Cotton bollworm age identification method

Through automated acquisition and UNet network segmentation technology, combined with feature selection algorithm, a feature space model is constructed, which solves the problem of insufficient adaptability of cotton bollworm age recognition under complex background and lighting conditions, and achieves high accuracy and rapid identification, which improves the Generalization ability and recognition speed.

CN119067986APending Publication Date: 2024-12-03SHENZHEN INST OF ADVANCED TECH CHINESE ACAD OF SCI
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410943966.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-15
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

The existing cotton bollworm age recognition technology is not adaptable under complex backgrounds and different lighting conditions, requires a large amount of labeling data, and the generalization ability needs to be improved.

Method used

Cotton bollworm images are collected through an automated acquisition system to improve image quality, use UNet network for segmentation and morphological parameter extraction, combine texture structure and feature information to build feature vectors, and filter out the most valuable features through feature selection algorithms to establish feature space The model is used for identification.

Benefits of technology

It realizes fast, automatic and high-accuracy recognition of cotton bollworm instars, improves the accuracy and robustness of identification in complex backgrounds and variable environments, reduces dependence on labeled data, and enhances the generalization of the model. Capability and recognition speed.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119067986A_ABST
    Figure CN119067986A_ABST
Patent Text Reader

Abstract

The invention provides a cotton bollworm age identification method comprising the following steps: collecting cotton bollworm images, and constructing a cotton bollworm image data set; the cotton bollworm image quality is improved, and key features of cotton bollworms are highlighted; performing accurate segmentation on the cotton bollworm image after quality improvement, extracting cotton bollworms and morphological parameters thereof, and taking the morphological parameters as features; converting the features into feature vectors; screening out most valuable features for cotton bollworm age identification through the feature vectors, and establishing a feature space model based on the most valuable features; and accurately identifying the age of the cotton bollworm by adopting the feature space model to realize automatic classification. The generalization ability and stability of the model under different environments and change conditions are improved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of agricultural monitoring technology, and in particular to a method for identifying the age of cotton bollworm. Background Art

[0002] Cotton bollworm is one of the major agricultural pests worldwide, causing serious damage to important economic crops such as cotton. Traditional pest management methods rely on the use of pesticides, but these methods are not only costly but may also have negative impacts on the environment and human health. Therefore, it is particularly important to develop an efficient and accurate technology for identifying the age of cotton bollworm, which can help implement more precise pest management strategies, such as biological control and integrated pest management (IPM).

[0003] In recent years, with the rapid development of machine learning and computer vision technology, the use of image recognition technology for pest identification and classification has become a research hotspot. These technologies provide technical support for precision agriculture by automatically analyzing pest images to identify their species and growth stages.

[0004] Existing methods usually include the following steps: 1) Data preparation: Collect and annotate a large amount of cotton bollworm image data, including cotton bollworms at different ages. These data are used to train and verify deep learning models. 2) Image preprocessing: Preprocess the collected images to improve the efficiency and accuracy of subsequent model training. Preprocessing steps may include image cropping, scaling, rotation, and color standardization. 3) Feature extraction and classification: Use deep convolutional neural networks to automatically learn features from preprocessed images and classify the ages of cotton bollworms based on these features. Typical models include AlexNet, VGGNet, ResNet, etc. 4) Model training and optimization: Train deep learning models with a large amount of training data, and use methods such as cross entropy loss functions and back propagation algorithms to optimize model parameters. In addition, techniques such as data augmentation and transfer learning may also be used to improve the generalization ability and accuracy of the model.

[0005] Although these existing solutions have achieved automatic recognition and classification of pest images to a certain extent, they still face some challenges, such as insufficient adaptability to complex backgrounds and different lighting conditions, the need for a large amount of labeled data for model training, and the need to improve generalization capabilities in practical applications. Summary of the invention

[0006] 1. Technical issues to be solved

[0007] Based on the existing solutions, automatic recognition and classification of pest images have been achieved to a certain extent, but they still face some challenges, such as insufficient adaptability to complex backgrounds and different lighting conditions, the need for a large amount of labeled data for model training, and the need to improve the generalization ability in practical applications. This application provides a method for identifying the age of cotton bollworm.

[0008] 2. Technical solution

[0009] In order to achieve the above-mentioned purpose, the present application provides a method for identifying the age of cotton bollworm, which includes collecting cotton bollworm images and constructing a cotton bollworm image dataset; improving the quality of the cotton bollworm images and highlighting the key features of the cotton bollworm; accurately segmenting the cotton bollworm images with improved quality, extracting cotton bollworms and their morphological parameters, and using the morphological parameters as features; converting the features into feature vectors; filtering out the most valuable features for identifying the age of cotton bollworm through the feature vectors, and establishing a feature space model based on the most valuable features; and using the feature space model to accurately identify the age of cotton bollworm and realize automatic classification.

[0010] In one implementation, the collecting of cotton bollworm images and the construction of a cotton bollworm image dataset include starting an automated collection system, setting collection parameters and environmental conditions, collecting the cotton bollworm images under different environmental conditions, classifying and labeling the cotton bollworm images, selecting the labeled cotton bollworm images as a dataset for training a machine learning model, and using the remaining labeled cotton bollworm images as a test set. The training dataset and the test set constitute the cotton bollworm image dataset.

[0011] In another implementation, improving the quality of the cotton bollworm image and highlighting the key features of the cotton bollworm includes adjusting the low-frequency details in the cotton bollworm image, enhancing the medium-frequency details in the cotton bollworm image, sharpening the high-frequency details in the cotton bollworm image, and combining the adjustment results of the low frequency, medium frequency and high frequency to obtain an improved cotton bollworm image at all scales, so that the key features of the cotton bollworm are highlighted.

[0012] In another implementation, the image is accurately segmented, and the cotton bollworm and its morphological parameters are extracted as features, including using a UNet network to segment the cotton bollworm image after quality improvement, separating the cotton bollworm from the background so that the outline and morphology of the cotton bollworm are clearly demarcated, extracting the morphological parameters of the cotton bollworm, and using the morphological parameters as features.

[0013] In another implementation, converting the features into feature vectors includes merging the features into a comprehensive feature vector.

[0014] In another implementation, the most valuable features for identifying the age of cotton bollworm are screened out through the feature vector, and a feature space model is established based on the most valuable features, including applying a feature selection algorithm to the merged feature vector to screen out the most valuable features, and establishing a feature space model based on the most valuable features.

[0015] In another implementation, the use of the feature space model to accurately identify the age of cotton bollworms and achieve automatic classification includes training the feature space model, inputting the most valuable features into the feature space model, and attempting to identify the age of cotton bollworms. If the feature space model successfully identifies the age of cotton bollworms, that is, the recognition result is consistent with the actual situation, then the recognition process will continue to the next step and output the recognition result. If the feature space model fails to successfully identify the age of cotton bollworms, or the recognition result is inaccurate, then the model parameters will need to be adjusted to improve the recognition performance; and automatic classification of cotton bollworms is achieved through the recognition result.

[0016] In another implementation, the morphological parameters include the length of the cotton bollworm, the width of the cotton bollworm, the area of ​​the cotton bollworm, the texture pattern of the cotton bollworm surface, the structural details of the cotton bollworm, and the morphological characteristics of the cotton bollworm.

[0017] In another implementation, the texture pattern, the structural details and the morphological features are combined into a feature vector, and the most valuable features for identifying the age of cotton bollworm are screened out through a feature selection algorithm.

[0018] 3. Beneficial effects

[0019] Compared with the prior art, the beneficial effects of the cotton bollworm age identification method provided by the present application are:

[0020] The method for identifying the age of cotton bollworm provided in the present application is a method for identifying the age of cotton bollworm based on machine learning. It realizes rapid, automatic and highly accurate identification of the age of cotton bollworm through machine learning technology, overcoming the low efficiency and inaccuracy of manual identification in traditional methods.

[0021] The cotton bollworm age identification method provided in this application solves the problem that the cotton bollworm images collected in the natural environment have different quality, are greatly affected by factors such as lighting and background interference, and affect the accuracy and efficiency of subsequent processing. The image data quality is improved through effective image enhancement technology.

[0022] The cotton bollworm age recognition method provided in the present application can accurately extract features useful for age recognition in the face of complex backgrounds and variable morphological features in cotton bollworm images.

[0023] The method for identifying the age of cotton bollworm provided in this application selects the features that are most helpful for age identification from among many possible features, and constructs a model with high generalization ability based on these features, ensuring that the proposed method can maintain high accuracy and stability in different environmental conditions and cotton bollworm populations.

[0024] The cotton bollworm age recognition method provided in the present application improves adaptability to complex backgrounds: through improved image preprocessing and enhancement technology, as well as more sophisticated feature extraction methods, the accuracy and robustness of the model in identifying the cotton bollworm age in complex backgrounds are improved.

[0025] The cotton bollworm age identification method provided in the present application reduces the reliance on a large amount of labeled data: it adopts semi-supervised learning, transfer learning or data enhancement techniques to reduce the need for a large amount of labeled data while maintaining or improving the recognition accuracy.

[0026] The cotton bollworm age identification method provided in this application enhances the generalization ability of the model: by designing more general feature extraction and classification strategies, and adopting advanced model training techniques (such as domain adaptation, model distillation, etc.), the generalization ability and stability of the model in different environments and changing conditions are improved.

[0027] The cotton bollworm age identification method provided in this application improves the identification speed and efficiency: optimizes the model structure and calculation method to achieve fast and accurate cotton bollworm age identification, especially on resource-constrained devices to meet the needs of real-time processing.

[0028] The cotton bollworm age identification method provided in the present application solves the problems existing in the prior art in cotton bollworm age identification by introducing advanced machine learning technology and algorithm improvement, and achieves more accurate, efficient and stable identification performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 This is a schematic diagram of the principle of the cotton bollworm age identification method of the present application;

[0030] Figure 2 It is a schematic diagram of the process of generating the cotton bollworm image dataset of the present application;

[0031] Figure 3 is a schematic diagram of the image enhancement process of the present application;

[0032] Figure 4 This is a schematic diagram of the UNet feature network segmentation and morphological parameter extraction process of this application;

[0033] Figure 5 It is a schematic diagram of the process of forming a feature vector composition and a feature space model of the present application;

[0034] Figure 6This is a schematic diagram of the insect age identification process of cotton bollworm in the present application. DETAILED DESCRIPTION

[0035] Hereinafter, the specific embodiments of the present application will be described in detail with reference to the accompanying drawings. According to these detailed descriptions, those skilled in the art can clearly understand the present application and can implement the present application. Without violating the principles of the present application, the features in the various embodiments can be combined to obtain new implementations, or certain features in certain embodiments can be replaced to obtain other preferred implementations.

[0036] See also Figures 1 to 6 The present application provides a method for identifying the age of cotton bollworm, the method comprising collecting cotton bollworm images and constructing a cotton bollworm image dataset; improving the quality of the cotton bollworm images and highlighting the key features of the cotton bollworm; accurately segmenting the cotton bollworm images with improved quality, extracting the cotton bollworms and their morphological parameters, and using the morphological parameters as features; converting the features into feature vectors; filtering out the most valuable features for identifying the age of cotton bollworm through the feature vectors, and establishing a feature space model based on the most valuable features; using the feature space model to accurately identify the age of cotton bollworm and realize automatic classification.

[0037] Specifically, (1) the acquisition system generates a cotton bollworm image dataset: an automated acquisition system is used to collect cotton bollworm images under different environmental conditions, and a dataset for training and testing is constructed. The training dataset is combined with the test dataset to form a cotton bollworm image dataset. (2) Image enhancement (multi-scale detail enhancement): the image quality is improved by multi-scale detail enhancement technology, and the key features of cotton bollworm are highlighted for subsequent processing. (3) Segmentation and morphological parameter extraction based on UNet feature network: the image is accurately segmented using the UNet network, and cotton bollworm and its related morphological parameters are extracted as features. (4) Feature vector construction (combining texture structure, texture features, etc.): the texture structure and texture features extracted from the image are combined into a feature vector to provide input data for the recognition model. (5) Feature selection and feature space model establishment: the most valuable features for cotton bollworm age recognition are selected through the feature selection algorithm, and a feature space model is established based on these features. (6) Cotton bollworm age recognition: the trained feature space model is used to accurately identify the cotton bollworm age and realize automatic classification.

[0038] Furthermore, the collecting of cotton bollworm images and the construction of a cotton bollworm image dataset include starting an automated acquisition system, setting acquisition parameters and environmental conditions, collecting the cotton bollworm images under different environmental conditions, classifying and labeling the cotton bollworm images, selecting the labeled cotton bollworm images as a dataset for training a machine learning model, and using the remaining labeled cotton bollworm images as a test set. The training dataset and the test set constitute the cotton bollworm image dataset.

[0039] The process of generating a cotton bollworm image dataset by the acquisition system involves using a highly automated acquisition system to systematically collect images of cotton bollworm under carefully set different environmental conditions. This process starts with starting the acquisition system and covers the setting of environmental conditions, automatic collection of images, and subsequent image classification and annotation. In this way, a rich and diverse collection of cotton bollworm images can be obtained, which not only reflects the physiological responses and appearance changes of cotton bollworm under different environmental conditions, but also provides high-quality raw data for subsequent image processing and feature extraction through accurate classification and annotation. Finally, these carefully prepared image data are organized into training and testing datasets to support the development and verification of cotton bollworm recognition models, aiming to achieve automated and accurate identification of cotton bollworm ages and types. This whole process not only improves the efficiency and quality of data collection, but also provides important technical support for cotton bollworm-related research and applications.

[0040] Specifically, (1) starting the automated acquisition system: initializing the automated acquisition system, including starting the camera, setting acquisition parameters (such as resolution, frame rate, etc.), and ensuring that all system components operate normally.

[0041] (2) Setting environmental conditions: According to research requirements, adjust the conditions of the laboratory or field collection environment, such as temperature, humidity, light intensity, etc., to simulate the natural growth of cotton bollworm or specific experimental conditions.

[0042] (3) Collecting images under different environmental conditions: Under the preset environmental conditions, the system automatically takes pictures of cotton bollworms and collects images of cotton bollworms under different environmental conditions. This step aims to capture the changes in the appearance of cotton bollworms in various environments.

[0043] (4) Image classification and annotation: The collected images are classified and grouped according to the different ages, types or other relevant features of the cotton bollworm. Then, the images are annotated, including the location, size, morphological characteristics and other information of the cotton bollworm, to provide accurate data for subsequent image analysis and model training.

[0044] (5) Constructing a training dataset: Select a portion of the classified and labeled images to construct a dataset for machine learning model training. These images will be used to train the model to identify and analyze the characteristics of cotton bollworms.

[0045] (6) Constructing a test dataset: Select another portion of the remaining classified and labeled images to construct a dataset for testing and verifying the model performance. These images are used to evaluate the accuracy and reliability of the model in identifying cotton bollworms.

[0046] (7) Cotton bollworm image dataset: The training dataset and the test dataset are combined to form a complete cotton bollworm image dataset. This dataset contains a wealth of cotton bollworm images, covering different environmental conditions, different ages, and diverse morphological characteristics, providing basic data for the automatic identification and analysis of cotton bollworms.

[0047] Through this series of detailed steps, the researchers were able to systematically collect and prepare high-quality cotton bollworm image data, laying the foundation for further image processing, feature extraction, and model training.

[0048] Furthermore, improving the quality of the cotton bollworm image and highlighting the key features of the cotton bollworm includes adjusting the low-frequency details in the cotton bollworm image, enhancing the medium-frequency details in the cotton bollworm image, sharpening the high-frequency details in the cotton bollworm image, and combining the adjustment results of the low frequency, medium frequency and high frequency to obtain an improved cotton bollworm image at all scales, so that the key features of the cotton bollworm are highlighted.

[0049] The image enhancement (multi-scale detail enhancement) process is a series of sophisticated image processing steps that aims to significantly improve the quality of cotton bollworm images through technical means, especially to highlight the key features of cotton bollworm for subsequent image analysis and recognition. This process starts with the original image and first applies multi-scale detail enhancement technology, which includes targeted adjustment and enhancement of the low-frequency, medium-frequency, and high-frequency details of the image. Low-frequency detail adjustment mainly improves the overall brightness and contrast of the image, medium-frequency detail enhancement focuses on improving the clarity of the texture and contour of the cotton bollworm, and high-frequency detail sharpening focuses on refining the edges and tiny parts of the cotton bollworm to make it sharper. Through these comprehensive adjustments, the image is optimized at different scales, and the final generated image is visually clearer. The key features of the cotton bollworm, such as shape and texture, are effectively highlighted, providing a high-quality foundation for subsequent feature extraction and recognition. This step is a key link in achieving high-precision cotton bollworm recognition, ensuring that the recognition algorithm can accurately analyze and identify the characteristics of the cotton bollworm.

[0050] Specifically, (1) Original cotton bollworm images: These are unprocessed cotton bollworm images obtained from the automated acquisition system, which serve as the starting point for the image enhancement process. These images may have inconsistent quality due to environmental conditions, lighting, and other factors.

[0051] (2) Multi-scale detail enhancement: The original image is processed using multi-scale detail enhancement technology. This technology improves the image quality by adjusting the image details at different scales (frequencies) to make the characteristics of the cotton bollworm more prominent.

[0052] Low-frequency detail adjustment: Adjusts the low-frequency details in the image, primarily affecting the image's large-scale brightness and contrast. This helps improve the overall visual effect of the image, making the contrast between the background and the bollworm more obvious.

[0053] Mid-frequency detail enhancement: Enhance the mid-frequency details in the image, which affects the texture and contour edges of the cotton bollworm. By enhancing the mid-frequency details, the morphological features of the cotton bollworm can be made clearer, which is convenient for subsequent identification and analysis.

[0054] High-frequency detail sharpening: sharpens the high-frequency details in the image, focusing on the small parts and edge details of the cotton bollworm. This step makes the details of the image sharper and helps to highlight the tiny features of the cotton bollworm.

[0055] Combined adjusted image: The adjustments of low, mid and high frequencies are combined to obtain an image that is improved at all scales. This step ensures that the image is optimized at different levels of detail.

[0056] Images highlighting key features of cotton bollworm: Images after multi-scale detail enhancement, in which key features of cotton bollworm are effectively highlighted. These images provide clearer and more accurate visual information for subsequent image processing, feature extraction, and cotton bollworm identification.

[0057] Through this series of detailed image enhancement steps, the quality of cotton bollworm images can be significantly improved, making the key features of cotton bollworm more obvious, thus providing high-quality input images for subsequent image analysis and cotton bollworm identification.

[0058] Furthermore, the image is accurately segmented, and the cotton bollworm and its morphological parameters are extracted as features, including using a UNet network to segment the cotton bollworm image after quality improvement, separating the cotton bollworm from the background, so that the outline and morphology of the cotton bollworm are clearly divided, extracting the morphological parameters of the cotton bollworm, and using the morphological parameters as features.

[0059] The UNet feature network segmentation and morphological parameter extraction step is a refined image processing process that aims to accurately segment the cotton bollworm from the enhanced cotton bollworm image and extract key morphological parameters from these segmentation results. This process first uses the UNet network, a deep learning model designed specifically for image segmentation, to process the image. By learning the features in the image, the UNet network can accurately segment the cotton bollworm from the background and provide a clear outline and morphology of the cotton bollworm.

[0060] After segmentation, morphological parameters such as length, width, and area are further extracted from these clearly defined bollworm images, which are key to understanding the physical characteristics of bollworms. These morphological parameters are then used as features for further identification and classification of bollworms, providing accurate input data for machine learning models. The entire process not only improves the accuracy of image segmentation, but also provides a solid foundation for the automatic identification and classification of bollworms by extracting meaningful morphological parameters. This method effectively combines deep learning technology with traditional morphological analysis, demonstrating the powerful capabilities and application potential of deep learning models in complex image processing tasks.

[0061] Specifically, (1) Enhanced bollworm images: This step involves using bollworm images that have been previously processed with multi-scale detail enhancement. These images have been optimized to ensure that the key features of the bollworm are highlighted, providing an ideal starting point for accurate segmentation and feature extraction.

[0062] (2) Applying UNet network for image segmentation: In this step, the enhanced cotton bollworm image is processed using the UNet network. UNet is a deep learning model designed for image segmentation tasks that can accurately segment objects of interest from images. In this process, the UNet network learns the characteristics of cotton bollworms in the image and achieves accurate segmentation of cotton bollworms and background.

[0063] (3) Segmentation of cotton bollworm images: After being processed by the UNet network, the resulting images are accurately segmented cotton bollworm images. In these images, the outline and morphology of the cotton bollworm are clearly demarcated and distinguished from the background, providing a clear visual basis for further analysis.

[0064] (4) Extracting morphological parameters: Morphological parameters are extracted from the segmented cotton bollworm images, including important physical and morphological characteristics of the cotton bollworm, such as length, width, and area. These parameters directly reflect the size and shape of the cotton bollworm and are an important basis for subsequent identification and classification.

[0065] (5) Morphological parameters as features: The extracted morphological parameters are used as features for further identification and classification of cotton bollworm. These features provide key input data for the machine learning model, enabling the model to accurately identify and classify cotton bollworm based on its physical and morphological characteristics.

[0066] Through this series of steps, cotton bollworms can be effectively segmented from images, and morphological parameters that facilitate identification and classification can be extracted from the segmentation results, providing key technical support for achieving high-precision automatic identification of cotton bollworms.

[0067] Furthermore, converting the features into feature vectors includes merging the features into a comprehensive feature vector.

[0068] In the process of "feature vector construction and feature space model establishment", key texture structures and feature information are first extracted from the preprocessed and enhanced cotton bollworm images, including but not limited to the texture patterns and structural details of the cotton bollworm surface, which are crucial for subsequent recognition and classification tasks. These extracted features are combined into a comprehensive feature vector to comprehensively describe the characteristics of cotton bollworm.

[0069] The feature selection algorithm is applied to analyze these feature vectors to evaluate the contribution of each feature to the task of identifying the age of cotton bollworm. This step uses logical judgment to screen out the features that contribute significantly to the identification task, while eliminating those that do not contribute significantly. This not only optimizes the feature set and improves the efficiency and accuracy of the model, but also avoids unnecessary increase in model complexity. Based on the contributing features retained after screening, a feature space model is established. This model integrates all key features and takes into account the interactions between features, providing strong data support for the accurate identification and classification of cotton bollworm. Through this process, it is ensured that the cotton bollworm age identification model can be trained based on the most informative features, thereby achieving high-precision identification results.

[0070] Furthermore, the most valuable features for identifying the age of cotton bollworm are screened out through the feature vector, and a feature space model is established based on the most valuable features, including applying a feature selection algorithm to the merged feature vector to screen out the most valuable features, and establishing a feature space model based on the most valuable features.

[0071] (1) Extracted texture structure and texture features: Extract key texture information, including texture structure and features, from the preprocessed and enhanced cotton bollworm images. This information captures the detailed patterns on the surface of the cotton bollworm, which is crucial for identification and classification.

[0072] (2) Merge into feature vector: Merge the extracted texture structure and feature information into a comprehensive feature vector. This step aims to create a data structure that comprehensively describes the characteristics of cotton bollworm and provides input for subsequent analysis and model training.

[0073] (3) Applying feature selection algorithm: Applying feature selection algorithm to the merged feature vector. This algorithm evaluates the contribution of each feature to the cotton bollworm age identification task to determine which features are most valuable.

[0074] (4) Logical reasoning

[0075] 1) Features with significant contributions: Features that are identified by the feature selection algorithm as having significant contributions to the recognition task will be retained.

[0076] 2) Features with no significant contribution: Features that are considered to have no significant contribution to the recognition task will be removed. This step ensures that the model will not be disturbed by irrelevant information and improves the efficiency and accuracy of the model.

[0077] (5) Establishing feature space model

[0078] Based on the contributing features retained after screening, a feature space model is established. This model integrates all key features for the identification of the age of cotton bollworm. The construction of the model takes into account the interaction and combination between features, aiming to maximize the performance of the identification task.

[0079] Through this series of detailed steps, the process of constructing the feature vector and establishing the feature space model not only optimizes the feature set for identifying cotton bollworms, but also ensures the efficiency and accuracy of the model through precise feature selection, providing a solid foundation for achieving accurate identification of cotton bollworm ages.

[0080] Furthermore, the use of the feature space model to accurately identify the age of cotton bollworms and realize automatic classification includes training the feature space model, inputting the most valuable features into the feature space model, and attempting to identify the age of cotton bollworms. If the feature space model successfully identifies the age of cotton bollworms, that is, the recognition result is consistent with the actual situation, then the recognition process will continue to the next step and output the recognition result. If the feature space model fails to successfully identify the age of cotton bollworms, or the recognition result is inaccurate, then the model parameters will need to be adjusted to improve the recognition performance; and automatic classification of cotton bollworms is realized through the recognition result.

[0081] In the process of "identifying the age of cotton bollworm", the carefully selected and optimized feature vectors are first input into the trained feature space model. This model is constructed using key features extracted from cotton bollworm images, such as texture structure and morphological parameters, with the aim of accurately identifying the age of cotton bollworm. Through the analysis and calculation of the model, attempts are made to predict the age of the input features, thereby achieving automated classification of cotton bollworms.

[0082] The recognition process includes a key logical judgment step, which is to evaluate whether the recognition result of the model is successful. If the model accurately identifies the age of cotton bollworm, the recognition result will be output, completing the automated classification process. However, if the recognition is unsuccessful or the result is inaccurate, measures will be taken to adjust the model parameters, such as optimizing the learning rate, adjusting the algorithm, or retraining the model to improve the recognition performance of the model. After the adjustment, the model will retry the recognition process until the age of cotton bollworm is successfully identified. This process not only demonstrates the application of deep learning models in image recognition tasks, but also reflects the importance of continuous optimization of model performance in practical applications. In this way, accurate identification of cotton bollworm ages can be effectively achieved, providing strong technical support for the research and management of cotton bollworm.

[0083] (1) Input feature space model

[0084] Before starting the recognition process, the selected and optimized feature vectors are first input into the trained feature space model, which is built based on key features previously extracted from the cotton bollworm images, such as texture structure and morphological parameters.

[0085] (2) Applying the model to identify insect age

[0086] The feature space model analyzes and calculates the input feature vector to try to identify the age of the cotton bollworm. This step uses the model's learning ability to predict the age of the bollworm based on the feature data.

[0087] (3) Successful identification

[0088] After the model attempts to recognize an object, it needs to determine whether the recognition process was successful. This logical judgment step is based on comparing the output of the model with the expected results to determine the accuracy of the recognition.

[0089] 1) Yes: If the model successfully identifies the age of the cotton bollworm, that is, the identification result is consistent with the actual situation, then the identification process will continue to the next step and output the identification result.

[0090] 2) No: If the model fails to successfully identify the age of cotton bollworm, or the identification result is inaccurate, then the model parameters will need to be adjusted to improve the identification performance.

[0091] (4) Output the identified insect age

[0092] After the model successfully identifies the age of the cotton bollworm, it will output the identification results. This step realizes the automatic identification and classification of the age of the cotton bollworm, providing important information for further research and application.

[0093] (5) Readjust model parameters

[0094] If the identification is unsuccessful or the result is inaccurate, the parameters of the model will be readjusted. This may include adjusting the learning rate, optimizing the algorithm, or retraining the model. After the adjustment, the feature vector is input into the model again and the insect age identification attempt is made again.

[0095] Through this series of detailed steps, the age identification process of cotton bollworm not only includes the application of the model and the output of results, but also covers the judgment of whether the identification is successful or not and the model optimization measures in the case of unsuccessful identification, ensuring the accuracy and effectiveness of the identification process.

[0096] This application is an integrated method of automated acquisition and image enhancement: Developing and using an automated acquisition system to collect cotton bollworm images under diverse environmental conditions is the basis for building a high-quality, diverse dataset, ensuring the breadth and representativeness of the data. This method effectively improves the quality and diversity of the dataset, providing high-quality training and test data for deep learning models.

[0097] This application is based on UNet's image segmentation and morphological feature extraction technology: multi-scale detail enhancement technology is used to pre-process the collected images, improve image quality and highlight the key features of cotton bollworms, laying the foundation for subsequent image analysis and feature extraction. The advanced UNet network is used to accurately segment the image and extract the morphological parameters of cotton bollworms, which are important features for subsequent identification and classification. The application of this technology improves the accuracy of image segmentation and the effectiveness of feature extraction, providing key information for accurate identification.

[0098] The feature vector composition and feature selection process in this application: the extracted texture structure, texture features and other information are combined into a feature vector, and the most valuable features for identifying the age of cotton bollworm are selected through a feature selection algorithm. This process optimizes the input of the model and improves the accuracy of recognition. This process significantly improves the performance and accuracy of the recognition model by optimizing the feature set of the model input.

[0099] The construction and application method of the feature space model of the present application: a feature space model is established based on the screened features, and the model is used to identify the ages of cotton bollworms, thereby realizing the automatic classification of the ages of cotton bollworms. This method not only involves the model construction process, but also includes how to effectively apply the model to accurately identify the ages of cotton bollworms, which is the core technology for realizing automatic classification.

[0100] This application improves image quality through multi-scale detail enhancement technology, especially highlighting the key features of cotton bollworm, which not only improves the accuracy of subsequent model recognition, but also enhances the model's sensitivity to detailed features in the image. Compared with traditional image preprocessing methods, this targeted image enhancement can prepare data more effectively and provide richer and more accurate feature information for deep learning models.

[0101] This application uses the UNet network for image segmentation and morphological parameter extraction. Compared with the traditional CNN feature extraction method, UNet can more accurately segment the cotton bollworm image and extract more critical morphological parameters from it. This method not only improves the accuracy of feature extraction, but also can obtain more features closely related to the task, thereby improving the performance of the recognition model.

[0102] This application combines information such as texture structure and texture features extracted from the image into a feature vector. This application forms a more comprehensive feature vector by combining multiple types of features. Compared with the traditional method that only relies on a single type of feature, this method can more comprehensively describe the characteristics of cotton bollworms and provide richer input data for the recognition model.

[0103] The most valuable features for identifying the age of cotton bollworm were selected through feature selection algorithm, and the feature space model was established based on these features. This step optimized the input features of the model, reduced the interference of irrelevant features on the model performance, and improved the accuracy and generalization ability of the model.

[0104] This application uses a trained feature space model to accurately identify the age of cotton bollworms and realize automatic classification. This application can achieve high-accuracy cotton bollworm age identification through a carefully designed feature extraction, selection and model training process, providing a more accurate and reliable automatic classification solution compared to the existing technology.

[0105] Compared with the existing technology, this application not only improves the quality of data preparation and the accuracy of feature extraction, but also optimizes the input features and classification performance of the model by comprehensively applying technologies such as multi-scale detail enhancement, UNet network segmentation, feature vector construction, feature selection, and feature space model establishment, and achieves high-accuracy automatic identification of cotton bollworm ages. It has significant practical value in improving the efficiency of agricultural pest monitoring and management.

[0106] Although the present application is described above with reference to specific embodiments, it should be understood by those skilled in the art that many modifications may be made to the configurations and details disclosed in the present application within the principles and scope disclosed in the present application. The scope of protection of the present application is determined by the appended claims, and the claims are intended to cover all modifications included in the equivalent literal meaning or scope of the technical features in the claims.

Claims

1. A method for identifying the age of cotton bollworm, characterized in that: The method comprises collecting cotton bollworm images and constructing a cotton bollworm image dataset; Improving the image quality of the cotton bollworm and highlighting the key features of the cotton bollworm; The cotton bollworm image after quality improvement is accurately segmented, the cotton bollworm and its morphological parameters are extracted, and the morphological parameters are used as features; the features are converted into feature vectors; the most valuable features for identifying the age of cotton bollworm are screened out through the feature vectors, and a feature space model is established based on the most valuable features; the feature space model is used to accurately identify the age of cotton bollworm and realize automatic classification.

2. The method for identifying the age of cotton bollworm according to claim 1, characterized in that: The collecting of cotton bollworm images and constructing of a cotton bollworm image data set include starting an automated collection system, setting collection parameters and environmental conditions, collecting the cotton bollworm images under different environmental conditions, classifying and labeling the cotton bollworm images, selecting the labeled cotton bollworm images as a data set for training a machine learning model, and using the remaining labeled cotton bollworm images as a test set. The training data set and the test set constitute the cotton bollworm image data set.

3. The method for identifying the age of cotton bollworm according to claim 2, characterized in that: The method of improving the quality of the cotton bollworm image and highlighting the key features of the cotton bollworm includes adjusting the low-frequency details in the cotton bollworm image, enhancing the medium-frequency details in the cotton bollworm image, sharpening the high-frequency details in the cotton bollworm image, and combining the adjustment results of the low frequency, medium frequency and high frequency to obtain an improved cotton bollworm image at all scales, so that the key features of the cotton bollworm are highlighted.

4. The method for identifying the age of cotton bollworm according to claim 3, characterized in that: The image is accurately segmented, and the cotton bollworm and its morphological parameters are extracted as features, including using a UNet network to segment the cotton bollworm image after quality improvement, separating the cotton bollworm from the background, so that the outline and morphology of the cotton bollworm are clearly divided, extracting the morphological parameters of the cotton bollworm, and using the morphological parameters as features.

5. The method for identifying the age of cotton bollworm according to claim 4, characterized in that: The converting the features into feature vectors includes merging the features into a comprehensive feature vector.

6. The method for identifying the age of cotton bollworm according to claim 5, characterized in that: The most valuable features for identifying the age of cotton bollworm are screened out through the feature vectors, and a feature space model is established based on the most valuable features, including applying a feature selection algorithm to the merged feature vectors to screen out the most valuable features, and establishing a feature space model based on the most valuable features.

7. The method for identifying the age of cotton bollworm according to any one of claims 1 to 6, characterized in that: The use of the feature space model to accurately identify the age of cotton bollworms and realize automatic classification includes training the feature space model, inputting the most valuable features into the feature space model, and trying to identify the age of cotton bollworms. If the feature space model successfully identifies the age of cotton bollworms, that is, the recognition result is consistent with the actual situation, then the recognition process will continue to the next step and output the recognition result. If the feature space model fails to successfully identify the age of cotton bollworms, or the recognition result is inaccurate, then the model parameters will need to be adjusted to improve the recognition performance; and the automatic classification of cotton bollworms is realized through the recognition result.

8. The method for identifying the age of cotton bollworm according to claim 4, characterized in that: The morphological parameters include the length of the cotton bollworm, the width of the cotton bollworm, the area of ​​the cotton bollworm, the texture pattern on the surface of the cotton bollworm, the structural details of the cotton bollworm and the morphological characteristics of the cotton bollworm.

9. The method for identifying the age of cotton bollworm according to claim 8, characterized in that: The texture pattern, the structural details and the morphological features are combined into a feature vector, and the most valuable features for identifying the age of cotton bollworm are screened out through a feature selection algorithm.