Method for Classifying Spores of Aerogenic Diseases of Greenhouse Crops Based on Microscopic Polarization Image Features

Abstract

The present invention discloses a method for classifying airborne disease spores of greenhouse crops based on microscopic polarization image features, which is applied to the technical field of classifying and identifying airborne disease spores, and includes: preprocessing the microscopic polarization image of airborne disease spores by using the Stokes vector method to obtain the polarization degree image and polarization angle image of airborne disease spores; adopting Gabor transform to convert the polarization angle image from the spatial domain to the frequency domain, extracting the texture features of the polarization angle image, and jointly using the relative light intensity distribution value of the polarization degree image as the input of the neural network to train and obtain an airborne disease spore classification model; inputting the image to be measured into the airborne disease spore classification model to obtain a classification result. The present invention can effectively reduce the interference of external environmental light, improve the recognition rate of disease spores with similar morphologies, and effectively improve the classification and recognition accuracy of airborne disease spores.

Description

Technical Field

[0001] The present invention relates to the technical field of airborne disease spore classification and recognition, and particularly to a method for classifying airborne disease spores of greenhouse crops based on microscopic polarization image features. Background Art

[0002] In recent years, the vegetable industry has developed rapidly, and tomatoes and cucumbers are deeply loved by consumers because of their rich taste and nutritional value. Tomatoes and cucumbers are inevitably exposed to biological stresses during the growth process, and the temperature and humidity conditions in the greenhouse environment are more conducive to the occurrence, prevalence, and spread of airborne fungal diseases. These airborne fungal diseases will become increasingly severe with the expansion of the cultivation area and the increase in the number of consecutive cropping years. In severe cases, it can cause a reduction in production, with a yield loss of 20 - 50%, or even a complete crop failure.

[0003] Currently, the diagnosis and control of greenhouse crop diseases are mainly based on the experience of producers and the results of conventional laboratory or on-site tests. Laboratory testing technologies mainly include electron microscopy testing technology, polymerase chain reaction (PCR), molecular biology testing technology, etc. These testing technologies can accurately determine the types of greenhouse crop diseases, but laboratory testing technologies have the disadvantages of being destructive, time-consuming, and labor-intensive. On-site conventional testing technologies mainly use spectral detection technology and image processing technology to detect known or specific crop diseases. These technologies can achieve the detection of crop diseases through statistical modeling inversion, and based on this, accurate guidance can be provided for timely prediction and treatment of diseases during the cultivation of greenhouse crops. However, these diagnostic technologies cannot achieve early warning before the epidemic of diseases, but can only detect them when the diseases occur, and by this time, the best window period for preventing and controlling crop diseases has been missed.

[0004] Actually, before the occurrence of airborne diseases in greenhouse crops and the large-scale spread of the diseases, the first thing that happens is the spread of airborne disease spores along with the air currents in the air. When the airborne disease spores in the spread come into contact with the plant leaves, they will enter the plant tissues through the stomata of the plant leaves and germinate, and then release more airborne disease spores to continue spreading among the plants. Therefore, as long as the spread information of the airborne disease spores can be obtained in real time from the transmission path, the prediction and forecasting of the airborne diseases in crops can be carried out in a timely manner. With the improvement of the spore trap by technicians, there are more and more spore traps sold on the market now, and using the spore trap to capture the airborne disease spores in the air has become a common method. For example, the existing portable spore trap, vehicle-mounted spore trap, and fixed spore trap all first capture the airborne disease spores in the air, and then use a microscope for identification, achieving a certain degree of detection of the airborne disease spores spreading in the air. However, the microscope is susceptible to the interference of the external environmental light when collecting the images of the disease spores, and the morphologies of different types of disease spores are similar. When classifying and identifying the disease spores based on the traditional microscopic image features, the accuracy is relatively low, and it is difficult to distinguish between similar disease spores.

[0005] Therefore, how to provide a classification method for airborne disease spores of greenhouse crops based on microscopic polarization image features that can effectively reduce the interference of the external environmental light, improve the recognition rate of disease spores with similar morphologies, and effectively improve the classification and recognition accuracy of airborne disease spores is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0006] In view of this, the present invention proposes a classification method for airborne disease spores of greenhouse crops based on microscopic polarization image features. By training a neural network-based classification model for airborne disease spores based on the relative light intensity distribution value of the polarization degree image of the airborne disease spores and the texture features of the polarization angle image of the airborne disease spores, the interference of the external environmental light is effectively reduced, the recognition rate of disease spores with similar morphologies is improved, and the classification and recognition accuracy of airborne disease spores is effectively improved.

[0007] In order to achieve the above object, the present invention adopts the following technical solutions:

[0008] A classification method for airborne disease spores of greenhouse crops based on microscopic polarization image features, comprising:

[0009] Step (1): Preprocess the microscopic polarization image of the airborne disease spores by using the Stokes vector method to obtain the polarization degree image and the polarization angle image of the airborne disease spores;

[0010] Step (2): Use Gabor transform to convert the polarization angle image from the spatial domain to the frequency domain, extract the texture features of the polarization angle image, and jointly use the relative light intensity distribution value of the polarization degree image as the input of the neural network to train and obtain an airborne disease spore classification model;

[0011] Step (3): Input the image to be measured into the airborne disease spore classification model to obtain the classification result.

[0012] Optionally, in step (1), use the Stokes vector method to preprocess the microscopic polarization image of airborne disease spores as follows:

[0013]

[0014] Among them, S0 is the total light field intensity of the incident light; S1 is the intensity difference of linear polarization at 0° and 90°; S2 is the intensity difference of linear polarization at 45° and 135°; S3 is the intensity difference of circular polarization in the right-handed and left-handed directions.

[0015] Optionally, in step (1), the calculation formula of the polarization degree image is as follows:

[0016]

[0017] Optionally, in step (1), the calculation formula of the polarization angle image is as follows:

[0018]

[0019] Optionally, in step (2), use Gabor transform to convert the polarization angle image from the spatial domain to the frequency domain as follows:

[0020]

[0021]

[0022]

[0023]

[0024] Among them, v is the wavelength of the Gabor filter; u is the direction of the Gabor kernel function; K is the total number of directions of the Gabor kernel function; σ / k is the size of the Gaussian window.

[0025] Optionally, in step (2), the texture features include: the contrast and entropy of airborne disease spores.

[0026] Optionally, in step (2), the method for obtaining the relative light intensity distribution value of the polarization degree image is specifically:

[0027] Save the degree of polarization image as a file in the bmp format;

[0028] Use the import data in the mat lab software to import the degree of polarization image saved in the bmp format into the mat lab software;

[0029] Use the double instruction to convert the imported image data format into a double-precision data type;

[0030] Use the mesh instruction to generate a three-dimensional image from the image data converted into a double-precision data type, and obtain the relative light intensity value of the degree of polarization image;

[0031] Based on the relative light intensity value, statistically obtain the relative light intensity distribution value of the degree of polarization image.

[0032] Optionally, in step (2), the relative light intensity distribution value input into the neural network includes: the maximum value, the minimum value, and the mean value of the relative light intensity distribution value.

[0033] Optionally, in step (2), the neural network is a BP neural network.

[0034] As can be seen from the above technical solutions, compared with the prior art, the present invention proposes a method for classifying airborne disease spores of greenhouse crops based on microscopic polarization image features. Aiming at the fact that the airborne disease spore image is easily interfered by ambient light during acquisition, and different types of airborne disease spores have their unique color characteristics, the relative light intensity distribution value of the airborne disease spore polarization degree image is selected as the first type of feature for classifying airborne disease spores; aiming at the fact that different types of airborne disease spores have unique texture characteristics on their surfaces, the texture feature of the airborne disease spore polarization angle image is selected as the second type of feature for classifying airborne disease spores; by fusing these two types of features, a classification model of airborne disease spores based on a neural network is trained, effectively reducing the interference of external ambient light, improving the recognition rate of disease spores with similar morphologies, effectively improving the classification and recognition accuracy of airborne disease spores, and having important significance and practical value for the field of facility environment disease detection and environmental control. Description of the Drawings

[0035] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained according to the provided drawings without creative efforts.

[0036] Figure 1 It is a schematic flow chart of the method of the present invention.

[0037] Figure 2(a) is a schematic diagram of the microscopic polarization image of the Botrytis cinerea spores of the present invention under 0° polarization.

[0038] Figure 2(b) is a schematic diagram of the microscopic polarization image of the Botrytis cinerea spores of the present invention under 45° polarization.

[0039] Figure 2(c) is a schematic diagram of the microscopic polarization image of the Botrytis cinerea spores of the present invention under 90° polarization.

[0040] Figure 2(d) is a schematic diagram of the microscopic polarization image of the Botrytis cinerea spores of the present invention under 135° polarization.

[0041] Figure 3(a) is a schematic diagram of the microscopic polarization image of the Podosphaera xanthii spores of the present invention under 0° polarization.

[0042] Figure 3(b) is a schematic diagram of the microscopic polarization image of the Podosphaera xanthii spores of the present invention under 45° polarization.

[0043] Figure 3(c) is a schematic diagram of the microscopic polarization image of the Podosphaera xanthii spores of the present invention under 90° polarization.

[0044] Figure 3(d) is a schematic diagram of the microscopic polarization image of the Podosphaera xanthii spores of the present invention under 135° polarization.

[0045] Figure 4(a) is a schematic diagram of the microscopic polarization image of the Pseudoperonospora cubensis spores of the present invention under 0° polarization.

[0046] Figure 4(b) is a schematic diagram of the microscopic polarization image of the Pseudoperonospora cubensis spores of the present invention under 45° polarization.

[0047] Figure 4(c) is a schematic diagram of the microscopic polarization image of the Pseudoperonospora cubensis spores of the present invention under 90° polarization.

[0048] Figure 4(d) is a schematic diagram of the microscopic polarization image of the Pseudoperonospora cubensis spores of the present invention under 135° polarization.

[0049] Figure 5(a) is a schematic diagram of the preprocessing result of the Botrytis cinerea spores of the present invention under S0.

[0050] Figure 5(b) is a schematic diagram of the preprocessing result of the Botrytis cinerea spores of the present invention under S1.

[0051] Figure 5(c) is a schematic diagram of the preprocessing result of the Botrytis cinerea spores of the present invention under S2.

[0052] Figure 5(d) is a schematic diagram of the preprocessing result of the Botrytis cinerea spores of the present invention under S3.

[0053] Figure 6(a) is a schematic diagram of the preprocessing result of the Podosphaera xanthii spores of the present invention under S0.

[0054] Figure 6(b) is a schematic diagram of the pretreatment result of cucumber powdery mildew spores of the present invention under S1.

[0055] Figure 6(c) is a schematic diagram of the pretreatment result of cucumber powdery mildew spores of the present invention under S2.

[0056] Figure 6(d) is a schematic diagram of the pretreatment result of cucumber powdery mildew spores of the present invention under S3.

[0057] Figure 7(a) is a schematic diagram of the pretreatment result of cucumber downy mildew spores of the present invention under S0.

[0058] Figure 7(b) is a schematic diagram of the pretreatment result of cucumber downy mildew spores of the present invention under S1.

[0059] Figure 7(c) is a schematic diagram of the pretreatment result of cucumber downy mildew spores of the present invention under S2.

[0060] Figure 7(d) is a schematic diagram of the pretreatment result of cucumber downy mildew spores of the present invention under S3.

[0061] Figure 8(a) is a schematic diagram of the degree of polarization image DOP of tomato gray mold spores of the present invention.

[0062] Figure 8(b) is a schematic diagram of the angle of polarization image AOP of tomato gray mold spores of the present invention.

[0063] Figure 9(a) is a schematic diagram of the degree of polarization image DOP of cucumber powdery mildew spores of the present invention.

[0064] Figure 9(b) is a schematic diagram of the angle of polarization image AOP of cucumber powdery mildew spores of the present invention.

[0065] Figure 10(a) is a schematic diagram of the angle of polarization image AOP of cucumber downy mildew spores of the present invention.

[0066] Figure 10(b) is a schematic diagram of the angle of polarization image AOP of cucumber downy mildew spores of the present invention.

[0067] Figure 11 It is a schematic diagram of the influence of the Gabor filtering direction on the contrast and entropy of the angle of polarization image of airborne disease spores of the present invention.

[0068] Figure 12(a) is a schematic diagram of the angle of polarization image of airborne disease spores before Gabor filtering treatment of the present invention.

[0069] Figure 12(b) is a schematic diagram of the angle of polarization image of airborne disease spores after Gabor filtering treatment of the present invention.

[0070] Figure 13(a) is a schematic diagram of the relative light intensity value of the degree of polarization image of tomato gray mold spores of the present invention.

[0071] Figure 13(b) is a schematic diagram of the relative light intensity value of the polarization degree image of cucumber powdery mildew spores of the present invention.

[0072] Figure 13(c) is a schematic diagram of the relative light intensity value of the polarization degree image of cucumber downy mildew spores of the present invention.

[0073] Figure 14(a) is a schematic diagram of the root mean square error at different iteration times under the operation of the BP neural network of the present invention.

[0074] Figure 14(b) is a schematic diagram of the network operation state under the operation of the BP neural network of the present invention.

[0075] Figure 14(c) is a schematic diagram of the regression result after the network training under the operation of the BP neural network of the present invention.

[0076] Figure 15 It is a schematic diagram of the classification result of airborne disease spores of the present invention. Specific embodiments

[0077] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0078] Embodiment 1:

[0079] Embodiment 1 of the present invention discloses a method for classifying airborne disease spores of greenhouse crops based on microscopic polarization image features, as Figure 1 shown, including:

[0080] Capture airborne disease spores spreading in the greenhouse through a spore capture device, and collect microscopic polarization images of the airborne disease spores using a microscopic polarization device, including: 200 microscopic polarization images of tomato gray mold spores, cucumber powdery mildew spores, and cucumber downy mildew spores at four polarization angles of 0°, 45°, 90°, and 135° respectively. The microscopic polarization images of tomato gray mold spores at four polarization angles of 0°, 45°, 90°, and 135° are shown in Figures 2(a), 2(b), 2(c), and 2(d) respectively; the microscopic polarization images of cucumber powdery mildew spores at four polarization angles of 0°, 45°, 90°, and 135° are shown in Figures 3(a), 3(b), 3(c), and 3(d) respectively; the microscopic polarization images of cucumber downy mildew spores at four polarization angles of 0°, 45°, 90°, and 135° are shown in Figures 4(a), 4(b), 4(c), and 4(d) respectively.

[0081] Step (1): Preprocess the microscopic polarization image of airborne disease spores using the Stokes vector method to obtain the polarization degree image and polarization angle image of the airborne disease spores;

[0082] Preprocess the microscopic polarization image of airborne disease spores using the Stokes vector method as follows:

[0083]

[0084] Among them, S0 is the total light field intensity of the incident light; S1 is the intensity difference of linear polarization at 0° and 90°; S2 is the intensity difference of linear polarization at 45° and 135°; S3 is the intensity difference of circular polarization between right-handed and left-handed. The preprocessing results of tomato gray mold spores under S0, S1, S2, and S3 are shown in Figures 5(a), 5(b), 5(c), and 5(d) respectively; the preprocessing results of cucumber powdery mildew spores under S0, S1, S2, and S3 are shown in Figures 6(a), 6(b), 6(c), and 6(d) respectively; the preprocessing results of cucumber downy mildew spores under S0, S1, S2, and S3 are shown in Figures 7(a), 7(b), 7(c), and 7(d) respectively.

[0085] The calculation formula of the polarization degree image is as follows:

[0086]

[0087] The calculation formula of the polarization angle image is as follows:

[0088]

[0089] Among them, the polarization degree image DOP and polarization angle image AOP of tomato gray mold spores are shown in Figures 8(a) and 8(b) respectively; the polarization degree image DOP and polarization angle image AOP of cucumber powdery mildew spores are shown in Figures 9(a) and 9(b) respectively; the polarization degree image DOP and polarization angle image AOP of cucumber downy mildew spores are shown in Figures 10(a) and 10(b) respectively.

[0090] Step (2): Use the Gabor transform (i.e., the windowed Fourier transform) to transform the polarization angle image from the spatial domain to the frequency domain, extract the texture features of the polarization angle image, and jointly use the relative light intensity distribution value of the polarization degree image as the input of the neural network to train and obtain an airborne disease spore classification model;

[0091] Use the Gabor transform to transform the polarization angle image from the spatial domain to the frequency domain as follows:

[0092]

[0093]

[0094]

[0095]

[0096] Among them, v is the wavelength of the Gabor filter; u is the direction of the Gabor kernel function; K is the total number of directions of the Gabor kernel function; σ / k is the size of the Gaussian window.

[0097] Texture features include the contrast and entropy of airborne disease spores.

[0098] The direction value of the Gabor filter is π / 15. Because at this time, the contrast and entropy of the polarization angle image of airborne disease spores are the largest, as Figure 11 shown. The light and dark degree of the polarization angle image of airborne disease spores contributes the most to its own texture features, and the direction of the filter also best matches the texture direction of the polarization angle image of airborne disease spores.

[0099] The polarization angle images of airborne disease spores before and after Gabor filtering are shown in Figures 12(a) and 12(b) respectively.

[0100] The method for obtaining the relative light intensity distribution value of the polarization degree image is specifically as follows:

[0101] Save the polarization degree image as a file in bmp format;

[0102] Use the import data in the mat lab software to import the polarization degree image saved in bmp format into the mat lab software;

[0103] Use the double instruction to convert the imported image data format into a double-precision data type;

[0104] Use the mesh instruction to generate a three-dimensional image from the image data converted into a double-precision data type to obtain the relative light intensity value of the polarization degree image; the relative light intensity values of the polarization degree images of tomato gray mold spores, cucumber powdery mildew spores, and cucumber downy mildew spores are shown in Figures 13(a), 13(b), and 13(c) respectively;

[0105] Based on the relative light intensity values, the relative light intensity distribution value of the polarization degree image is statistically obtained, as shown in Table 1.

[0106] Table 1 Relative light intensity distribution value of the polarization degree image

[0107]

[0108] The relative light intensity distribution values input into the neural network include: the maximum value, the minimum value, and the mean value of the relative light intensity distribution values. The neural network is a BP neural network. That is, three eigenvalue features of the maximum value, the minimum value, and the mean value of the relative light intensity distribution value of the airborne disease spore polarization degree image, and two texture features of the contrast and entropy of the airborne disease spore polarization angle image are combined with the BP neural network to realize the classification of multiple airborne disease spores. When performing classification and recognition, the number of network iterations is set to 300, the learning rate is set to 0.01, the set value of the root mean square error is 0.001, and the root mean square error (Mean Squared Error) is used to evaluate the network performance. The root mean square error, the network running state, and the regression result after network training at different iteration times under the operation of the BP neural network are shown in Figures 14(a), 14(b), and 14(c) respectively.

[0109] Step (3): Input the image to be measured into the airborne disease spore classification model to obtain the classification result, as Figure 15 shown; the classification result statistics are shown in Table 2.

[0110] Table 2 Airborne disease spore classification results

[0111] Spore species Sample quantity Number of correctly predicted samples Recognition rate (%) Spores of cucumber downy mildew 60 45 75 Spores of tomato gray mold 60 53 83.33 Spores of cucumber powdery mildew 60 58 96.67

[0112] The embodiment of the present invention discloses a method for classifying airborne disease spores of greenhouse crops based on microscopic polarization image features. Aiming at the fact that the airborne disease spore images are easily interfered by ambient light during acquisition, and different types of airborne disease spores have their unique color characteristics, the relative light intensity distribution value of the airborne disease spore polarization degree image is selected as the first type of feature for classifying airborne disease spores; aiming at the fact that different types of airborne disease spores have unique texture features on their surfaces, the texture features of the airborne disease spore polarization angle image are selected as the second type of feature for classifying airborne disease spores; by fusing these two types of features, an airborne disease spore classification model based on a neural network is trained, effectively reducing the interference of external ambient light, improving the recognition rate of disease spores with similar morphologies, effectively improving the classification and recognition accuracy of airborne disease spores, and having important significance and practical value for the field of facility environment disease detection and environmental control.

[0113] In this specification, each embodiment is described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. The same or similar parts among the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the description of the method part.

[0114] The foregoing description of the disclosed embodiments enables those skilled in the art to practice or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for classifying airborne disease spores of greenhouse crops based on microscopic polarization image features, characterized in that, Including: Step (1): Preprocess the microscopic polarization image of airborne disease spores by using the Stokes vector method to obtain the polarization degree image and polarization angle image of airborne disease spores; Step (2): Use Gabor transform to convert the polarization angle image from the spatial domain to the frequency domain, extract the texture features of the polarization angle image, and jointly use the relative light intensity distribution value of the polarization degree image as the input of the neural network to train an airborne disease spore classification model; Step (3): Input the image to be measured into the airborne disease spore classification model to obtain a classification result; In step (1), preprocess the microscopic polarization image of airborne disease spores by using the Stokes vector method as follows: Wherein, S0 is the total light field intensity of the incident light; S1 is the intensity difference of linear polarization at 0° and 90°; S2 is the intensity difference of linear polarization at 45° and 135°; S3 is the intensity difference of circular polarization at right-handed and left-handed; In step (1), the calculation formula of the polarization degree image is as follows: In step (1), the calculation formula of the polarization angle image is as follows: In step (2), use Gabor transform to convert the polarization angle image from the spatial domain to the frequency domain as follows: Wherein, v is the wavelength of the Gabor filter; u is the direction of the Gabor kernel function; K is the total number of directions of the Gabor kernel function; σ / k is the size of the Gaussian window; In step (2), the relative light intensity distribution values input into the neural network include: the maximum value, minimum value and mean value of the relative light intensity distribution value.

2. The method for classifying airborne disease spores of greenhouse crops based on microscopic polarization image features according to claim 1, wherein In step (2), the texture features include: the contrast and entropy of airborne disease spores.

3. The method for classifying airborne disease spores of greenhouse crops based on microscopic polarization image features according to claim 1, wherein, In step (2), the method for obtaining the relative light intensity distribution value of the polarization degree image is specifically: Save the polarization degree image as a file in bmp format; Use importdata in matlab software to import the polarization degree image saved in bmp format into matlab software; Use the double instruction to convert the imported image data format into a double-precision data type; Use the mesh instruction to generate a three-dimensional image from the image data converted into a double-precision data type to obtain the relative light intensity value of the polarization degree image; Based on the relative light intensity value, statistically obtain the relative light intensity distribution value of the polarization degree image.

4. The method for classifying airborne disease spores of greenhouse crops based on microscopic polarization image features according to claim 1, wherein In step (2), the neural network is a BP neural network.

Images