Aloe disease grading system based on improved deeplab v3+ and method thereof

Abstract

The application discloses an aloe disease grading system and method based on an improved Deeplab v3+, which comprises a data acquisition module, a data preprocessing module, a semantic segmentation module, a performance estimation module, an input module, an identification module, a grading module, an output module and a database; the semantic segmentation module is used for feature extraction of aloe leaves and surface disease spots to obtain aloe leaf identification and disease spot segmentation results; the semantic segmentation module adopts a UAMT network framework, comprises a teacher model and a student model, and increases constraint on uncertain information in a "prediction consistency" part, so that when a consistency loss is calculated, only the results with small prediction deviation of the teacher model and the student model are learned, and the results with large prediction deviation are inhibited; the teacher model and the student model both adopt an improved Deeplab v3+ network; the application can realize segmentation and identification of surface diseases of aloe leaves and grading of disease degrees, and has high precision and efficiency.

Description

Technical Field

[0001] This invention belongs to the technical field of aloe vera disease segmentation, specifically relating to an aloe vera disease grading system and method based on an improved Deeplab v3+. Background Technology

[0002] Aloe vera is a succulent plant widely distributed throughout the world. It is not only edible but also has numerous medicinal and cosmetic applications, and its global market size continues to expand. However, aloe vera is also highly susceptible to diseases, with leaf spot and anthracnose being particularly prevalent, severely impacting yields. In modern agricultural production, timely identification and accurate classification of crop diseases are crucial for crop health and yield, especially in aloe vera cultivation. Disease management directly affects economic benefits and quality stability; failure to promptly detect, assess, and effectively control disease severity can lead to market losses. Therefore, the segmentation and severity classification of aloe vera diseases are of great significance for both agricultural production and market assessment.

[0003] Currently, traditional methods for disease identification mainly rely on manual observation and experience-based judgment. These methods suffer from high subjectivity, low efficiency, and low diagnostic accuracy. Furthermore, for large-scale disease surveys of aloe vera fields, traditional methods are not only extremely time-consuming and labor-intensive but also prone to errors, making it difficult to achieve the desired results. Summary of the Invention

[0004] The main objective of this invention is to overcome the shortcomings and deficiencies of existing technologies and provide an aloe vera disease grading system and method based on an improved Deeplabv3+. The improved Deeplabv3+ semantic segmentation algorithm is used to segment aloe vera leaves and their surface lesions, realizing the segmentation and identification of diseases on the surface of aloe vera leaves and the grading of their severity. It has high accuracy and efficiency, and can realize the identification and grading research of aloe vera diseases, effectively solving the shortcomings of manual identification.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] This invention provides an aloe vera disease grading system based on an improved Deeplab v3+, including a data acquisition module, a data preprocessing module, a semantic segmentation module, a performance estimation module, an input module, a recognition module, a grading module, an output module, and a database;

[0007] The data acquisition module is used to acquire data of aloe vera leaves taken from above or from the side.

[0008] The data preprocessing module preprocesses the collected aloe vera data video to obtain an image dataset, and annotates the aloe vera leaves and their leaf surface lesions. Finally, the dataset is divided.

[0009] The semantic segmentation module is used to extract features from aloe vera leaves and surface lesions to obtain aloe vera leaf recognition and lesion segmentation results. The semantic segmentation module adopts the UAMT network framework, which includes a teacher model (Tnet) and a student model (Snet). In the "prediction consistency" part, constraints on uncertain information are added so that when calculating the consistency loss, only results with small prediction deviations between the teacher model (Tnet) and the student model (Snet) are learned, while results with large prediction deviations are suppressed. The teacher model (Tnet) and the student model (Snet) have the same structure, both using an improved Deeplab v3+ network. The UAMT network framework uses an improved Deeplab v3+ network for deep learning during the aloe vera plant detection process. By evaluating the uncertainty in the teacher network's prediction process, the confidence of the network output is improved. The lightweight and efficient MobileNetV2 architecture is used to optimize the model's backbone network, and a multi-scale attention module and CBAM attention mechanism are introduced into the backbone network to improve the efficiency and real-time performance of the semantic segmentation module.

[0010] The performance estimation module is used to evaluate the detection results, test the MIoU, MPA and accuracy of the detection results. If they are lower than the set threshold, the module returns to the semantic segmentation module to improve the scheme and re-extract features.

[0011] The input module is used to input the image information required for aloe vera disease detection into the model;

[0012] The recognition module is used to identify and segment aloe vera leaves and their surface lesions in the input video or image;

[0013] The grading module is used to calculate the segmentation results of the identification module and to grade the degree of aloe vera disease.

[0014] The output module is used to output and present the disease identification and grading data to the terminal display device, and at the same time output the weighted calculation results.

[0015] The database is used to store the characteristics of aloe vera leaves and surface lesions.

[0016] As a preferred technical solution, the specific steps for obtaining aloe vera leaf data by overhead or side-view photography are as follows:

[0017] Use drones or handheld devices to take photos and obtain video data of aloe vera leaves from above or the side.

[0018] As a preferred technical solution, the preprocessing is mean normalization, and the feature scaling formula is:

[0019]

[0020] In the formula, x′ is the value after being normalized by average, and x is the original value; max(x) represents the maximum value of the sample data, and min(x) represents the minimum value of the sample data.

[0021] As a preferred technical solution, the improved Deeplab v3+ network includes an encoder and a decoder. The Multi-Scale Attention Module (MSA) introduces the Spatial Channel Attention Module (CBAM) on features at different scales, and then fuses the enhanced feature representations at different scales to further enhance the representational capability of feature maps at different scales. The Deeplab v3+ network selects layers 4 and 8 of the backbone network MobilenetV2 as key layers for feature extraction. These two layers correspond to 1 / 4 and 1 / 16 scales of the input image, respectively, with initial channel numbers of 24 and 64, effectively capturing features at different levels, from fine texture to high-level semantics. To ensure the consistency of feature maps in spatial dimension during subsequent processing, the smaller layer 8 feature map is first upsampled by 4 times to match the spatial resolution of the layer 4 feature map. Then, the sampled feature maps are input into the CBAM module. Finally, the two feature maps enhanced by the CBAM module are effectively combined. The combined feature map contains feature information at different levels and has richer semantic and spatial information.

[0022] As a preferred technical solution, the CBAM module can adaptively adjust the feature maps in the convolutional neural network. Given an intermediate feature map, the CBAM module calculates attention maps along both the channel and spatial dimensions, and then applies these attention maps to the input feature map for adaptive feature improvement.

[0023] As a preferred technical solution, the performance estimation module compares the received data with the feature extraction results of the target detection module. The results include four parts: TP is positive and is determined to be positive; TN is negative and is determined to be negative; FN is positive and is determined to be negative; and FP is negative and is determined to be positive.

[0024] MIoU, MPA, and accuracy were selected as evaluation metrics for the algorithm.

[0025] The formula for calculating MIoU is:

[0026] The formula for calculating MPA is:

[0027] The formula for calculating accuracy is:

[0028] The closer all evaluation metrics are to 1, the better the results.

[0029] Secondly, the present invention provides a method for grading aloe vera diseases based on an improved Deeplab v3+, comprising the following steps:

[0030] S1. Obtain data of aloe vera leaves by overhead or side view;

[0031] S2. Preprocess the captured video data. Use a Python program to capture a screenshot every 0.5-1 second, select usable images, and randomly distribute them into training and test sets according to a set ratio.

[0032] S3. Use the Labelme tool to label the aloe vera objects in the experimental data, convert the output JSON file into VOC format PNG images for storage, and classify the different colors of the pixels in the images to represent healthy aloe vera leaves, disease spots, and filtered backgrounds.

[0033] S4. Put the labeled data into the semantic segmentation module, perform deep learning on the improved Deeplab v3+, and use MIoU, MPA and accuracy as evaluation indicators of the algorithm. If the evaluation indicators are less than the set threshold, the improved scheme is to retrain the detection model with the incorrectly identified data and store the corrected deep learning results in the model library, thereby achieving the segmentation and detection effect of aloe leaf diseases.

[0034] S5. Users can input images of aloe vera leaves to be detected through the visualization platform. The model can segment the lesions on the surface of the aloe vera leaves and, based on the calculation of the proportion of the lesion area to the total leaf area, achieve the classification study of aloe vera diseases.

[0035] As a preferred technical solution, step S1 specifically includes:

[0036] Operate a drone to fly above the aloe vera field at a height of 1.5 to 2.5 meters, with the camera tilt angle set to be perpendicular to the ground, and take aerial photos of the aloe vera field at a horizontal flight speed of less than 1 meter per second to record video data; or manually walk through the aloe vera field and use a mobile phone to take direct photos of the surface of the aloe vera leaves, with shooting angles including shooting vertically downwards from above and shooting obliquely downwards at a 45-degree angle from the side.

[0037] As a preferred technical solution, in step S4, deep learning is performed on the improved Deeplab v3+, specifically as follows:

[0038] S41. The labeled data is put into the semantic segmentation module, which is based on the improved Deeplab v3+ model;

[0039] S42 and DeepLab employ a Convolutional Neural Network (CNN) architecture, based on improvements to a selected standard network, and introduce dilated convolutions, enabling the network to extract features at different scales without losing resolution.

[0040] S43. During training, DeepLab uses the cross-entropy loss function to evaluate the difference between the prediction and the true label, and the model will continuously adjust the parameters based on the feedback from the loss.

[0041] S44. After training, the performance of the model is evaluated using a test set. Typically, metrics such as Intersection over Union (IoU) and pixel accuracy are used to measure the segmentation quality of the model.

[0042] As a preferred technical solution, in step S4, the threshold is 0.9.

[0043] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0044] This invention provides a deep learning-based method for classifying the severity of aloe vera diseases. It employs a UAMT semi-supervised learning architecture and utilizes the existing Deeplab v3+ deep network learning architecture for aloe vera plant detection. By assessing the uncertainty in the teacher network's prediction process, the confidence of the network output is improved. The lightweight and efficient MobileNetV2 architecture is used to optimize the model's backbone network, and a multi-scale attention module and CBAM attention mechanism are introduced into the backbone, enhancing model efficiency and real-time performance. This method exhibits strong learning effectiveness, with accuracy and efficiency exceeding those of conventional manual observation and subjective judgment.

[0045] This invention combines images automatically acquired by drones to segment aloe vera leaves and their surface lesions, and calculates the area and proportion of both, thereby analyzing the results of aloe vera disease. It has high pixel accuracy and F1 value, and can realize disease monitoring and disease severity rating of aloe vera growth status. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0047] Figure 1 This is a structural diagram of the aloe vera disease grading system based on the improved Deeplab v3+ of this invention;

[0048] Figure 2 This is a diagram of the UAMT network architecture of the present invention;

[0049] Figure 3 This is a structural diagram of the CBAM attention mechanism module of the present invention;

[0050] Figure 4 This is a flowchart of the multi-scale attention module and Deeplabv3+ network of the present invention. Detailed Implementation

[0051] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of the present application, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present application without creative effort are within the scope of protection of the present application.

[0052] In this application, the reference to "embodiment" means that a specific feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described in this application can be combined with other embodiments.

[0053] like Figure 1 As shown, this embodiment provides an aloe vera disease grading system based on an improved Deeplab v3+, including a data acquisition module, a data preprocessing module, a semantic segmentation module, a performance estimation module, an input module, a recognition module, a grading module, an output module, and a database;

[0054] The data acquisition module is used to acquire data of aloe vera leaves taken from above or from the side.

[0055] The data preprocessing module preprocesses the collected aloe vera data video to obtain an image dataset, and annotates the aloe vera leaves and their leaf surface lesions. Finally, the dataset is divided.

[0056] The semantic segmentation module is used to extract features from aloe vera leaves and surface lesions to obtain aloe vera leaf recognition and lesion segmentation results. The semantic segmentation module adopts the UAMT network framework, which includes a teacher model Tnet and a student model Snet. In the "prediction consistency" part, constraints on uncertain information are added so that when calculating the consistency loss, only the results with small prediction deviations between the teacher model Tnet and the student model Snet are learned, while the results with large prediction deviations are suppressed. The teacher model Tnet and the student model Snet have the same structure and both adopt the improved Deeplab v3+ network.

[0057] The performance estimation module is used to evaluate the detection results, test the MIoU, MPA and accuracy of the detection results. If they are lower than the set threshold, the module returns to the semantic segmentation module to improve the scheme and re-extract features.

[0058] The input module is used to input the image information required for aloe vera disease detection into the model;

[0059] The recognition module is used to identify and segment aloe vera leaves and their surface lesions in the input video or image;

[0060] The grading module is used to calculate the segmentation results of the identification module and to grade the degree of aloe vera disease.

[0061] The output module is used to output and present the disease identification and grading data to the terminal display device, and at the same time output the weighted calculation results.

[0062] The database is used to store the characteristics of aloe vera leaves and surface lesions.

[0063] Furthermore, in the data acquisition module, acquiring aloe vera leaf data by overhead or side view specifically involves:

[0064] By using drones for aerial photography or handheld devices such as smartphones and tablets to photograph aloe vera plants from the ground, image data of aloe vera leaves can be collected. These images can be still photographs or dynamic video footage, aiming to comprehensively record the morphological characteristics of aloe vera leaves from different angles for subsequent analysis of their growth status, health, and other information. This method allows for efficient monitoring of large-scale aloe vera cultivation areas and enables regular data acquisition to track changes in aloe vera growth.

[0065] In a more specific embodiment, a drone is operated to fly above the aloe vera field at a height of 1.5 to 2.5 meters, with the camera tilt angle set to be perpendicular to the ground, to take a bird's-eye view of the aloe vera field and record video data at a horizontal flight speed of less than 1 meter per second; or a person walks through the aloe vera field and uses a mobile phone to take direct photos of the aloe vera leaf surface, with shooting angles including shooting vertically downwards from above and shooting obliquely downwards at a 45-degree angle from the side.

[0066] Furthermore, in the data preprocessing module, the preprocessing is mean normalization, and the feature scaling formula is:

[0067]

[0068] In the formula, x′ is the value after being normalized by average, and x is the original value; max(x) represents the maximum value of the sample data, and min(x) represents the minimum value of the sample data.

[0069] Furthermore, in the semantic segmentation module, the UAMT network framework uses an improved Deeplab v3+ network for deep learning during the aloe vera plant detection process. By evaluating the uncertainty in the teacher network prediction process, the confidence of the network output is improved. The lightweight and efficient MobileNetV2 architecture is used to optimize the model backbone network, and a multi-scale attention module and CBAM attention mechanism are introduced into the backbone network to improve the efficiency and real-time performance of the semantic segmentation module.

[0070] Furthermore, such as Figure 2 As shown, the Uncertainty-Aware Mean Teacher (UAMT) network architecture, while maintaining the core idea of ​​the Mean Teacher framework, adds constraints on uncertain information to the "prediction consistency" part. Adding these constraints allows the consistency loss to be calculated by learning only from results where the predictions of the two networks have small discrepancies, while suppressing results with large discrepancies. Therefore, Snet is optimized and receives more reliable supervision, which in turn encourages Tnet to make higher-quality predictions.

[0071] The teacher model Tnet and the student model Snet have the same structure, both based on the improved DeeplabV3+. Labeled data XL is input into Snet, and the output ysL is used for supervised learning with the label Y. The loss function is the cross-entropy loss function, denoted as Lsup. Unlabeled data XU is input into Snet and Tnet, and predictions for ysU and ytU are made respectively. The consistency loss between the two is denoted as Lcon.

[0072] like Figure 3As shown, the CBAM module can adaptively adjust the feature maps in a convolutional neural network. Given an intermediate feature map, the CBAM module computes attention maps along both the channel and spatial dimensions, and then applies these attention maps to the input feature map for adaptive feature improvement.

[0073] like Figure 4 As shown, the Multi-scale Attention (MSA) module introduces a Spatial Channel Attention (CBAM) module on features at different scales, and then fuses enhanced feature representations at different scales to further enhance the representational power of feature maps at different scales. Specifically, it selects layers 4 and 8 of the backbone network MobileNetV2 (17 layers) as key layers for feature extraction. These two layers correspond to 1 / 4 and 1 / 16 scales of the input image, respectively, with initial channel numbers of 24 and 64, effectively capturing features at different levels from fine texture to high-level semantics. To ensure the consistency of feature maps in spatial dimension in subsequent processing, the smaller layer 8 feature map (1 / 16 scale) is first upsampled by 4 times to match the spatial resolution of the layer 4 feature map (1 / 4 scale), and then input into the CBAM module. Finally, the two feature maps enhanced by the CBAM module are effectively combined, containing feature information at different levels and possessing richer semantic and spatial information.

[0074] Furthermore, in the performance estimation module, the received data is compared with the feature extraction results of the target detection module. The results include four parts: TP is positive and is determined to be positive; TN is negative and is determined to be negative; FN is positive and is determined to be negative; and FP is negative and is determined to be positive.

[0075] MIoU, MPA, and accuracy were selected as evaluation metrics for the algorithm.

[0076] The formula for calculating MIoU is:

[0077] The formula for calculating MPA is:

[0078] The formula for calculating accuracy is:

[0079] The closer all evaluation metrics are to 1, the better the results.

[0080] Further improvements include, but are not limited to: changing the ratio of teacher and student model labels in the Mean teacher model; changing the attention mechanism; using other loss functions, etc.; this paper only provides one preferred solution for aloe vera lesion segmentation.

[0081] In another embodiment of this application, an aloe vera disease grading method based on an improved Deeplab v3+ is provided, the method comprising the following steps:

[0082] S1. Operate a drone to fly above the aloe vera field at a height of 1.5 to 2.5 meters, with the camera angle set to be perpendicular to the ground. Take a bird's-eye view of the aloe vera field and record video data at a horizontal flight speed of less than 1 meter per second. Alternatively, manually walk through the aloe vera field and use a mobile phone to take direct photos of the aloe vera leaf surface. The shooting angles include shooting vertically downwards from above and shooting at a 45-degree angle from the side. You can choose one of the two options or do both at the same time.

[0083] S2. Preprocess the captured video data. Use a Python program to capture a screenshot every 0.5-1 second. Manually select usable images and randomly distribute them into training and test sets in a 9:1 ratio.

[0084] S3. Use the Labelme tool to label the aloe vera objects in the experimental data, convert the output JSON file into VOC format PNG images for storage, and classify the different colors of the pixels in the images to represent healthy aloe vera leaves, disease spots, and filtered backgrounds.

[0085] S4. Place the labeled data into the semantic segmentation module and perform deep learning on the improved Deeplab v3+. Use MIoU, MPA, and accuracy as evaluation metrics for the algorithm. If the evaluation metrics are less than 0.9, then the improved scheme is adopted. The erroneously identified data is retrained on the detection model, and the deep learning results are corrected and stored in the model library, thereby achieving the segmentation and detection effect of aloe vera leaf diseases.

[0086] Furthermore, step S4 specifically involves:

[0087] S41. Place the labeled data into the semantic segmentation module, which is based on the improved Deeplab v3+ model.

[0088] S42 and DeepLab employ a convolutional neural network (CNN) architecture, typically based on improvements to certain standard networks (such as Mobilenet or Xception), and introduce atrous convolution, which allows the network to extract features at different scales without losing resolution.

[0089] S43. During training, DeepLab uses the cross-entropy loss function to evaluate the difference between the prediction and the true label. The model continuously adjusts its parameters based on the feedback from the loss.

[0090] S44. After training, the performance of the model is evaluated using a test set. Metrics such as Intersection over Union (IoU) and pixel accuracy are typically used to measure the segmentation quality of the model.

[0091] S45. Users can input the images of aloe vera leaves to be detected through the visualization platform. The model can segment the lesions on the surface of the aloe vera leaves and, based on the calculation of the proportion of the lesion area to the total leaf area, realize the classification study of aloe vera diseases.

[0092] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and RAMbus dynamic RAM (RDRAM), etc.

[0093] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0094] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments. Any changes, modifications, substitutions, combinations, or simplifications made without departing from the spirit and principle of the present invention shall be considered equivalent substitutions and shall be included within the protection scope of the present invention.

Claims

1. An aloe vera disease grading system based on an improved Deeplab v3+, characterized in that, It includes a data acquisition module, a data preprocessing module, a semantic segmentation module, a performance estimation module, an input module, a recognition module, a classification module, an output module, and a database; The data acquisition module is used to acquire data of aloe vera leaves taken from above or from the side. The data preprocessing module preprocesses the collected aloe vera data video to obtain an image dataset, and annotates the aloe vera leaves and their leaf surface lesions, and divides the dataset. The semantic segmentation module is used to extract features from aloe vera leaves and surface lesions, obtaining aloe vera leaf recognition and lesion segmentation results. The semantic segmentation module uses the UAMT network framework, which includes a teacher model (Tnet) and a student model (Snet). The teacher model (Tnet) and the student model (Snet) have the same structure, both using an improved Deeplab v3+ network. The UAMT network framework uses the improved Deeplab v3+ network for aloe vera plant detection. The MobileNetV2 architecture is used to optimize the model's backbone network. The improved Deeplab v3+ network includes an encoder and a decoder, and the Multi-Scale Attention Module (MSA) introduces a Spatial Channel Attention Module (CBAM) on features at different scales. The Deeplab v3+ network selects layers 4 and 8 of the MobileNetV2 backbone network as key layers for feature extraction. These two layers correspond to 1 / 4 and 1 / 16 scales of the input image, respectively, with initial channel numbers of 24 and 64. The feature map of layer 8 is first upsampled by 4 times to match the feature map of layer 4. The spatial resolution of the layer feature maps is determined, and the sampled feature maps are then input into the CBAM module. Finally, the two feature maps enhanced by the CBAM module are combined together. The performance estimation module is used to evaluate the detection results, test the MIoU, MPA and accuracy of the detection results. If they are lower than the set threshold, the system returns to the semantic segmentation module to re-extract features. The input module is used to input the image information required for aloe vera disease detection into the model; The recognition module is used to identify and segment aloe vera leaves and their surface lesions in the input video or image; The grading module is used to calculate the segmentation results of the identification module and to grade the degree of aloe vera disease. The output module is used to output and present the disease identification and grading data to the terminal display device, and at the same time output the weighted calculation results. The database is used to store the characteristics of aloe vera leaves and surface lesions.

2. The aloe vera disease grading system based on the improved Deeplab v3+ as described in claim 1, characterized in that, The specific steps for obtaining aloe vera leaf data by overhead or side-view photography are as follows: Use drones or handheld devices to take photos and obtain video data of aloe vera leaves from above or the side.

3. The aloe vera disease grading system based on the improved Deeplab v3+ as described in claim 1, characterized in that, The preprocessing is mean normalization, and the feature scaling formula is: ； In the formula, The value after being normalized by average. x The original value; max ( x () represents the maximum value of the sample data. min ( x ) represents the minimum value of the sample data.

4. The aloe vera disease grading system based on the improved Deeplab v3+ as described in claim 1, characterized in that, The CBAM module can adaptively adjust the feature maps in the convolutional neural network. Given an intermediate feature map, the CBAM module calculates attention maps along both channel and spatial dimensions, and then applies these attention maps to the input feature map for adaptive feature improvement.

5. The aloe vera disease grading system based on the improved Deeplab v3+ as described in claim 1, characterized in that, The performance estimation module compares the received data with the feature extraction results of the target detection module. The results include four parts: TP is positive and is determined to be positive; TN is negative and is determined to be negative; FN is positive and is determined to be negative; and FP is negative and is determined to be positive. MIoU, MPA, and accuracy were selected as evaluation metrics for the algorithm. The formula for calculating MIoU is: ; The formula for calculating MPA is: ; The formula for calculating accuracy is: .

6. A method for classifying aloe vera diseases based on an improved Deeplab v3+, applied to the aloe vera disease classification system based on an improved Deeplab v3+ as described in claim 1, characterized in that, Includes the following steps: S1. Obtain data of aloe vera leaves by overhead or side view; S2. Preprocess the captured video data. Use a Python program to capture a screenshot every 0.5-1 second, select usable images, and randomly distribute them into training and test sets according to a set ratio. S3. Use the Labelme tool to label the aloe vera objects in the experimental data, convert the output JSON file into VOC format PNG images for storage, and classify the different colors of the pixels in the images to represent healthy aloe vera leaves, disease spots, and filtered backgrounds. S4. Put the labeled data into the semantic segmentation module, perform deep learning on the improved Deeplab v3+, and use MIoU, MPA and accuracy as evaluation indicators of the algorithm. If the evaluation indicators are less than the set threshold, the improved scheme is to retrain the detection model with the incorrectly identified data and store the corrected deep learning results in the model library, thereby achieving the segmentation and detection effect of aloe leaf diseases. S5. Users can input images of aloe vera leaves to be detected through the visualization platform. The model can segment the lesions on the surface of the aloe vera leaves and, based on the calculation of the proportion of the lesion area to the total leaf area, achieve the classification study of aloe vera diseases.

7. The aloe vera disease classification method based on the improved Deeplab v3+ according to claim 6, characterized in that, Step S1 is as follows: Operate a drone to fly above the aloe vera field at a height of 1.5 to 2.5 meters, with the camera tilt angle set to be perpendicular to the ground, and take aerial photos of the aloe vera field at a horizontal flight speed of less than 1 meter per second to record video data; or manually walk through the aloe vera field and use a mobile phone to take direct photos of the surface of the aloe vera leaves, with shooting angles including shooting vertically downwards from above and shooting obliquely downwards at a 45-degree angle from the side.

8. The aloe vera disease classification method based on the improved Deeplab v3+ according to claim 7, characterized in that, In step S4, deep learning is performed on the improved Deeplab v3+, specifically as follows: S41. The labeled data is put into the semantic segmentation module, which is based on the improved Deeplab v3+ model; S42 and DeepLab employ a Convolutional Neural Network (CNN) architecture, based on improvements to a selected standard network, and introduce dilated convolutions, enabling the network to extract features at different scales. S43. During training, DeepLab uses the cross-entropy loss function to evaluate the difference between the prediction and the true label. The model will continuously adjust the parameters based on the feedback from the loss. S44. After training, the performance of the model is evaluated using a test set. Typically, metrics such as Intersection over Union (IoU) and pixel accuracy are used to measure the segmentation quality of the model.

9. The aloe vera disease classification method based on the improved Deeplab v3+ according to claim 6, characterized in that, In step S4, the threshold is 0.9.

Images