Classification method and system for mature and diseased spots of flue-cured tobacco based on texture and shape features

Abstract

The present invention discloses a method and system for classifying mature and diseased spots in flue-cured tobacco based on texture and shape features. The method comprises: obtaining a currently collected flue-cured tobacco image; inputting the currently collected flue-cured tobacco image into a trained flue-cured tobacco mature and diseased spot classification model to classify mature and diseased tobacco leaves in the currently collected flue-cured tobacco image; the diseased tobacco leaves include brown spot disease leaves and frog eye disease leaves; the trained flue-cured tobacco mature and diseased spot classification model is determined based on an Adaboost classifier, a circular LBP texture feature extraction module, a shape feature extraction module, and a sample data set; the sample data set includes multiple flue-cured tobacco sample images and corresponding annotation information; the annotation information includes information characterizing the flue-cured tobacco state. The present invention has fast classification speed and high accuracy, and can effectively increase the proportion of medium and high-quality tobacco leaves during machine grading after flue-cured tobacco leaves, thereby increasing economic benefits.

Description

Technical Field

[0001] The present invention relates to the technical field of tobacco production quality inspection, and in particular to a method and system for classifying mature spots and diseased spots of flue-cured tobacco based on texture and shape features. Background Art

[0002] As a pillar of the national economy, the tobacco industry has made significant contributions to its economic development. However, tobacco is susceptible to numerous diseases and pests, each requiring different purchasing conditions. Therefore, to improve tobacco quality and promote national economic development, it is essential to classify tobacco leaf pests and diseases at the time of purchase.

[0003] Identifying tobacco pests and diseases is the first step in tobacco leaf procurement and a crucial foundation for improving the economic benefits of flue-cured tobacco. However, since lesions on flue-cured tobacco leaves vary in shape and size, and are more similar in color and shape to mature lesions, especially brown star lesions and frog-eye lesions, which are visually very similar to mature lesions on tobacco leaves, it is difficult to distinguish them at a glance during procurement, resulting in reduced tobacco leaf quality. Therefore, identifying lesions and mature lesions on flue-cured tobacco leaves is more complex.

[0004] In summary, the quality research of tobacco leaves after flue-curing, especially the diseased spots and mature spots that are easily confused during purchase, the quality and utilization rate of tobacco leaves, and the long-term development of tobacco companies are of great significance. Summary of the Invention

[0005] The purpose of the present invention is to provide a method and system for classifying mature spots and diseased spots of flue-cured tobacco based on texture and shape characteristics, which has a fast classification speed and high accuracy, can effectively increase the proportion of medium and high-quality tobacco leaves during machine grading after flue-cured tobacco leaves, and thus increase economic benefits.

[0006] To achieve the above object, the present invention provides the following solutions:

[0007] In a first aspect, the present invention provides a method for classifying mature and diseased spots of flue-cured tobacco based on texture and shape features, comprising:

[0008] Get the currently collected flue-cured tobacco pictures;

[0009] Inputting the currently collected flue-cured tobacco image into a trained flue-cured tobacco mature spot and diseased spot classification model to classify mature spotted tobacco leaves and diseased spotted tobacco leaves in the currently collected flue-cured tobacco image; the diseased spotted tobacco leaves include brown spot diseased tobacco leaves and frog eye diseased tobacco leaves;

[0010] The trained flue-cured tobacco mature spot and diseased spot classification model is determined based on an Adaboost classifier, a circular LBP texture feature extraction module, a shape feature extraction module, and a sample data set; the sample data set includes multiple flue-cured tobacco sample images and corresponding annotation information; the annotation information includes: information characterizing the state of the flue-cured tobacco.

[0011] Preferably, the process of determining the sample data set is:

[0012] Obtain multiple single tobacco leaves;

[0013] Laying each single tobacco leaf flat, and photographing the front of each single tobacco leaf vertically to obtain multiple flue-cured tobacco sample pictures;

[0014] The flue-cured tobacco states in the plurality of flue-cured tobacco sample images are annotated to obtain a sample data set; the flue-cured tobacco states include normal and spotted; the spotted include mature spots and diseased spots; the diseased spots include red star spots and frog eye spots.

[0015] Preferably, the process of determining the trained flue-cured tobacco mature spot and diseased spot classification model is as follows:

[0016] Divide the sample data set into training set and test set;

[0017] Construct a classification model for mature and diseased spots of flue-cured tobacco;

[0018] The training set and the test set are respectively input into the flue-cured tobacco mature spot and lesion spot classification model for training and verification; when the classification accuracy reaches the expected level, the training ends and a trained flue-cured tobacco mature spot and lesion spot classification model is obtained.

[0019] Preferably, the training process of the flue-cured tobacco mature spot and diseased spot classification model is:

[0020] The circular LBP texture feature extraction module is used to extract texture features of normal tobacco leaves and spotted tobacco leaves in the training set.

[0021] The extracted normal tobacco leaf feature values ​​and spotted tobacco leaf feature values ​​are respectively input into the Adaboost classifier for training to classify normal tobacco leaves and spotted tobacco leaves; and the classification results are sampled and verified;

[0022] Using a shape feature extraction module to extract shape features of the mature spotted tobacco leaves and the diseased spotted tobacco leaves respectively;

[0023] The extracted feature values ​​of mature spotted tobacco leaves and diseased spotted tobacco leaves were respectively input into the Adaboost classifier for training to classify mature spotted tobacco leaves and diseased spotted tobacco leaves; and the classification results were sampled and verified.

[0024] Preferably, the circular LBP texture feature extraction module is used to extract texture features of normal tobacco leaves and spotted tobacco leaves in the training set, specifically:

[0025] (1)

[0026] (2)

[0027] (3)

[0028] in, It is the center point of the flue-cured tobacco image area. These are the sampling points in the flue-cured tobacco picture. is the number of feature points around the sampling point, is the radius of the feature points around the sampling point, is the pixel value of the sampling point in the flue-cured tobacco image.

[0029] Preferably, the maximum number of iterations of the Adaboost classifier is set to 10, and the number of random seeds is 0.

[0030] Preferably, the sampling classification test is a three-fold cross validation.

[0031] Preferably, the shape feature extraction module uses a shape complexity model to extract shape features of mature spotted tobacco leaves and diseased spotted tobacco leaves.

[0032] Preferably, the calculation formula of the shape complexity model is:

[0033] (4)

[0034] Where, is the pixel value of the sampling point in the flue-cured tobacco image; is the perimeter of the lesion; It represents the perimeter of the lesion per unit area; N is the total number of sampling points.

[0035] In a second aspect, the present invention further provides a classification system for flue-cured tobacco mature spots and diseased spots based on texture and shape features, comprising:

[0036] The flue-cured tobacco image acquisition module is used to obtain the currently collected flue-cured tobacco image;

[0037] a flue-cured tobacco mature spot and diseased spot classification module, configured to input the currently collected flue-cured tobacco image into a trained flue-cured tobacco mature spot and diseased spot classification model, so as to classify the mature-spotted tobacco leaves and diseased-spotted tobacco leaves in the currently collected flue-cured tobacco image; the diseased-spotted tobacco leaves include brown spot diseased tobacco leaves and frog eye diseased tobacco leaves;

[0038] The trained flue-cured tobacco mature spot and diseased spot classification model is determined based on an Adaboost classifier, a circular LBP texture feature extraction module, a shape feature extraction module, and a sample data set; the sample data set includes multiple flue-cured tobacco sample images and corresponding annotation information; the annotation information includes: information characterizing the state of the flue-cured tobacco.

[0039] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0040] The present invention classifies mature spots and diseased spots of flue-cured tobacco based on texture and shape features. Among them, the texture feature uses the different textures of mature spots on tobacco leaves and red star disease and frog eye disease spots to distinguish the textures of tobacco leaf disease spots, so that a large number of leaves that do not meet the requirements of diseased tobacco leaves can be quickly screened out, thereby greatly shortening the classification time. Especially at the current stage, flue-cured tobacco is classified by machine, and its classification time is slow and it cannot distinguish between mature spots and two common diseased spots, red heart disease spots and frog eye disease spots. The present invention has the advantages of fast classification speed and high accuracy; and combined with the shape characteristics of the diseased spots, the tobacco leaves are judged twice, ensuring the accuracy of the judgment of whether the tobacco leaves have diseased spots. In summary, the present invention has a fast classification speed and higher accuracy, and can reduce the confusion between mature spots and diseased spots when purchasing tobacco leaves, laying a good foundation for subsequent tobacco leaf machine classification, grading and other applications, thereby effectively increasing the proportion of medium and high-quality tobacco leaves during machine grading after flue-cured tobacco leaves, and increasing economic benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0042] Figure 1 This is a flow chart of a method for classifying mature and diseased spots of flue-cured tobacco based on texture and shape features according to an embodiment of the present invention;

[0043] Figure 2 This is a structural diagram of a classification system for mature and diseased spots of flue-cured tobacco based on texture and shape features according to an embodiment of the present invention. DETAILED DESCRIPTION

[0044] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0045] The purpose of the present invention is to provide a method and system for classifying mature spots and diseased spots of flue-cured tobacco based on texture and shape characteristics. The method and system have fast classification speed and high accuracy, and can effectively increase the proportion of medium and high-quality tobacco leaves during machine grading after flue-cured tobacco leaves, thereby increasing economic benefits.

[0046] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0047] like Figure 1 As shown, an embodiment of the present invention provides a method for classifying mature spots and diseased spots of flue-cured tobacco based on texture and shape features, including:

[0048] Step 100: Obtain the currently collected flue-cured tobacco image.

[0049] Step 200: Input the currently collected flue-cured tobacco image into a trained flue-cured tobacco mature spot and diseased spot classification model to classify the mature spotted tobacco leaves and diseased spotted tobacco leaves in the currently collected flue-cured tobacco image; the diseased spotted tobacco leaves include brown spot diseased tobacco leaves and frog eye diseased tobacco leaves.

[0050] The trained flue-cured tobacco mature spot and diseased spot classification model is determined based on an Adaboost classifier, a circular LBP texture feature extraction module, a shape feature extraction module, and a sample data set; the sample data set includes multiple flue-cured tobacco sample images and corresponding annotation information; the annotation information includes: information characterizing the state of the flue-cured tobacco.

[0051] Furthermore, the process of determining the sample data set is as follows:

[0052] Step A: Obtain multiple single tobacco leaves.

[0053] Step B: each single tobacco leaf is laid out flat, and each single tobacco leaf is photographed vertically from the front to obtain a plurality of flue-cured tobacco sample pictures.

[0054] Step C: annotating the flue-cured tobacco status in the plurality of flue-cured tobacco sample images to obtain a sample data set; the flue-cured tobacco status includes normal and spotted; the spotted includes mature spots and diseased spots; the diseased spots include red star spots and frog eye spots.

[0055] Furthermore, the process of determining the trained classification model for flue-cured tobacco mature spots and diseased spots is as follows:

[0056] Step a: Divide the sample dataset into training set and test set.

[0057] Step b: Construct a classification model for mature and diseased spots of flue-cured tobacco.

[0058] Step c: inputting the training set and the test set into the flue-cured tobacco mature spot and lesion spot classification model for training and verification respectively; when the classification accuracy reaches the expected level, the training ends and a trained flue-cured tobacco mature spot and lesion spot classification model is obtained.

[0059] The training process of the flue-cured tobacco mature spot and diseased spot classification model is as follows:

[0060] Step c1: Using the circular LBP texture feature extraction module, texture features are extracted from normal tobacco leaves and spotted tobacco leaves in the training set respectively;

[0061] The step c1 specifically includes: inputting normal tobacco leaves and spotted tobacco leaves into the LBP texture feature extraction module respectively. The circular LBP feature formula included in the circular LBP feature extraction is as follows:

[0062] (1)

[0063] (2)

[0064] (3)

[0065] in, It is the center point of the flue-cured tobacco image area. These are the sampling points in the flue-cured tobacco picture. is the number of feature points around the sampling point, is the radius of the feature points around the sampling point, is the pixel value of the sampling point in the flue-cured tobacco image. In this embodiment, R=100 and P=5 are selected to perform LBP feature extraction.

[0066] The coordinates of any sampling point in the flue-cured tobacco image are obtained by using equations (1) and (2), and the pixel value of the sampling point is obtained by using equation (3), which is the LBP feature value of the point.

[0067] Step c2: inputting the extracted normal tobacco leaf feature values ​​and spotted tobacco leaf feature values ​​into an Adaboost classifier for training to classify normal tobacco leaves and spotted tobacco leaves; and performing sampling classification verification on the classification results;

[0068] The sampling classification test is a three-fold cross-validation, which can avoid that the number of diseased spot images is too small compared to the number of normal tobacco leaf images, thereby improving the fault tolerance rate and reducing the specificity and limitations of the sample data set.

[0069] The maximum number of iterations of the Adaboost classifier is set to 10, and the number of random seeds is 0.

[0070] Step c3: using a shape feature extraction module to extract shape features of the mature spotted tobacco leaves and the diseased spotted tobacco leaves respectively;

[0071] The step c3 specifically comprises: extracting shape features of tobacco leaf spots using the spot complexity of the mature spotted tobacco leaves and the diseased spotted tobacco leaves screened out in c2. The present invention uses a shape complexity model to extract shape features of the mature spotted tobacco leaves and the diseased spotted tobacco leaves.

[0072] The calculation formula of the shape complexity model is:

[0073] (4)

[0074] Where, is the pixel value of the sampling point in the flue-cured tobacco image; is the perimeter of the lesion; It represents the perimeter of the lesion per unit area; N is the total number of sampling points.

[0075] Among them, complexity refers to the perimeter length of tobacco leaf spots per unit area. The smaller the value, the simpler the graph, and vice versa.

[0076] Step c4: The extracted feature values ​​of mature spotted tobacco leaves and diseased spotted tobacco leaves are respectively input into the Adaboost classifier for training to classify mature spotted tobacco leaves and diseased spotted tobacco leaves; and the classification results are sampled and verified.

[0077] The parameters and verification method of step c4 are the same as those of step c2 and will not be repeated here.

[0078] like Figure 2 As shown, the present invention also provides a classification system for flue-cured tobacco mature spots and diseased spots based on texture and shape characteristics, comprising:

[0079] The flue-cured tobacco image acquisition module is used to obtain the currently collected flue-cured tobacco image.

[0080] The flue-cured tobacco mature spot and diseased spot classification module is used to input the currently collected flue-cured tobacco image into the trained flue-cured tobacco mature spot and diseased spot classification model to classify the mature spotted tobacco leaves and diseased tobacco leaves in the currently collected flue-cured tobacco image; the diseased tobacco leaves include brown spot disease tobacco leaves and frog eye disease tobacco leaves.

[0081] The trained flue-cured tobacco mature spot and diseased spot classification model is determined based on an Adaboost classifier, a circular LBP texture feature extraction module, a shape feature extraction module, and a sample data set; the sample data set includes multiple flue-cured tobacco sample images and corresponding annotation information; the annotation information includes: information characterizing the state of the flue-cured tobacco.

[0082] This invention takes into account the complexity and similarity of the appearance of mature tobacco spots, brown star spots, and frog-eye spots. Using LBP texture features alone cannot achieve good recognition results, and the accuracy of classification and recognition directly affects the subsequent grading and procurement of flue-cured tobacco. Therefore, a method combining shape features and an Adaboost classifier is used to improve classification results. Through experiments and testing of this method, a method for distinguishing mature tobacco spots from brown star spots and frog-eye spots was ultimately determined. This method achieves the classification and recognition of mature tobacco spots and two types of brown star spots and frog-eye spots after flue-cured tobacco leaves in an efficient and highly accurate manner.

[0083] The present invention includes a method for distinguishing between texture features and shape features. The texture feature uses the different textures of mature tobacco leaves, brown spot, and frog-eye spots to distinguish between the textures of tobacco leaf spots, allowing a large number of leaves that do not meet the diseased tobacco leaf classification criteria to be quickly screened out, thereby greatly shortening the classification time. The present invention has the advantages of fast classification speed and high accuracy. The current method of machine classification of flue-cured tobacco results in slow classification time and an inability to distinguish between mature tobacco leaves and red heart and frog-eye spots, two common diseased spots. Combining the shape characteristics of the diseased spots allows for a secondary determination of the tobacco leaves, ensuring the accuracy of the determination of whether the tobacco leaves have diseased spots.

[0084] In summary, the judgment results of the present invention are highly accurate, and can reduce the confusion between mature spots and diseased spots when purchasing tobacco leaves, laying a good foundation for subsequent tobacco leaf machine classification, grading and other applications, and improving its accuracy, thereby effectively increasing the proportion of medium and high-quality tobacco leaves during machine grading after flue-cured tobacco leaves, thereby increasing economic benefits.

[0085] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the various embodiments can be referenced to each other.

[0086] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The above examples are only intended to help understand the method and core concept of the present invention. At the same time, those skilled in the art will find that the specific implementation methods and application scopes may vary based on the concept of the present invention. In summary, the contents of this specification should not be construed as limiting the present invention.

Claims

1. A method for classifying mature and diseased spots of flue-cured tobacco based on texture and shape features, characterized in that: include: Get the currently collected flue-cured tobacco pictures; Inputting the currently collected flue-cured tobacco image into a trained flue-cured tobacco mature spot and diseased spot classification model to classify mature spotted tobacco leaves and diseased spotted tobacco leaves in the currently collected flue-cured tobacco image; the diseased spotted tobacco leaves include brown spot diseased tobacco leaves and frog eye diseased tobacco leaves; The trained flue-cured tobacco mature spot and diseased spot classification model is determined based on an Adaboost classifier, a circular LBP texture feature extraction module, a shape feature extraction module, and a sample data set; The specific training process of the classification model for mature and diseased tobacco spots is as follows: The circular LBP texture feature extraction module is used to extract texture features of normal tobacco leaves and spotted tobacco leaves in the training set. The extracted normal tobacco leaf feature values ​​and spotted tobacco leaf feature values ​​are respectively input into the Adaboost classifier for training to classify normal tobacco leaves and spotted tobacco leaves; and the classification results are sampled and verified; Using a shape feature extraction module to extract shape features of the mature spotted tobacco leaves and the diseased spotted tobacco leaves respectively; The extracted mature spotted tobacco leaf feature values ​​and diseased spotted tobacco leaf feature values ​​were input into Adaboost classifier for training respectively to classify mature spotted tobacco leaves and diseased spotted tobacco leaves; And conduct sampling classification verification on the classification results; The sample data set includes multiple flue-cured tobacco sample images and corresponding annotation information; The labeling information includes information indicating the state of the flue-cured tobacco.

2. The method for classifying mature and diseased spots of flue-cured tobacco based on texture and shape features according to claim 1, characterized in that: The process of determining the sample data set is as follows: Obtain multiple single tobacco leaves; Laying each single tobacco leaf flat, and photographing the front of each single tobacco leaf vertically to obtain multiple flue-cured tobacco sample pictures; The flue-cured tobacco states in the plurality of flue-cured tobacco sample images are annotated to obtain a sample data set; the flue-cured tobacco states include normal and spotted; the spotted include mature spots and diseased spots; the diseased spots include red star spots and frog eye spots.

3. The method for classifying mature and diseased spots of flue-cured tobacco based on texture and shape features according to claim 1, characterized in that: The process of determining the trained flue-cured tobacco mature spot and diseased spot classification model is as follows: Divide the sample data set into training set and test set; Construct a classification model for mature and diseased spots of flue-cured tobacco; The training set and the test set are respectively input into the flue-cured tobacco mature spot and lesion spot classification model for training and verification; when the classification accuracy reaches the expected level, the training ends and a trained flue-cured tobacco mature spot and lesion spot classification model is obtained.

4. The method for classifying mature and diseased spots of flue-cured tobacco based on texture and shape features according to claim 1, wherein: The circular LBP texture feature extraction module is used to extract texture features of normal tobacco leaves and spotted tobacco leaves in the training set, specifically: （1） （2） （3） in, It is the center point of the flue-cured tobacco image area. These are the sampling points in the flue-cured tobacco picture. is the number of feature points around the sampling point, is the radius of the feature points around the sampling point, is the pixel value of the sampling point in the flue-cured tobacco image.

5. The method for classifying mature and diseased spots of flue-cured tobacco based on texture and shape features according to claim 1, wherein: The maximum number of iterations of the Adaboost classifier is set to 10, and the number of random seeds is 0.

6. The method for classifying mature and diseased spots of flue-cured tobacco based on texture and shape features according to claim 1, wherein: The sampling classification test is a three-fold cross validation.

7. The method for classifying mature and diseased spots of flue-cured tobacco based on texture and shape features according to claim 1, wherein: The shape feature extraction module uses a shape complexity model to extract shape features of mature spotted tobacco leaves and diseased spotted tobacco leaves.

8. The method for classifying mature and diseased spots of flue-cured tobacco based on texture and shape features according to claim 7, characterized in that: The calculation formula of the shape complexity model is: （4） Where, is the pixel value of the sampling point in the flue-cured tobacco image; is the perimeter of the lesion; It represents the perimeter of the lesion per unit area; N is the total number of sampling points.

9. A classification system for mature and diseased spots of flue-cured tobacco based on texture and shape features, characterized in that: include: The flue-cured tobacco image acquisition module is used to obtain the currently collected flue-cured tobacco image; a flue-cured tobacco mature spot and diseased spot classification module, configured to input the currently collected flue-cured tobacco image into a trained flue-cured tobacco mature spot and diseased spot classification model, so as to classify the mature-spotted tobacco leaves and diseased-spotted tobacco leaves in the currently collected flue-cured tobacco image; the diseased-spotted tobacco leaves include brown spot diseased tobacco leaves and frog eye diseased tobacco leaves; The trained flue-cured tobacco mature spot and diseased spot classification model is determined based on an Adaboost classifier, a circular LBP texture feature extraction module, a shape feature extraction module, and a sample data set; The specific training process of the flue-cured tobacco mature spot and diseased spot classification model is as follows: The circular LBP texture feature extraction module is used to extract texture features of normal tobacco leaves and spotted tobacco leaves in the training set. The extracted normal tobacco leaf feature values ​​and spotted tobacco leaf feature values ​​are respectively input into the Adaboost classifier for training to classify normal tobacco leaves and spotted tobacco leaves; and the classification results are sampled and verified; Using a shape feature extraction module to extract shape features of the mature spotted tobacco leaves and the diseased spotted tobacco leaves respectively; The extracted mature spotted tobacco leaf feature values ​​and diseased spotted tobacco leaf feature values ​​were input into Adaboost classifier for training respectively to classify mature spotted tobacco leaves and diseased spotted tobacco leaves; And conduct sampling classification verification on the classification results; The sample data set includes multiple flue-cured tobacco sample images and corresponding annotation information; The labeling information includes information indicating the state of the flue-cured tobacco.

Images