Hyperspectrum-based citrus leaf diagnosis system and diagnosis method

Abstract

The invention provides a hyperspectrum-based citrus leaf lesion diagnosis system and a hyperspectrum-based citrus leaf lesion diagnosis method, and the hyperspectrum-based citrus leaf lesion diagnosis system and the hyperspectrum-based citrus leaf lesion diagnosis method disclosed by the invention have the advantages that on the basis of analyzing hyperspectral imaging data of citrus leaves without diseases, lack of nutrients, black spots and yellow shoot; three parameters of yellow wave band reflectivity, infrared wave band slope and inflection point wavelength are used as characteristic quantities, and classification of four types of blades is realized by applying a support vector machine (RBF-SVM) classification model based on a Gaussian radial basis kernel function. The method solves the problems that when citrus tree disease information is detected through a field detection method at present, long-time observation with eyes is needed, subjective judgment of observers is depended, and misjudgment is likely to be caused; according to the present invention, the problem that the citrus tree disease information detection by using the chemical detection method needs the special person to detect by using the professional equipment so as not to accurately and rapidly detect each production stage of the citrus tree can be solved, and the citrus leaf disease can be rapidly and accurately diagnosed.

Description

Technical Field

[0001] This invention relates to the field of leaf diagnostic technology, and in particular to a hyperspectral-based diagnostic system and method for citrus leaves. Background Technology

[0003] The existing methods for detecting citrus tree diseases mainly include field testing and chemical testing. Field testing requires observers to observe for extended periods and relies heavily on their subjective judgment, which can easily lead to misjudgments. Chemical testing, on the other hand, is more complex and requires specialized personnel and equipment, which limits its widespread application and makes it impossible to accurately and quickly detect diseases at various stages of citrus growth. Summary of the Invention

[0004] This invention discloses a hyperspectral-based leaf-sensing diagnostic system and method, which solves the problems of current field testing methods for detecting citrus tree diseases, which require long-term visual observation and rely heavily on the observer's subjective judgment, easily leading to misjudgment; and chemical testing methods for detecting citrus tree diseases, which require specialized personnel and equipment, and cannot accurately and quickly detect diseases at various stages of citrus production. This invention enables rapid and accurate diagnosis of diseases in citrus leaves.

[0005] To achieve the above objectives, the technical solution of the present invention is specifically implemented as follows:

[0006] This invention discloses a hyperspectral-based citrus leaf lesion diagnosis system, comprising a dark box, a micro-displacement platform, a light source, a hyperspectral imager, a data processing unit, and a display unit. The dark box provides a diagnostic environment free from ambient light interference for the citrus leaf to be diagnosed. The micro-displacement platform fixes the citrus leaf and adjusts its position. Several light sources are fixed within the dark box to illuminate the citrus leaf and generate reflected signal light. The hyperspectral imager receives the reflected signal light, forms a hyperspectral image of the citrus leaf, and transmits the image to the data processing unit. The data processing unit receives the hyperspectral image of the citrus leaf from the hyperspectral imager and diagnoses whether lesions are present on the leaf. The display unit displays the type of lesion on the citrus leaf.

[0007] Furthermore, the light source is a 20W halogen lamp.

[0008] Furthermore, two light sources are fixed inside the dark box.

[0009] Furthermore, the wavelength range of the hyperspectral imager is 392–1003 nm, and the spectral resolution is 2.8 nm.

[0010] Another aspect of the present invention discloses a diagnostic method based on a hyperspectral sensory leaf lesion diagnostic system, comprising the following steps:

[0011] Several disease-free citrus leaves and various diseased citrus leaves were collected separately.

[0012] The collected citrus leaves without disease and various diseased citrus leaves were divided into a modeling set and a prediction set.

[0013] Collect hyperspectral images of several disease-free citrus leaves and citrus leaves with various diseases;

[0014] Several regions of interest (ROIs) were selected on the hyperspectral images of each disease-free citrus leaf and different diseased citrus leaves, and the average spectrum of the ROI for each leaf was calculated.

[0015] Calculate three characteristic parameters: yellow band reflectance, infrared band slope, and inflection point wavelength of disease-free citrus leaves and citrus leaves with various diseases.

[0016] Calculate the correlation between the three feature parameters;

[0017] Using three characteristic parameters—yellow band reflectance, infrared band slope, and inflection point wavelength—of disease-free citrus leaves and various diseased citrus leaves in the modeling set, an RBF-SVM recognition model was established, and the optimal recognition model was obtained.

[0018] The effectiveness of the optimal identification model was validated using disease-free citrus leaves and various diseased citrus leaves from the prediction set.

[0019] The optimal recognition model was used to diagnose citrus leaves.

[0020] Furthermore, after acquiring hyperspectral images of several disease-free citrus leaves and various diseased citrus leaves, the acquired hyperspectral images were calibrated in black and white.

[0021] Furthermore, the collected citrus leaves with different diseases included nutrient deficiency leaves, black spot disease leaves, and Huanglongbing (HLB) leaves.

[0022] Furthermore, when selecting the Region of Interest (ROI) on each leaf, avoid the main vein and select the position in the middle of the left or right side of the leaf vein.

[0023] Furthermore, after obtaining the average spectrum of the ROI for each leaf, the obtained average spectrum is processed using a second-order SG smoothing algorithm.

[0024] Furthermore, the expression for the RBF-SVM recognition model is as follows:

[0025]

[0026] Among them, a i For Lagrange multipliers, y i Here, b represents the label of the corresponding sample result, and K(x) represents the intercept. i ·x j ) is the RBF kernel function, and

[0027]

[0028] x i x j Let x be the eigenvector. i ∈R n .

[0029] Beneficial technical effects:

[0030] 1. This invention discloses a hyperspectral-based diagnostic system and method for citrus leaf lesions, comprising a dark box, a micro-displacement platform, a light source, a hyperspectral imager, a data processing unit, and a display unit. The dark box provides a diagnostic environment free from ambient light interference for the citrus leaf to be diagnosed; the micro-displacement platform fixes the citrus leaf and adjusts its position; several light sources are fixed within the dark box to irradiate the citrus leaf and generate reflected signal light; the hyperspectral imager receives the reflected signal light, forms a hyperspectral image of the citrus leaf, and displays the formed hyperspectral image of the citrus leaf. The image is transmitted to the data processing unit; the data processing unit receives the hyperspectral image of the citrus leaf to be diagnosed sent by the hyperspectral imager and diagnoses whether there are lesions on the citrus leaf; the display unit displays the types of lesions on the citrus leaf to be diagnosed. This solves the problems of current field detection methods for detecting citrus tree diseases, which require long-term visual observation and rely heavily on the observer's subjective judgment, easily leading to misjudgment; and chemical detection methods for detecting citrus tree diseases, which require specialized personnel and equipment, and cannot accurately and quickly detect citrus tree diseases at various production stages. This method can quickly and accurately diagnose citrus leaf diseases.

[0031] 2. In this invention, after acquiring hyperspectral images of several disease-free citrus leaves and various diseased citrus leaves, the acquired hyperspectral images are black and white calibrated to compensate for the influence of dark current and uneven illumination in the system, making the diagnostic results more accurate.

[0032] 3. In this invention, an RBF-SVM recognition model is established using three characteristic parameters: yellow band reflectance, infrared band slope, and inflection point wavelength of disease-free citrus leaves and various diseased citrus leaves. This enables rapid detection of citrus leaf diseases with high accuracy. Attached Figure Description

[0033] To more clearly illustrate the technical solution of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below.

[0034] Figure 1 This is a schematic diagram of the structure of a hyperspectral-based citrus leaf lesion diagnostic system according to the present invention;

[0035] Figure 2 This is a flowchart of the diagnostic method of a hyperspectral-based citrus leaf lesion diagnostic system according to the present invention.

[0036] Figure 3a The average spectrum of the ROI of disease-free leaves, along with its first and second derivative spectral images.

[0037] Figure 3b The average spectrum of the ROI of Huanglongbing leaves and its first and second derivative spectral images;

[0038] Figure 3c The average spectrum of the ROI of the nutrient-deficient leaf and its first and second derivative spectral images.

[0039] Figure 3d The average spectrum of the ROI and its first and second derivatives are shown in the image.

[0040] Figure 4 Extreme point images of disease-free leaves, Huanglongbing (HLB) leaves, nutrient-deficient leaves, and black spot leaves;

[0041] Figure 5 Images showing the inflection points of disease-free leaves, leaves affected by Huanglongbing (HLB), leaves with nutrient deficiencies, and leaves with black spot disease;

[0042] Figure 6 Two-dimensional plots of reflectance in the yellow band for disease-free leaves, leaves with Huanglongbing (HLB), leaves with nutrient deficiency, and leaves with black spot disease;

[0043] Figure 7 Two-dimensional images of disease-free leaves, Huanglongbing leaves, nutrient-deficient leaves, and black spot leaves were obtained using inflection points as feature values.

[0044] Figure 8 The slope images of disease-free leaves, Huanglongbing leaves, nutrient-deficient leaves, and black spot leaves in the infrared band;

[0045] Figure 9 Reflectance images of disease-free leaves, leaves with Huanglongbing (HLB), leaves with nutrient deficiency, and leaves with black spot disease in the infrared band;

[0046] Figure 10 Two-dimensional images of leaves without disease, leaves with Huanglongbing (HLB), leaves with nutrient deficiency, and leaves with black spot disease, characterized by infrared reflectance. Figure 11 Three-dimensional images of the selected disease-free leaves, Huanglongbing leaves, nutrient-deficient leaves, and black spot leaves, characterized by yellow band reflectance, infrared band slope, inflection point, and infrared band reflectance.

[0047] Among them, 1-dark box, 2-micro-displacement platform, 3-light source, 4-hyperspectral imager, 5-data processing unit, and 6-display unit. Detailed Implementation

[0048] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0049] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0050] Unless otherwise specifically stated, the relative arrangement, numerical expressions, and values ​​of the components and steps set forth in these embodiments do not limit the scope of the invention. It should also be understood that, for ease of description, the dimensions of the various parts shown in the drawings are not drawn to actual scale. Techniques, methods, and devices known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and devices should be considered part of the specification. In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values. It should be noted that similar reference numerals and letters in the following figures denote similar items; therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.

[0051] In the description of this invention, it should be understood that the orientation or positional relationship indicated by directional terms such as "front, back, up, down, left, right", "horizontal, vertical, horizontal" and "top, bottom" is generally based on the orientation or positional relationship shown in the accompanying drawings, and is only for the convenience of describing this invention and simplifying the description. Unless otherwise stated, these directional terms do not indicate or imply that the device or element referred to must have a specific orientation or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation on the scope of protection of this invention; the directional terms "inner" and "outer" refer to the inner and outer contours relative to the outline of each component itself.

[0052] For ease of description, spatial relative terms such as "above," "on top of," "on the upper surface of," "above," etc., are used herein to describe the spatial positional relationship of a device or feature as shown in the figures to other devices or features. It should be understood that spatial relative terms are intended to encompass different orientations in use or operation beyond the orientation of the device as described in the figures. For example, if the device in the figures were inverted, a device described as "above" or "on top of" other devices or structures would subsequently be positioned as "below" or "under" other devices or structures. Thus, the exemplary term "above" can include both "above" and "below." The device may also be positioned in other different ways (rotated 90 degrees or in other orientations), and the spatial relative descriptions used herein will be interpreted accordingly.

[0053] Furthermore, it should be noted that the use of terms such as "first" and "second" to define components is merely for the purpose of distinguishing the corresponding components. Unless otherwise stated, the above terms have no special meaning and therefore should not be construed as limiting the scope of protection of this invention.

[0054] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0055] This invention discloses, in one aspect, a hyperspectral-based diagnostic system for citrus leaf diseases, see [link to relevant documentation]. Figure 1Specifically, the system includes a dark box 1, a micro-displacement platform 2, a light source 3, a hyperspectral imager 4, a data processing unit 5, and a display unit 6. The dark box 1 provides a diagnostic environment free from ambient light interference for the citrus leaf to be diagnosed. The micro-displacement platform 2 is used to fix the citrus leaf to be diagnosed and adjust its position. Several light sources 3 are fixed inside the dark box 1 to illuminate the citrus leaf and generate reflected signal light. Preferably, two light sources 3 are fixed inside the dark box 1, and the light sources 3 are 20W halogen lamps. The hyperspectral imager 4 receives the reflected signal light, forms a hyperspectral image of the citrus leaf to be diagnosed, and transmits the formed hyperspectral image to the data processing unit 5. Preferably, the wavelength range of the hyperspectral imager 4 is 392–1003 nm, and the spectral resolution is 2.8 nm. The data processing unit 5 receives the hyperspectral image of the citrus leaf to be diagnosed sent by the hyperspectral imager 4 and diagnoses whether there is a lesion on the citrus leaf. The display unit 6 displays the type of lesion on the citrus leaf to be diagnosed.

[0056] Another aspect of this invention discloses a diagnostic method for a citrus leaf lesion diagnostic system based on hyperspectral imaging, see [link to relevant documentation]. Figure 2 Specifically, it includes the following steps:

[0057] Several disease-free citrus leaves and various diseased citrus leaves were collected separately.

[0058] Specifically, the diseased leaves collected in this embodiment are common citrus leaf diseases, including nutrient deficiency leaves, black spot disease leaves and Huanglongbing (HLB) leaves. In this embodiment, 60 disease-free leaves (normal leaves), nutrient deficiency leaves, black spot disease leaves and HLB leaves were collected from a citrus planting base.

[0059] The collected citrus leaves without disease and various diseased citrus leaves were divided into a modeling set and a prediction set.

[0060] Specifically, in this embodiment, 50 leaves each of disease-free leaves (normal leaves), nutrient-deficient leaves, black spot disease leaves, and Huanglongbing leaves are randomly selected as the modeling set, and the remaining 10 leaves each of disease-free leaves (normal leaves), nutrient-deficient leaves, black spot disease leaves, and Huanglongbing leaves are selected as the prediction set.

[0061] Collect hyperspectral images of several disease-free citrus leaves and citrus leaves with various diseases;

[0062] The hyperspectral image of each citrus leaf is acquired using the hyperspectral-based citrus leaf disease diagnosis system disclosed in this invention. Preferably, to compensate for the effects of dark current and uneven illumination in the system, the acquired hyperspectral image is calibrated using the following formula:

[0063]

[0064] Among them, R λ For the hyperspectral image of the calibrated wavelength λ channel, B λ To create a completely black calibration image with the camera shutter wavelength λ channel closed, I λ The raw hyperspectral data for the wavelength λ channel, W λ This is a full-white image of a standard PTFE white board at wavelength λ.

[0065] Several regions of interest (ROIs) were selected on the hyperspectral images of each disease-free citrus leaf and different diseased citrus leaves, and the average spectrum of the ROI for each leaf was calculated.

[0066] Specifically, when acquiring the Region of Interest (ROI), the main vein should be avoided. A rectangular region with a horizontal and vertical pixel size of 5 should be selected at the middle of the left or right side of the vein as the ROI. In this embodiment, 20 ROIs are selected for each leaf. The average spectrum is obtained by calculating the average reflectance of each pixel in each ROI. Preferably, the average spectrum is processed using a second-order SG smoothing algorithm (i.e., calculating the first and second derivatives of the average spectrum). See [link to relevant documentation]. Figure 3a , 3b 3c and 3d are the average ROI spectra and their first and second derivatives for disease-free leaves, Huanglongbing leaves, nutrient-deficient leaves, and black spot leaves, respectively.

[0067] Calculate three characteristic parameters: yellow band reflectance, infrared band slope, and inflection point wavelength of disease-free citrus leaves and citrus leaves with various diseases.

[0068] Specifically, through Figure 3a , 3b The characteristics of leaves without disease, nutrient-deficient leaves, Huanglongbing leaves, and black spot disease leaves are not obvious when using the first or second derivative. Therefore, it is necessary to extract other characteristics to effectively distinguish these types of leaves.

[0069] Further treatment of disease-free leaves, nutrient-deficient leaves, Huanglongbing (HLB) leaves, and black spot leaves was conducted to determine the extreme points and inflection points, such as... Figure 4 and Figure 5 As shown, in Figure 4 In the analysis, the four leaf types exhibit a minimum value P1 between the orange and red bands (622nm–780nm), a maximum value P2 between blue (492nm) and P1, a minimum value P3 between violet (455nm) and P2, and a maximum value P4 in the red (770nm) to 900nm infrared region. Figure 4It can be seen that the total reflectance of the four types of blades from P3 to P1 varies significantly, reflecting the different reflective abilities of the blades for yellow light. Using the reflective area of ​​the P3 to P1 band as a characteristic, the characteristic values ​​of 50 blades were calculated and plotted in a two-dimensional graph as shown below. Figure 6 As shown.

[0070] analyze Figure 6 It can be seen that there is a clear dividing line between the yellow band reflectance of normal leaves and leaves with black spot disease, and the yellow band reflectance of leaves with Huanglongbing disease and leaves lacking nutrients. The yellow band reflectance of normal leaves and leaves with black spot disease lies below 36.05 on the y-axis, while the yellow band reflectance of leaves with Huanglongbing disease and leaves lacking nutrients lies above 38.65 on the y-axis. While yellow band reflectance can be used to classify normal leaves, leaves with black spot disease, and leaves with Huanglongbing disease and leaves lacking nutrients, it is insufficient to further distinguish between normal leaves and leaves with black spot disease, or between leaves with Huanglongbing disease and leaves lacking nutrients. To quickly distinguish between the four types of leaves, other characteristic parameters need to be introduced for further analysis. Figure 4 and Figure 5 It can be seen that the four types of blades have an inflection point around 700 nm, and these differences are obvious. Therefore, this inflection point is used as a characteristic value to determine the characteristics of the four types of blades, and a two-dimensional graph is drawn as follows. Figure 7 As shown.

[0071] analyze Figure 7 It can be seen that the inflection points of nutrient-deficient leaves and leaves with black spot disease are basically within 696.6 nm on the y-axis, while the inflection points of leaves with Huanglongbing disease and normal leaves are basically greater than 696.6 nm. Some inflection points of both types of leaves fall on the dividing line of 696.6 nm, and the inflection points of some Huanglongbing leaves are also less than 696.66 nm. Therefore, in general, nutrient-deficient leaves and leaves with black spot disease can be distinguished from Huanglongbing leaves and normal leaves by using the inflection points. However, during the classification process, errors will occur, and Huanglongbing leaves that are at the 696.6 nm dividing line or deviate from the dividing line cannot be correctly distinguished. The classification accuracy rate of the two types of leaves is 92%. Taking into account the inflection point and the reflectance results of the yellow band, a classification model was established, and the classification results are shown in Table 1.

[0072] Table 1 shows the results of two specific positive classification methods.

[0073]

[0074] From Table 1 and the above Figure 6 and Figure 7 It can be seen that the accuracy rates of classification using yellow band reflectance and inflection point as features are 92% for nutrient-deficient leaves, 91% for Huanglongbing leaves, 90% for normal leaves, and 91% for black spot disease leaves. The classification accuracy needs to be further improved.

[0075] To further improve the accuracy of classification, analysis Figure 4It can be seen that after point P4, the slopes of normal leaves, Huanglongbing leaves, nutrient-deficient leaves, and black spot diseased leaves are significantly different. The slopes obtained by selecting typical normal leaves, Huanglongbing leaves, nutrient-deficient leaves, and black spot diseased leaves are as follows: Figure 8 As shown. For the four types of leaves, starting from the maximum value P4 obtained from the infrared band to 900nm, the least squares method was used for fitting to obtain the slope of the straight line. The graph shows that the slope of the Huanglongbing (HLB) leaf band is less than 0, while the slopes of nutrient-deficient leaves, normal leaves, and black spot disease leaves are greater than 0. The relationship is: slope of black spot disease leaves > slope of nutrient-deficient leaves > slope of HLB leaves. The slope of black spot disease leaves is the largest and much greater than the other leaves. The slopes of nutrient-deficient leaves and HLB leaves are relatively close to 0. Figure 9 It can be seen that the reflectance of the infrared band can easily distinguish the leaves with black spot disease from the other three types of leaves. Among them, the leaves with Huanglongbing disease have the lowest reflectance and are at the bottom. However, the other three types of leaves cannot be accurately distinguished by reflectance alone, and the accuracy of the distinction is not high.

[0076] Determine the characteristic values ​​of 50 blades based on infrared reflectance, such as... Figure 10 As shown in the figure, it can be seen that the reflectance of leaves with black spot disease is significantly lower than that of the other three types of leaves in the infrared band. The leaves with Huanglongbing disease have the highest infrared reflectance at the top. There is overlap in the infrared reflectance of normal leaves and nutrient-deficient leaves. The classification boundaries of the various leaf characteristics are relatively blurred and the differentiation is not high.

[0077] Therefore, this embodiment of the invention uses a combination of three characteristic parameters: yellow band reflectivity, infrared band slope, and inflection point wavelength, to effectively distinguish between disease-free leaves, Huanglongbing leaves, nutrient-deficient leaves, and black spot disease leaves.

[0078] Calculate the correlation between the three feature parameters;

[0079] Specifically, it was not possible to successfully identify the four types of leaves using a single feature. The Pearson correlation coefficients between the three features were calculated as follows: Figure 11 As shown, the correlation is calculated using the following formula:

[0080] Two sets of feature data X: (X1, X2, ..., X...) n Y: (Y1, Y2, ..., Y) n Sample mean: Sample covariance: Then the sample Pcarson correlation coefficient: Among them, S X (sigma X) is the sample standard deviation of X.

[0081] As shown in the figure, the correlation coefficient between the yellow band reflectance characteristics and the inflection point characteristics is 0.1897, indicating no obvious linear relationship between them. The Pearson correlation coefficients between the yellow band reflectance characteristics and the infrared band reflectance characteristics and the infrared band slope characteristics are 0.7979 and -0.5932, respectively, showing a strong linear relationship. The Pearson correlation coefficients between the infrared band slope characteristics and the inflection point characteristics and the infrared band reflectance characteristics are -0.5775 and 0.7966, respectively, showing a strong correlation, with a stronger linear relationship with the infrared band reflectance. The Pearson correlation coefficient between the inflection point characteristics and the infrared band reflectance characteristics is 0.3548, showing a moderate linear correlation.

[0082] Using three characteristic parameters—yellow band reflectance, infrared band slope, and inflection point wavelength—of disease-free citrus leaves and various diseased citrus leaves in the modeling set, an RBF-SVM recognition model was established, and the optimal recognition model was obtained.

[0083] Specifically, the four types of leaves cannot be well distinguished by a single feature or even just two features. Therefore, this study considers using multiple features to identify the four types of leaves and compares their characteristics. Figure 4 and Figure 5 It is known that leaves with black spot disease can be easily distinguished by inflection point or infrared reflectance. Therefore, three characteristics—yellow band reflectance, infrared band slope, and inflection point—are used to distinguish the four types of leaves. A three-dimensional graph with the selected three types of leaves as coordinate axes is drawn as follows. Figure 11 As shown, analysis Figure 11 It is evident that the four types of blades exhibit significant separation in three-dimensional space. Therefore, a recognition model is established using three parameters: yellow band reflectivity, infrared band slope, and inflection point. Preferably, the expression for the RBF-SVM recognition model is as follows:

[0084]

[0085] Among them, a i For Lagrange multipliers, y i Here, b represents the label of the corresponding sample result, and K(x) represents the intercept. i ·x j ) is the RBF kernel function, and

[0086]

[0087] x i x j Let x be the eigenvector. i ∈R n .

[0088] In this embodiment of the invention, the three parameters of yellow band reflectance, inflection point, and infrared band slope of the extracted disease-free leaves, nutrient-deficient leaves, Huanglongbing leaves, and black spot leaves are used as feature recognition quantities. Fifty samples are selected as training samples, and the training data are fed into the RBF-SVM recognition model for training using a one-against-all method. The optimal values ​​of the training results σ are 25.1798, 3.6372, and 6.2207, respectively.

[0089] The effectiveness of the optimal identification model was validated using disease-free citrus leaves and various diseased citrus leaves from the prediction set.

[0090] Specifically, the remaining 10 samples were used as test samples. When the test samples were fed into the classification model, the accuracy rate was 100%. These three features can achieve rapid detection of citrus leaves, and the results show the effectiveness of the feature selection.

[0091] The optimal recognition model was used to diagnose citrus leaves.

[0092] The present invention discloses a hyperspectral-based diagnostic system and method for citrus leaf lesions. Based on the analysis of hyperspectral imaging data of disease-free, nutrient-deficient, black spot, and Huanglongbing citrus leaves, it proposes to use three parameters as feature quantities: yellow band reflectance, infrared band slope, and inflection point wavelength. A support vector machine (RBF-SVM) classification model based on Gaussian radial basis kernel function is applied to classify the four types of leaves. Test results show that the classification accuracy using these three feature quantities can reach 100%, effectively achieving rapid and accurate diagnosis of citrus leaves.

[0093] In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0094] The above embodiments are merely descriptions of preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A citrus leaf lesion diagnostic system based on hyperspectral imaging, characterized in that, include: Dark box (1) is used to provide a diagnostic environment free from ambient light for the citrus leaves to be diagnosed; The micro-displacement platform (2) is used to fix the citrus leaf to be diagnosed and to adjust the position of the citrus leaf to be diagnosed. Light source (3), several of the light sources (3) are fixed inside the dark box (1) to illuminate the citrus leaf to be diagnosed and generate reflected signal light; The hyperspectral imager (4) is used to receive reflected signal light, form a hyperspectral image of the citrus leaf to be diagnosed, and transmit the formed hyperspectral image of the citrus leaf to be diagnosed to the data processing unit (5). The data processing unit (5) is used to receive the hyperspectral image of the citrus leaf to be diagnosed sent by the hyperspectral imager (4) and diagnose whether there is a lesion in the citrus leaf to be diagnosed. Display unit (6) is used to display the types of lesions on the citrus leaves to be diagnosed.

2. The citrus leaf lesion diagnostic system based on hyperspectral imaging according to claim 1, characterized in that, The light source (3) is a halogen lamp with a power of 20W.

3. The citrus leaf lesion diagnostic system based on hyperspectral imaging according to claim 2, characterized in that, Two light sources (3) are fixed inside the dark box (1).

4. The citrus leaf lesion diagnostic system based on hyperspectral imaging according to claim 1, characterized in that, The wavelength range of the hyperspectral imager (4) is 392–1003 nm, and the spectral resolution is 2.8 nm.

5. A diagnostic method for a hyperspectral-based citrus leaf lesion diagnostic system as described in any one of claims 1-4, characterized in that, Includes the following steps: Several disease-free citrus leaves and various diseased citrus leaves were collected separately. The collected citrus leaves without disease and various diseased citrus leaves were divided into a modeling set and a prediction set. Collect hyperspectral images of several disease-free citrus leaves and citrus leaves with various diseases; Several regions of interest (ROIs) were selected on the hyperspectral images of each disease-free citrus leaf and different diseased citrus leaves, and the average spectrum of the ROI for each leaf was calculated. Calculate three characteristic parameters: yellow band reflectance, infrared band slope, and inflection point wavelength of disease-free citrus leaves and citrus leaves with various diseases. Calculate the correlation between the three feature parameters; Using three characteristic parameters—yellow band reflectance, infrared band slope, and inflection point wavelength—of disease-free citrus leaves and various diseased citrus leaves in the modeling set, an RBF-SVM recognition model was established, and the optimal recognition model was obtained. The effectiveness of the optimal identification model was validated using disease-free citrus leaves and various diseased citrus leaves from the prediction set. The optimal recognition model was used to diagnose citrus leaves.

6. The diagnostic method of the citrus leaf lesion diagnostic system based on hyperspectral imaging according to claim 5, characterized in that, After acquiring hyperspectral images of several disease-free citrus leaves and various diseased citrus leaves, the acquired hyperspectral images were calibrated in black and white.

7. The diagnostic method of the citrus leaf lesion diagnostic system based on hyperspectral imaging according to claim 5, characterized in that, The collected citrus leaves with different diseases included nutrient deficiency leaves, black spot disease leaves, and Huanglongbing (HLB) leaves.

8. The diagnostic method of the citrus leaf lesion diagnostic system based on hyperspectral imaging according to claim 5, characterized in that, When selecting a region of interest (ROI) on each leaf, avoid the main vein and select the area in the middle of the left or right side of the vein.

9. The diagnostic method of the citrus leaf lesion diagnostic system based on hyperspectral imaging according to claim 5, characterized in that, After obtaining the average spectrum of the ROI for each leaf, the average spectrum is processed using a second-order SG smoothing algorithm.

10. The diagnostic method of the citrus leaf lesion diagnostic system based on hyperspectral imaging according to claim 5, characterized in that, The expression for the RBF-SVM recognition model is as follows: Among them, a i For Lagrange multipliers, y i Here, b represents the label of the corresponding sample result, and K(x) represents the intercept. i ·x j ) is the RBF kernel function, and x i x j Let x be the eigenvector. i ∈R n .

Images