral index calculation formula introduces an adaptive segmented index, which can dynamically adjust the response characteristics of each spectral band, and can better capture the complex nonlinear relationship between spectral data and chlorophyll content, thereby improving the detection accuracy. It can be applied to different plant leaves in different growth environments, has strong applicability, and strong noise resistance, so that this method can still maintain a high prediction accuracy when facing high-noise data.

[0038] The weights corresponding to the wavelength pairs are preset values. For example, they can be obtained using the following two methods: (1) The wavelength pairs belonging to the visible light region, the wavelength pairs belonging to the red edge region, and the wavelength pairs belonging to the near infrared region are respectively pre-set with corresponding weight values; (2) Pre-calculate the mutual information corresponding to each wavelength in the wavelength pair, and multiply the mutual information corresponding to the two wavelengths in the wavelength pair by the value obtained as the weight value corresponding to the wavelength pair; Calculate wavelength λ i1 The corresponding mutual information is calculated as: Take a number of plant leaves of the same type as the plant leaf sample to be tested, and detect the wavelength λ of each plant leaf. i1 The spectral reflectance at the wavelength λ is used to detect the chlorophyll content of each plant leaf by chemical method, and the mutual information between the spectral reflectance and the chlorophyll content is calculated. i1 The corresponding mutual information.

[0039] Embodiment 2: A method for detecting chlorophyll content in plant leaves of this embodiment further includes the following method based on the method of Embodiment 1: The weight w corresponding to the i-th wavelength pair i The calculation formula is as follows: , , , Among them, H i1 is the wavelength λ i1 The corresponding information entropy, H i2 is the wavelength λ i2 The corresponding information entropy, p k (i1) is the wavelength λ i1 The spectral reflectance at is R i1 The probability that the chlorophyll content of a plant leaf sample is in the kth chlorophyll content interval (i.e., if a plant leaf sample is at wavelength λ i1 The spectral reflectance at is R i1 , then the probability that the plant leaf sample is in the kth chlorophyll content interval is p k (i1)), p k (i2) is the wavelength λ i2 The spectral reflectance at is R i2 The probability that the chlorophyll content of the plant leaf sample is in the kth chlorophyll content interval, m is the number of chlorophyll content intervals, 1≤k≤m.

[0040] Compared with the method in Example 1, the method in this embodiment optimizes the weight wi The calculation method of spectral index is the same as that of spectral index. The calculation formula of spectral index introduces adaptive weights, which are determined by the information entropy corresponding to the wavelength, further improving the detection accuracy. Information entropy reflects the uncertainty or amount of information in the data. In spectral data, spectral reflectances of different wavelengths often carry different information, and their contributions can be measured by information entropy. The weight is calculated by the inverse of the information entropy of the spectral reflectance distribution. The smaller the information entropy, the larger the weight, indicating that the corresponding wavelength pair is more important; the larger the information entropy, the smaller the weight, indicating that the corresponding wavelength pair is less important.

[0041] In this embodiment, the information entropy corresponding to the wavelength is a measure of the distribution of spectral reflectance, indicating the degree of "dispersion" of the spectral reflectance of the wavelength in different chlorophyll content ranges. The larger the information entropy, the wider the distribution of spectral reflectance and the less certain it is; the smaller the information entropy, the more concentrated the distribution of spectral reflectance and the more certain it is. The weight w i It is the inverse of information entropy, that is, the smaller the information entropy (i.e., the more concentrated the spectral reflectance distribution), the higher the weight; conversely, the larger the information entropy (i.e., the more dispersed the spectral reflectance distribution), the lower the weight. A high weight means that the spectral reflectance distribution is more concentrated, that is, most of the spectral reflectance values ​​of this wavelength are concentrated within a certain chlorophyll content range, with less fluctuation, which means that this wavelength is more sensitive to changes in chlorophyll content, which means that this wavelength has a higher certainty for chlorophyll content and is more important for predicting chlorophyll content.

[0042] p k (i1), p k The value of (i2) is obtained by looking up the table. For example: the plant leaves to be tested are perilla leaves, and the chlorophyll content of perilla leaves is divided into 4 chlorophyll content intervals (i.e., k=4), the first chlorophyll content interval is [0.5,1.5) mg / g, the second chlorophyll content interval is [1.5,2.5) mg / g, the third chlorophyll content interval is [2.5,3.5) mg / g, and the fourth chlorophyll content interval is [3.5,4.5] mg / g; The hyperspectral reflectance data of the perilla leaves to be tested are collected. The hyperspectral reflectance data include n wavelength pairs and the spectral reflectance corresponding to each wavelength in the wavelength pair. Assuming that the nth wavelength pair is (800, 810), the spectral reflectance of the perilla leaves to be tested at a wavelength of 800nm ​​is detected to be 0.47. According to the spectral reflectance of the perilla leaves to be tested at a wavelength of 800nm ​​being 0.47, the table is looked up to obtain that the probability that the chlorophyll content of the perilla leaves to be tested is located in the first chlorophyll content interval is 10% (i.e., p1(n2) is 10%), the probability of being located in the second chlorophyll content interval is 50% (i.e., p2(n2) is 50%), the probability of being located in the third chlorophyll content interval is 30% (i.e., p3(n2) is 30%), and the probability of being located in the fourth chlorophyll content interval is 10% (i.e., p4(n2) is 10%).

[0043] Take a plurality of perilla leaves, respectively use the existing chlorophyll content detection method based on spectral analysis technology, the method of this embodiment, and the chemical method to detect the chlorophyll content of the perilla leaves. The relationship between the detection value of the chlorophyll content detection method based on spectral analysis technology and the detection value of the chemical method is shown in the figure Figure 2 As shown, the relationship between the detection value of the method of this embodiment and the detection value of the chemical method is shown in Figure 3 As shown. Figure 2 , Figure 3 As can be seen in Figure 3 The two-dimensional scatter points composed of the detection values ​​of the method of this embodiment and the detection values ​​of the chemical method are closer to the 1:1 line, indicating that the detection accuracy of the method of this embodiment is higher than that of the existing chlorophyll content detection method based on spectral analysis technology.

Claims

1. A method for detecting chlorophyll content in plant leaves, characterized in that: The following steps are involved: S1: Collecting hyperspectral reflectance data of a plant leaf sample to be tested, wherein the hyperspectral reflectance data includes n wavelength pairs and the spectral reflectance corresponding to each wavelength in the wavelength pair, where n>1; S2: Calculate the spectral index corresponding to the plant leaf sample to be tested based on the hyperspectral reflectance data; S3: The spectral index is input into a chlorophyll content prediction model, and the chlorophyll content prediction model outputs the chlorophyll content of the plant leaf sample to be tested.

2. A method for detecting chlorophyll content in plant leaves according to claim 1, characterized in that: The chlorophyll content prediction model is obtained by the following method: N1: prepare g plant leaf samples of the same type as the plant leaf samples to be tested as training samples; N2: Collect the hyperspectral reflectance data of each training sample and use the chemical method to detect the chlorophyll content of each training sample; N3: Calculate the spectral index corresponding to each training sample based on the hyperspectral reflectance data; N4: The chlorophyll content and spectral index of each training sample constitute the sample data of the training sample. All sample data are input into the machine learning model for training, and the trained machine learning model is used as the chlorophyll content prediction model.

3. A method for detecting chlorophyll content in plant leaves according to claim 1, characterized in that: The method for collecting hyperspectral reflectance data of plant leaf samples is as follows: M1: Collect spectral data of plant leaf samples; M2: n1 wavelength pairs are selected in the wavelength interval [550nm, 680nm), n2 wavelength pairs are selected in the wavelength interval [680nm, 750nm), and n3 wavelength pairs are selected in the wavelength interval [750nm, 850nm], where n=n1+n2+n3, and the wavelength pairs include two wavelengths of different sizes; M3: Calculate the spectral reflectance of the plant leaf sample at each wavelength in each wavelength pair based on the spectral data.

4. A method for detecting chlorophyll content in plant leaves according to claim 3, characterized in that: The formula for calculating the spectral index corresponding to the plant leaf sample based on the hyperspectral reflectance data is as follows: ， ， Where CI is the spectral index, (λ i1 ,λ i2 ) is the i-th wavelength pair, λ i1 is the larger wavelength in the i-th wavelength pair, λ i2 is the smaller wavelength in the i-th wavelength pair, λ i1 ≠λ i2 , w i is the weight corresponding to the i-th wavelength pair, a is a positive number, is the plant leaf sample at wavelength λ i1 The spectral reflectance at is the plant leaf sample at wavelength λ i2 The spectral reflectance at is the segment index corresponding to the i-th wavelength pair, 1≤i≤n.

5. A method for detecting chlorophyll content in plant leaves according to claim 3, characterized in that: An absolute value of a difference between two wavelengths in the wavelength pair is greater than or equal to 10 nm.

6. A method for detecting chlorophyll content in plant leaves according to claim 3, characterized in that: The n2>n1=n3.

7. A method for detecting chlorophyll content in plant leaves according to claim 3, characterized in that: The wavelengths in any two wavelength pairs are different.

8. A method for detecting chlorophyll content in plant leaves according to claim 4, characterized in that: The weight w corresponding to the i-th wavelength pair i The calculation formula is as follows: ， ， ， Among them, H i1 is the wavelength λ i1 The corresponding information entropy, H i2 is the wavelength λ i2 The corresponding information entropy, p k (i1) is the wavelength λ i1 The spectral reflectance at is R i1 The probability that the chlorophyll content of the plant leaf sample is in the kth chlorophyll content interval, p k (i2) is the wavelength λ i2 The spectral reflectance at is R i2 The probability that the chlorophyll content of the plant leaf sample is in the kth chlorophyll content interval, m is the number of chlorophyll content intervals, 1≤k≤m.

9. A method for detecting chlorophyll content in plant leaves according to claim 2, characterized in that: The machine learning model is a partial least squares regression model, a random forest model or a gradient boosting decision tree model.

10. A method for detecting chlorophyll content in plant leaves according to claim 2, characterized in that: In the step N2, the chlorophyll content of each training sample is detected by spectrophotometry or high performance liquid chromatography.

Images