Nondestructive detection method for degradation of Pleurotus geesteranus strains based on microscopic hyperspectral images

Abstract

The present invention proposes a nondestructive detection method for degradation of Pleurotus geesteranus strains based on microscopic hyperspectral images, comprising: step S1: selecting Pleurotus geesteranus strains for activation culture; step S2: analyzing the degradation of the selected Pleurotus geesteranus strains and defining the strain degradation evaluation level; step S3: collecting microscopic hyperspectral images of degraded and non-degraded Pleurotus geesteranus strains respectively; step S4: extracting the average spectral transmittance of the region of interest from the microscopic hyperspectral images; step S5: extracting the features of interest as learning features through a convolutional neural network, establishing a rapid detection model for strain degradation, and completing classification. The present invention establishes a model based on a one-dimensional convolutional neural network algorithm, effectively learns corresponding features from high-dimensional data, avoids complex feature extraction processes, reduces network calculation time, improves the accuracy of model classification, improves the detection efficiency of Pleurotus geesteranus strain degradation, and realizes rapid nondestructive detection of Pleurotus geesteranus strain degradation.

Description

Technical Field

[0001] The invention relates to the technical field of edible fungus information breeding, and in particular to a nondestructive detection method for degradation of Pleurotus geesteranus strains based on microscopic hyperspectral images. Background Art

[0002] Edible fungi, i.e. large edible fungi, are rich in nutrients and trace elements. They are delicious and have a high medical and health care effect, so they are deeply loved by consumers at home and abroad. my country is the largest producer of edible fungi, and its output accounts for about 70% of the world's total output of edible fungi. In recent years, with the continuous advancement of cultivation technology, the continuous increase in varieties, and the increasing maturity of factory cultivation, the edible fungi industry has become the fifth largest planting industry in my country's agriculture, playing an important role in my country's economic construction.

[0003] The seeds of edible fungi are called edible fungi strains. The quality of strains directly affects the results of cultivation, that is, the level of yield, and is one of the key factors in generating economic benefits. The main varieties of Pleurotus geesteranus cultivated in my country are very close in terms of kinship, and strains are prone to degeneration and aging after years of expansion and propagation. Strain degeneration refers to the phenomenon that the economic traits of edible fungi strains deteriorate, resulting in changes in edible fungi yield, quality, resistance, etc. that deviate from human needs. Strain degeneration is a serious problem common in many edible fungi cultivation. In actual production, some excellent edible fungi varieties often degenerate within a few generations. Once degraded strains are used for cultivation, they often lead to late fruiting, low yield, poor quality, etc., which bring significant economic losses to edible fungi producers. Therefore, it has always been a problem that edible fungi production needs to face and solve to quickly identify degraded strains and effectively eliminate the hidden dangers brought by degraded strains to edible fungi production, and prevent losses caused by degraded strains flowing into the production process.

[0004] Microscopic hyperspectral imaging technology is a new spectral technology that integrates microscopic optical imaging technology and spectral analysis technology, and has been widely used in medical diagnosis, food safety testing and other fields. Due to the many advantages of hyperspectral imaging technology, such as non-contact, non-destructive, fast and accurate, and the fact that microscopic technology can observe the morphological structure and characteristics of tiny objects that cannot be distinguished by the naked eye, the combination of hyperspectral imaging technology and microscopic technology makes information collection faster, the exploration level deeper, and the captured information more persistent and stable, which has good development prospects in the field of bioinformatics breeding technology. Summary of the invention

[0005] Considering that the current research on the degeneration of edible fungi strains mainly uses chemical or biological technology to detect and analyze the changes in physiological information of fungi before and after degeneration, traditional detection methods are inefficient and time-consuming, and the entire process is complicated and cumbersome. There is still a lack of efficient and rapid means of detecting the degeneration of edible fungi strains, which cannot meet the production needs well.

[0006] In order to overcome the shortcomings of the prior art, the present invention takes Pleurotus geesteranus strains that are prone to degeneration during the passage process as the object, and proposes a non-destructive detection method for the degeneration of Pleurotus geesteranus strains based on microscopic hyperspectral images.

[0007] The solution of the present invention is based on the fact that the internal physiological information of Pleurotus geesteranus strains is an important indicator for determining whether the strains are degenerated, and the spectral image of the strains can effectively reflect the internal physiological information of the strains without causing damage to the strains. Therefore, the detection model is established using the microscopic hyperspectral image information of the strains, which has a fast detection speed and high accuracy, and can quickly and non-destructively identify the degeneration of Pleurotus geesteranus strains.

[0008] The scheme includes: step S1: selecting Pleurotus geesteranus strains for activation culture; step S2: analyzing the degradation of the selected Pleurotus geesteranus strains and defining the strain degradation evaluation level; step S3: collecting microscopic hyperspectral images of degraded and non-degraded Pleurotus geesteranus strains respectively; step S4: extracting the average spectral transmittance of the region of interest from the microscopic hyperspectral image; step S5: extracting the features of interest as learning features through a convolutional neural network, establishing a rapid detection model for strain degradation, and completing classification. The present invention establishes a model based on a one-dimensional convolutional neural network algorithm, effectively learns corresponding features from high-dimensional data, avoids a complex feature extraction process, reduces network calculation time, improves the accuracy of model classification, improves the detection efficiency of Pleurotus geesteranus strain degradation, and realizes rapid non-destructive detection of Pleurotus geesteranus strain degradation.

[0009] The present invention specifically adopts the following technical solutions:

[0010] A nondestructive detection method for degradation of Pleurotus geesteranus strains based on microscopic hyperspectral images, characterized in that it comprises the following steps: step S1: selecting Pleurotus geesteranus strains for activation culture; step S2: analyzing the degradation of the selected Pleurotus geesteranus strains and defining the strain degradation evaluation level; step S3: collecting microscopic hyperspectral images of degraded and non-degraded Pleurotus geesteranus strains respectively; step S4: extracting the average spectral transmittance of an area of ​​interest from the microscopic hyperspectral image; step S5: extracting features of interest as learning features through a convolutional neural network, establishing a rapid detection model for strain degradation, and completing classification.

[0011] Further, in step S1, the selected Pleurotus geesteranus strains are inoculated onto a solid culture medium and cultured at 25° C. for 10 days to perform primary strain activation culture;

[0012] In step S2, the activated Pleurotus geesteranus strains are transferred to a liquid potato dextrose agar medium and cultured in a dark and shaking manner for 5 days to obtain a secondary liquid strain of Pleurotus geesteranus, and the Pleurotus geesteranus spore liquid is filtered through 4 layers of filter paper to obtain the Pleurotus geesteranus spore liquid and transferred to a colorimetric medium, and the decolorization rate of the Pleurotus geesteranus strains in the colorimetric medium is measured to define the Pleurotus geesteranus strain degradation evaluation grade;

[0013] In step S3, expansion culture is carried out according to the degradation level of Pleurotus geesteranus strains divided in step S2, microscopic hyperspectral images of degraded and non-degraded Pleurotus geesteranus strains are collected, and the microscope magnification, wavelength range, exposure time, resolution and acquisition speed are set.

[0014] Furthermore, step S4 processes the collected microscopic hyperspectral images of degraded and non-degraded Pleurotus geesteranus strains, creates a region of interest, extracts the average spectral transmittance, and completes the establishment of a data set;

[0015] In step S5, the data set established in step S4 is used as input data, divided into a training set and a test set, a strain degradation detection model is established through a convolutional neural network algorithm, the accuracy of the detection model in detecting whether the strain is degraded or not is analyzed, and the strain degradation detection model is evaluated.

[0016] Furthermore, in step S2, the method for distinguishing the undegraded and degraded strains is as follows: the decolorization condition is set as T = OD of the culture supernatant 615 / OD of control solution supernatant 615 When T>0.7, the Pleurotus geesteranus strain is judged as a degenerated strain; otherwise, it is a non-degenerated strain.

[0017] Further, in step S3, a hyperspectral imager is set just above the microscope to collect microscopic hyperspectral images of degraded and non-degraded Pleurotus geesteranus strains; the collection method is transmission mode, the microscope eyepiece magnification is 10 times, the objective lens magnification is 10 times, the light source is a 20W halogen lamp, the exposure time is set to 15ms, the collection speed is 0.06cm / s, and the wavelength range is 400-1046nm; at the same time, the microscopic hyperspectral image data of the reference plate and the dark current are obtained for the correction of the original spectral image before data processing to reduce the influence of uneven illumination and dark current on the spectral image.

[0018] Furthermore, the region of interest created in step S4 is the spectral value within the image range of any mycelium part with a pixel size of 50×15 pixels, and finally the data set is established as a 361*200 matrix, which contains 200 spectral information, and each spectral information consists of 360 bands.

[0019] Furthermore, the rapid detection model for strain degradation in step S5 adopts a classification algorithm based on a one-dimensional convolutional neural network. The model includes 6 convolutional layers, in which the filter dimension is 3, and the activation function uses a hyperbolic tangent function to improve the expressiveness of the model; a pooling layer is added after every two convolutional layers to reduce the amount of calculation of the model and improve the calculation speed of the model; and a Flatten layer is added between the convolutional neural network and the fully connected layer to make the multidimensional data one-dimensional, and a cross entropy loss function is set as the loss function of the model training to describe the difference between the model feature output and the expected output.

[0020] Further, in step S3, the microscopic hyperspectral image data of the blank PDA culture medium and the dark current are obtained, according to the correction formula:

[0021]

[0022] The original spectral image is corrected before data processing to reduce the influence of uneven illumination and dark current on the spectral image; R is the corrected spectral image, I dark is the dark current image, I raw is the original spectral image, I ref This is the spectrum image of blank PDA culture medium.

[0023] Furthermore, in step S5, the expression of the activation function used is as follows:

[0024]

[0025] in For W k The value at a certain position (i, j), W k is the kth filter, x is the value at a certain position (i, j), b k is a constant;

[0026] The loss function is as follows:

[0027]

[0028] Where L is the cross entropy loss function value, is the probability of label being 1, is the probability that the label is 0. When the predicted output is closer to the true sample label 1, the loss function L is smaller; the closer the predicted output is to 0, the larger L is.

[0029] Compared with the prior art, the present invention and its preferred embodiments have the following beneficial effects:

[0030] The spectral image of Pleurotus geesteranus strains is collected by a microscopic hyperspectral image acquisition device, so as to quickly and effectively obtain the spectral characteristics of the internal physiological information of Pleurotus geesteranus strains;

[0031] In order to obtain the effective spectral information of Pleurotus geesteranus strains in a short time, the degradation of Pleurotus geesteranus strains can be quickly and accurately identified through model analysis, thereby improving the efficiency and accuracy of Pleurotus geesteranus strain degradation detection.

[0032] It combines the advantages of microscopic hyperspectral imaging technology, such as faster information collection, deeper exploration, and more persistent and stable information capture. It uses a convolutional neural network algorithm to classify whether the Pleurotus geesteranus strains are degraded, thereby improving the speed and accuracy of detection. This method uses a microscopic hyperspectral imaging system, optical equipment, and sensors to digitize data from manual observations and measurements, providing new ideas and methods for early identification of degraded edible mushroom strains. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] The present invention is further described in detail below with reference to the accompanying drawings and specific embodiments:

[0034] Figure 1 Schematic diagram of the structure of the test device used in the embodiment of the present invention, in which:

[0035] 1- portable hyperspectral imager; 2- eyepiece; 3- objective lens; 4- Pleurotus geesteranus strain sample; 5- stage; 6- light source; 7- microscope; 8- computer;

[0036] Figure 2 is a specific operation flow chart of an embodiment of the present invention;

[0037] Figure 3 is a graph showing average spectral transmittance of degraded and non-degraded Pleurotus geesteranus strains in an embodiment of the present invention;

[0038] Figure 4 This is a diagram of classification detection results of a training set of Pleurotus geesteranus strain samples according to an embodiment of the present invention;

[0039] Figure 5 This is a diagram of classification detection results of a test set of Pleurotus geesteranus strain samples according to an embodiment of the present invention. DETAILED DESCRIPTION

[0040] In order to make the features and advantages of this patent more obvious and easy to understand, the following embodiments are specifically described in detail with the accompanying drawings as follows:

[0041] like Figure 2 As shown, the main design points of the present invention are centered around the following steps:

[0042] Step S1: Select bacteria for activation culture (A)

[0043] The selected Pleurotus geesteranus strains were inoculated onto a solid culture medium and cultured at 25°C for 10 days to perform primary strain activation culture;

[0044] Step S2: Determine the bacterial species degradation evaluation level (B)

[0045] The activated Pleurotus geesteranus strains were transferred to a liquid potato dextrose agar medium (liquid PDA) and cultured in a dark and shaking manner for 5 days to obtain a secondary liquid strain of Pleurotus geesteranus, and the Pleurotus geesteranus spore liquid was obtained by filtering through 4 layers of filter paper and transferred to a colorimetric medium, and the decolorization rate of the Pleurotus geesteranus strains in the colorimetric medium was determined to define the degeneration evaluation grade of the Pleurotus geesteranus strains;

[0046] Step S3: Collecting microscopic hyperspectral images of bacterial samples (C)

[0047] Carry out expansion culture according to the degradation level of Pleurotus geesteranus strains divided in step S2, collect microscopic hyperspectral images of degraded and non-degraded Pleurotus geesteranus strains, and set the microscope magnification, wavelength range, exposure time, resolution and acquisition speed;

[0048] Step S4: Extract the average spectral transmittance (D) of the region of interest

[0049] The collected microscopic hyperspectral images of degraded and non-degraded Pleurotus geesteranus strains were processed to create regions of interest, extract the average spectral transmittance, and complete the establishment of the data set;

[0050] Step S5: Establish a rapid detection model for the degradation of Pleurotus geesteranus strains and complete classification (E)

[0051] The data set established in step S4 is used as input data and divided into a training set and a test set. A strain degradation detection model is established through a convolutional neural network algorithm. The accuracy of the detection model in detecting whether the strain is degraded is analyzed, and the strain degradation detection model is evaluated.

[0052] The specific design details of the embodiments of the present invention are described below:

[0053] 1. Device:

[0054] like Figure 1 The experimental device used in this scheme includes: a microscope 7 and a sample 4 of the Pleurotus geesteranus strain of the test object.

[0055] A portable hyperspectral imager 1, an eyepiece 2, an objective lens 3, a stage 5, a light source 6 and a computer 8 are provided;

[0056] The specific connection method is:

[0057] The microscope 7 is provided with an eyepiece 2, an objective lens 3, a stage 5 and a light source 6; wherein the objective lens 3 is placed above the stage 5, and a Pleurotus geesteranus strain sample 4 is placed on the stage 5; and the light source 6 is directly below the stage 5;

[0058] A portable hyperspectral imager 1 is arranged directly above the microscope 7;

[0059] The portable hyperspectral imager 1 is externally connected to a computer 8;

[0060] The above-mentioned functional components are all universal parts.

[0061] Its working mechanism is: turn on the computer 8 and the portable hyperspectral imager 1, connect the light source 6 of the microscope 7, set the exposure time, resolution and camera acquisition speed on the software provided by the portable hyperspectral imager 1, adjust the position of the stage 6, and obtain a clear image of the Pleurotus geesteranus strain sample 4; click the software start acquisition button to complete the acquisition of the microscopic hyperspectral image of the Pleurotus geesteranus strain sample.

[0062] 2. Methods:

[0063] 1. Culture medium:

[0064] Step A1: Prepare solid culture medium: 200 g potatoes, 20 g glucose, 10 g peptone, 18 g agar, add water to 1 L.

[0065] Step A2: prepare liquid potato dextrose agar medium (liquid PDA): 200 g potatoes, 20 g glucose, 2 g potassium dihydrogen phosphate, 1 g magnesium sulfate, 20 mg vitamin B1, and add water to 1 L.

[0066] Step A3: Prepare colorimetric culture medium: 4.5 g yeast powder, 7.5 g bacterial peptone, 0.025 g bromothymol blue, 5 g lactose, and add water to 1 L.

[0067] The above culture medium is sterilized at 121°C for 30 minutes and is used after cooling. Vitamin B will be inactivated at high temperatures and is filtered and added during the test operation.

[0068] 2. Evaluation of Pleurotus geesteranus degeneration:

[0069] Step B1: inoculate the Pleurotus geesteranus strains onto a solid culture medium and culture them at 25° C. for 10 days to perform primary strain activation culture.

[0070] Step B2: The activated Pleurotus geesteranus strains were transferred to a liquid PDA culture medium, two 5 mm diameter bacterial blocks were inoculated into each bottle of liquid PDA culture medium, and the culture was shaken at 25° C. and 160 r / min in the dark for 5 days, and secondary liquid culture was performed to obtain secondary liquid strains of Pleurotus geesteranus. The Pleurotus geesteranus spore liquid was filtered through 4 layers of filter paper and transferred to a colorimetric culture medium, and shaken and cultured at 25° C. and 120 r / min for 10 days. The same level was repeated 3 times, and the control group was a blank PDA culture solution. The culture solution and the control solution were centrifuged respectively, and the centrifugation was performed at 12000 r / min for 5 min. The supernatant was taken to measure the OD615nm value.

[0071] Step B3: Calculate the decolorization status based on the measured absorbance T = OD of the culture supernatant 615 / OD of control solution supernatant 615 , and it is set that when T>0.7, the Pleurotus geesteranus strain is judged as a degenerated strain.

[0072] 3. Data collection:

[0073] According to the degradation level of Pleurotus geesteranus strains divided in step S2, the culture was expanded, and the microscopic hyperspectral images of degraded and non-degraded Pleurotus geesteranus strains were collected using a microscopic hyperspectral image acquisition device, wherein the eyepiece magnification was 10 times, the objective lens magnification was 10 times, the light source was a 20W halogen lamp, the exposure time was set to 15ms, the acquisition speed was 0.06cm / s, and the microscopic hyperspectral image data of the blank PDA culture medium and the dark current were obtained at the same time. According to the correction formula:

[0074]

[0075] Where R is the corrected spectral image, I dark is the dark current image, I raw is the original spectral image, I ref This is the spectral image of the blank PDA culture medium. The original spectral image was corrected before data processing to reduce the impact of uneven illumination and dark current on the spectral image.

[0076] 4. Data extraction:

[0077] Import the corrected microscopic hyperspectral image into ENVI, create a region of interest as any part of the hyphae, extract the average spectral transmittance of the region of interest and save it, and the data obtained are the average spectral transmittance at a wavelength of 400nm-1046nm.

[0078] Figure 3 The extracted average transmittance spectral curves of degraded and non-degraded bacterial species.

[0079] Import the extracted and saved average spectral transmittance into an Excel table, set each row to represent the average spectral reflectance information of a bacterial species, each column to represent the spectral reflectance at different wavelengths, and the last column to be the true degradation category of the bacterial species to complete the establishment of the data set.

[0080] 5. Model building

[0081] In the MATLAB environment, the data set created in step S4 is used as input data, X is set to be the feature information of the data, and Y represents the true category, where X increases one dimension when reading the matrix, and is divided into a training set and a validation set in a ratio of 7:3.

[0082] In the MATLAB environment, the one-dimensional convolutional neural network model extracts interesting features from complex data sets through the convolution operation of the convolution layer, and uses the hyperbolic tangent function as the activation function to improve the expressiveness of the model. The expression of the activation function used is as follows:

[0083]

[0084] in For W k The value at a certain position (i, j), W k is the kth filter, x is the value at a certain position (i, j), b k is a constant.

[0085] At the same time, in order to ensure that the main features are still retained and reduce the amount of calculation, a maximum pooling layer is added after every two convolution layers. Taking the input data size of 360*1 as an example, after 16 filters with a dimension of 3 and two convolution operations, the output data size is 16@356*1. After removing redundant information through pooling, the output data size is 16@89*1.

[0086] A flatten layer is added between the convolutional layer and the fully connected layer to convert the multidimensional data into one dimension.

[0087] The cross entropy loss function is used as the loss function for model training to describe the difference between the feature output of the training model and the actual calibration output. The smaller the loss function, the smaller the difference between the two, and the higher the model accuracy. The loss function is as follows:

[0088]

[0089] Where L is the cross entropy loss function value, is the probability of label being 1, is the probability that the label is 0. When the predicted output is closer to the true sample label 1, the loss function L is smaller; the closer the predicted output is to 0, the larger L is.

[0090] 3. Test results:

[0091] The divided test set is input into the above-mentioned classification model based on one-dimensional convolutional neural network. The convolution layer in the model completes the feature extraction of the average spectral information of Pleurotus geesteranus species, the pooling layer completes the dimensionality reduction of the average spectral information, and finally the fully connected layer completes the classification and recognition of Pleurotus geesteranus species samples, and the classification effect is intuitively displayed through the confusion matrix.

[0092] Figure 4 , 5 These are the classification detection results of the training set and test set of Pleurotus geesteranus samples, respectively.

[0093] As can be seen from the figure, after 800 iterations of training, the average accuracy of the training set of the Pleurotus geesteranus strain degradation detection model is 95.7%, and the average accuracy of the test set is 93.3%. The constructed detection model has obtained good detection results.

[0094] This patent is not limited to the above-mentioned optimal implementation mode. Anyone can derive other forms of non-destructive detection methods for degradation of Pleurotus geesteranus based on microscopic hyperspectral images under the inspiration of this patent. All equivalent changes and modifications made according to the scope of the patent application of the present invention shall fall within the scope of coverage of this patent.

Claims

1. A nondestructive detection method for degradation of Pleurotus geesteranus strains based on microscopic hyperspectral images, characterized in that: The following steps are included: Step S1: Selecting Pleurotus geesteranus strains for activation culture; Step S2: Analyzing the degradation of the selected Pleurotus geesteranus strains and defining the strain degradation evaluation level; Step S3: Collecting microscopic hyperspectral images of degraded and non-degraded Pleurotus geesteranus strains respectively; Step S4: Extracting the average spectral transmittance of the region of interest from the microscopic hyperspectral image; Step S5: Extracting the features of interest as learning features through a convolutional neural network, establishing a rapid detection model for strain degradation, and completing classification; In step S1, the selected Pleurotus geesteranus strains are inoculated onto a solid culture medium and cultured at 25° C. for 10 days to perform primary strain activation culture; In step S2, the activated Pleurotus geesteranus strains are transferred to a liquid potato dextrose agar medium and cultured in a dark and shaking manner for 5 days to obtain a secondary liquid strain of Pleurotus geesteranus, and the Pleurotus geesteranus spore liquid is filtered through 4 layers of filter paper to obtain the Pleurotus geesteranus spore liquid and transferred to a colorimetric medium, and the decolorization rate of the Pleurotus geesteranus strains in the colorimetric medium is measured to define the Pleurotus geesteranus strain degradation evaluation grade; In step S2, the method for distinguishing the undegraded and degraded strains is as follows: the decolorization condition is set as T = OD of the culture supernatant 615 / OD of control solution supernatant 615 ,When T>0.7, the Pleurotus geesteranus strain is judged as a degenerated strain; otherwise, it is a non-degenerated strain; In step S3, expansion culture is carried out according to the degradation level of Pleurotus geesteranus strains divided in step S2, microscopic hyperspectral images of degraded and non-degraded Pleurotus geesteranus strains are collected, and the microscope magnification, wavelength range, exposure time, resolution and acquisition speed are set; Step S4 processes the collected microscopic hyperspectral images of degraded and non-degraded Pleurotus geesteranus strains, creates a region of interest, extracts the average spectral transmittance, and completes the establishment of a data set; the region of interest created in step S4 is any mycelium part in the image, and the pixel size is 50×15 pixels within the spectral value of the image range; Step S5 uses the data set established in step S4 as input data, divides it into a training set and a test set, establishes a strain degradation detection model through a convolutional neural network algorithm, analyzes the accuracy of the detection model in detecting whether the strain is degraded, and evaluates the strain degradation detection model; The rapid detection model for strain degradation described in step S5 adopts a classification algorithm based on a one-dimensional convolutional neural network. The model includes 6 convolutional layers, in which the filter dimension is 3, and the activation function uses a hyperbolic tangent function to improve the expressiveness of the model; a pooling layer is added after every two convolutional layers to reduce the amount of calculation of the model and improve the calculation speed of the model; and a Flatten layer is added between the convolutional neural network and the fully connected layer to make the multidimensional data one-dimensional, and a cross entropy loss function is set as the loss function of the model training to describe the difference between the model feature output and the expected output.

2. The nondestructive detection method for degradation of Pleurotus geesteranus strains based on microscopic hyperspectral images according to claim 1 is characterized in that: In step S3, a hyperspectral imager is arranged directly above the microscope to collect microscopic hyperspectral images of degraded and non-degraded Pleurotus geesteranus strains; The acquisition method is transmission, the microscope eyepiece magnification is 10 times, the objective lens magnification is 10 times, the light source is a 20W halogen lamp, the exposure time is set to 15ms, the acquisition speed is 0.06cm / s, and the wavelength range is 400-1046nm; at the same time, the microscopic hyperspectral image data of the reference plate and the dark current are obtained for the correction of the original spectral image before data processing to reduce the influence of uneven illumination and dark current on the spectral image.

3. The nondestructive detection method for degradation of Pleurotus geesteranus strains based on microscopic hyperspectral images according to claim 2 is characterized in that: In step S3, the microscopic hyperspectral image data of the blank PDA culture medium and the dark current are obtained, and the correction formula is: The original spectral image is corrected before data processing to reduce the influence of uneven illumination and dark current on the spectral image; R is the corrected spectral image, I dark is the dark current image, I raw is the original spectral image, I ref This is the spectrum image of blank PDA culture medium.

4. The nondestructive detection method for degradation of Pleurotus geesteranus strains based on microscopic hyperspectral images according to claim 1 is characterized in that: In step S5, the expression of the activation function used is as follows: in For W k The value at a certain position (i, j), W k is the kth filter, x is the value at a certain position (i, j), b k is a constant; The loss function is as follows: Where L is the cross entropy loss function value, is the probability of label being 1, is the probability that the label is 0. When the predicted output is closer to the true sample label 1, the loss function L is smaller; the closer the predicted output is to 0, the larger L is.

Images