Wearable crop disease and insect pest intelligent identification device and method

Abstract

The invention discloses a wearable crop disease and insect pest intelligent identification device and method. According to the method, crop environment parameters and cloud pest and disease damage images are utilized to carry out low-illumination enhancement processing on pest and disease damage crop image features collected by wearable equipment, and pest and disease damage identification features are obtained based on a data enhancement fusion method in combination with crop growth features; constructing a wearable crop disease and insect pest recognition model by using the disease and insect pest recognition features and the disease and insect pest classification evaluation result; a neural network model is constructed based on disease and insect pest crop image features collected and processed by a wearable device in combination with crop growth features, so that highly structured feature data is efficiently processed, disease and insect pest image data is processed by using cloud insect pest data, more suitable insect pest recognition features are obtained, and the disease and insect pest recognition accuracy is improved. The trained model is lighter in weight and stronger in robustness, so that crop disease and insect pest recognition is more objective, a subsequent disease and insect pest treatment method is convenient to put forward, and the agricultural production efficiency is improved.

Description

Technical Field

[0001] This invention belongs to the field of emerging agricultural software and new information technology service technology, and in particular relates to a wearable intelligent identification device and method for crop diseases and pests. Background Technology

[0002] Traditional agricultural monitoring relies on manual inspections, drone inspections, and remote sensing satellites, which are inefficient and require a high level of expertise. Existing intelligent identification devices mostly adopt a fixed camera + cloud processing mode, which suffers from problems such as strong network dependence, high latency, and high power consumption.

[0003] Existing technologies indicate that, on the one hand, current detection of crop diseases and pests relies on manual operation of equipment, with the equipment used to inspect crops. However, when analysis of crop diseases and pests is required, it is still done manually, which is time-consuming, inefficient, and prone to errors. On the other hand, the use of models for crop disease and pest assessment lacks a clear direction for improvement, such as utilizing wearable devices to comprehensively optimize image features and combining data from the entire crop growth cycle to adaptively construct lightweight neural network models and refine the models to obtain more objective intelligent identification results for crop diseases and pests.

[0004] Therefore, further research is needed on how to construct a neural network model based on crop growth characteristics and the features of crop images collected and processed by wearable devices, thereby efficiently processing highly structured feature data and using cloud-based pest data to process pest image data to derive more suitable pest identification features; how to make the trained model more lightweight and robust, making crop pest identification more objective, so as to improve agricultural production efficiency in the subsequent development of pest control methods; and what image processing methods to use to solve the problem of image blurring in low light conditions, so that the pest identification features obtained by the data enhancement and fusion of the subsequent input model training data can make the trained model more closely resemble mobile intelligent identification of crop pests. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention proposes a wearable intelligent identification device and method for crop diseases and pests.

[0006] In a first aspect of the present invention, a wearable intelligent identification method for crop diseases and pests is provided, the method comprising:

[0007] N1. Collect and process crop environmental parameters, and obtain crop image features of pests and diseases based on wearable devices. The crop data storage module stores cloud-based image data of pests and diseases and crop growth characteristics, and obtains corresponding pest classification and assessment results.

[0008] N2. Based on the crop environmental parameters and the cloud-based pest and disease image data, perform low-light enhancement processing on the pest and disease crop image features to obtain optimized pest and disease crop image features;

[0009] N3. Based on the optimized image features of the pest-infested crops and the crop growth features, pest identification features are obtained using a data augmentation fusion method.

[0010] N4. Based on the pest identification features and the pest classification evaluation results, input them into a lightweight neural network model to construct a wearable crop pest identification model.

[0011] Furthermore, the crop environmental parameters are obtained by processing soil environmental parameters, meteorological environmental parameters, and water quality parameters during the actual growth cycle of the crop.

[0012] Furthermore, the process of performing low-light enhancement processing on the crop disease and pest image features based on the crop environmental parameters and the cloud-based pest and pest image data to obtain optimized crop disease and pest image features is specifically calculated using an improved light reflection decomposition formula.

[0013] Furthermore, the crop growth characteristics include leaf area index, leaf thickness, and stem diameter.

[0014] Furthermore, the data enhancement and fusion method specifically refers to obtaining the optimized image features of the diseased and pest-infested crops and the crop growth features by matrix vectorization using three-dimensional feature vectors.

[0015] Furthermore, the wearable crop pest and disease identification model adopts a lightweight neural network model based on improved crop growth characteristics.

[0016] Furthermore, the lightweight neural network model utilizes the activation function of the lightweight neural network model modified based on crop growth characteristics, and uses the activation function value to perform lightweight modification of the input value.

[0017] A wearable intelligent identification device for crop diseases and pests is also provided. This device implements a wearable intelligent identification method for crop diseases and pests, including a crop environmental parameter acquisition module, a multispectral imaging module, an edge computing module, a wearable crop disease and pest identification model construction module, and an intelligent identification module for crop diseases and pests.

[0018] The crop environmental parameter acquisition module is used to collect and process crop environmental parameters.

[0019] The multispectral imaging module acquires and processes image features of pests and diseases in crops based on wearable devices.

[0020] The edge computing module includes a crop data storage module, which stores cloud-based pest and disease image data and crop growth characteristics, obtains corresponding pest classification and evaluation results, and performs low-light enhancement processing on the pest and disease crop image features based on the crop environmental parameters and the cloud-based pest and disease image data to obtain optimized pest and disease crop image features; furthermore, based on the optimized pest and disease crop image features and the crop growth characteristics, it obtains pest identification features using a data augmentation fusion method.

[0021] The wearable crop pest and disease identification model construction module: Based on the pest identification features and the pest classification evaluation results, the wearable crop pest and disease identification model is constructed by inputting them into a lightweight neural network model.

[0022] The intelligent crop disease and pest identification module: Based on the wearable crop disease and pest identification model, it performs crop disease and pest identification and evaluation on the real-time crop disease and pest image features captured by personnel who are wearing or re-wearing the wearable device.

[0023] Furthermore, the process of performing low-light enhancement processing on the crop disease and pest image features based on the crop environmental parameters and the cloud-based pest and pest image data to obtain optimized crop disease and pest image features is specifically calculated using an improved light reflection decomposition formula.

[0024] Furthermore, the lightweight neural network model utilizes the activation function of the lightweight neural network model modified based on crop growth characteristics, and uses the activation function value to perform lightweight modification of the input value.

[0025] Therefore, the beneficial effects of this invention are as follows: a lightweight neural network model is constructed based on crop growth characteristics combined with crop image features collected and processed by wearable devices. By pruning the neural networks of the hidden layers of the activation function, highly structured feature data is processed efficiently. Furthermore, by utilizing high-quality crop pest and disease data from the cloud to process and correct the timely collected pest and disease image data, more suitable pest identification features are obtained. This makes the trained model more lightweight and robust, resulting in more objective identification of crop pests and diseases. This facilitates the development of subsequent pest and disease control methods, improves agricultural production efficiency, and solves the problem of image blurring in low-light conditions for pest and disease images. The improved low-light enhancement image processing method solves the problem of image blurring in pest and disease images, enabling the pest identification features obtained by enhancing and fusing the data input to the model training to make the trained model more closely resemble mobile intelligent identification of crop pests and diseases.

[0026] Further embodiments and improvements of the present invention will be described in conjunction with the accompanying drawings and specific examples. Attached Figure Description

[0027] Figure 1 This is a flowchart of a wearable intelligent identification method for crop diseases and pests according to the present invention;

[0028] Figure 2 This is a schematic diagram of a wearable intelligent identification device for crop diseases and pests according to the present invention.

[0029] Figure 3 This is an example diagram of the multispectral imaging module of the wearable device used in the embodiments of the present invention;

[0030] Figure 4 This is a schematic diagram of the activation function of the neural network model in the embodiments of the present invention;

[0031] Figure 5 This is a schematic diagram of an electronic device structure for implementing the method of the present invention in an embodiment of the present invention. Detailed Implementation

[0032] The invention will now be further described with reference to the accompanying drawings and specific embodiments. The neural network model used in this invention is an improved version of the neural network model, employing an edge-cloud collaborative architecture: real-time recognition using a local lightweight model and an incremental update mechanism for the cloud model. It also incorporates multimodal data fusion and algorithms for correcting image visual data and surrounding environmental parameters.

[0033] like Figure 2 As shown, the device of the present invention belongs to the category of agricultural Internet of Things application services, and therefore belongs to emerging software and new information technology services, as well as agricultural big data services utilizing Internet platforms.

[0034] In a first aspect of the present invention, a wearable intelligent identification method for crop diseases and pests is provided, the method comprising:

[0035] N1. Collect and process crop environmental parameters, and obtain crop image features of pests and diseases based on wearable devices. The crop data storage module stores cloud-based image data of pests and diseases and crop growth characteristics, and obtains corresponding pest classification and assessment results.

[0036] N2. Based on the crop environmental parameters and the cloud-based pest and disease image data, perform low-light enhancement processing on the pest and disease crop image features to obtain optimized pest and disease crop image features;

[0037] N3. Based on the optimized image features of the pest-infested crops and the crop growth features, pest identification features are obtained using a data augmentation fusion method.

[0038] N4. Based on the pest identification features and the pest classification evaluation results, input them into a lightweight neural network model to construct a wearable crop pest identification model.

[0039] Furthermore, the crop environmental parameters are obtained by processing soil environmental parameters, meteorological environmental parameters, and water quality parameters during the actual growth cycle of the crop: ;

[0040] In the formula, Here, D represents the crop's environmental parameters, and D represents the actual growth cycle of the crop. Let be the average soil moisture of the crop on day i. The light intensity for crops on day i. The optimal humidity for crop growth. For optimal light intensity for crop growth, Let be the average soil temperature of the crop on day i. The optimal soil temperature for crop growth. Let be the average meteorological temperature of the crop on day i. The optimal weather temperature for crop growth For average per The pH level of the irrigation water.

[0041] The value is determined based on the type of crop, and generally ranges from 1 to 5 days. In this embodiment, 3 days are used, so irrigation is carried out every 3 days to obtain the preferred pH value among the irrigation water quality parameters at this time.

[0042] Based on actual research, this application innovatively integrates soil, meteorological, and water quality factors that affect pests and diseases during the actual growth cycle. The application optimizes the influencing factors and transforms them into data representations based on objective facts. The difference between the data of this category and the optimal growth environment parameters of the corresponding crop is used to identify crop pests and diseases. This reduces the amount of uninterpretable data collected by conventional neural network models while making the training of the subsequent lightweight neural network model more accurate and reliable.

[0043] Because research has found that different crops are susceptible to different pests and diseases at different times during the identification process, and are greatly affected by pests and diseases at different times, in order to meet the needs of precision agricultural pest and disease identification, this invention uses the average data of all-weather cycle deviation of key crops in soil, meteorology and water quality to refine the actual crop images of pests and diseases, so as to obtain more intense and adaptable crop image features of pests and diseases, so as to facilitate the rapid and efficient processing of low-dimensional data by the subsequent lightweight neural network model.

[0044] Furthermore, the process of performing low-light enhancement processing on the crop pest and disease image features based on the crop environmental parameters and the cloud-based pest and disease image data to obtain optimized crop pest and disease image features specifically employs an improved light reflection decomposition formula for calculation: ;

[0045] In the formula, x and y are the coordinates of the image pixels. To optimize image features for crops affected by pests and diseases at corresponding coordinates, where m represents the amount of pest and disease image data in the cloud, low-light enhancement processing is performed by comparing the features of the collected pest and disease images with those in the cloud database. This allows for more accurate acquisition of the reflectance after low-light enhancement. This refers to the maximum crop environmental parameter obtained during the crop growth cycle. Based on the minimum crop environmental parameters obtained during the crop growth cycle and their impact on crop diseases and pests, this invention utilizes the deviation values ​​of these parameters to correct the reflectance under low-light enhancement. This is a creative improvement discovered in this application, which can make crop images of diseases and pests more closely resemble real-world image features under the low-light enhancement method. The image features of the diseased and pest-infested crops are obtained at the corresponding coordinates. The Gaussian filter kernel function is the general form of Gaussian filter kernel function adopted in this field. It is usually in the form of an exponential function and is a conventional image filtering method for those skilled in the art, so it will not be described in detail here.

[0046] The core of the illumination-reflection decomposition formula is based on the simulation of color constancy in the human visual system. Its mathematical model achieves image enhancement by separating the illumination component and the reflection component of the image.

[0047] Furthermore, the crop growth characteristics include leaf area index, leaf thickness, and stem diameter.

[0048] In terms of crop disease and pest control, the occurrence of crop diseases and pests is closely related to the plant characteristics of the crop itself. These characteristics include physiological characteristics, morphological structure, growth cycle and other aspects. Among them, varieties with thick leaves and more wax (such as some wheat varieties) can physically block the attachment of pathogen spores or the feeding of pests. If the root system and leaves are not fully developed, they are susceptible to soil-borne diseases (such as damping-off) and underground pests.

[0049] Leaf area index refers to the ratio of total leaf area per unit of land surface area. The proportion of leaves, as well as leaf thickness and stem diameter, are significantly related to crop diseases and pests.

[0050] Furthermore, the data augmentation and fusion method specifically refers to obtaining the optimized image features of the pest-damaged crops and the crop growth features as three-dimensional feature vectors through matrix vectorization. The specific folded feature matrix vectorization is represented using the following embodiment: If the optimized image features of the pest-damaged crops are ( , ,..., Then, the fused pest identification features obtained by the data augmentation fusion method are represented as follows: .

[0051] Furthermore, the wearable crop pest and disease identification model adopts a lightweight neural network model based on improved crop growth characteristics.

[0052] Furthermore, the lightweight neural network model utilizes an activation function modified based on crop growth characteristics, and uses the activation function value to perform lightweight modification of the input value. The calculation formula for the activation function is as follows: ;

[0053] In the formula, P represents the activation function value, where P is the feature value of the input pest identification features after linear transformation through weights and biases. Leaf area index, For the thickness of the blade, The diameter is the stem diameter.

[0054] Activation functions are suitable for lightweight neural networks, but the appropriate activation function must be selected based on the characteristics of the network design.

[0055] The role of activation functions in lightweight networks is to introduce non-linear characteristics, enabling neural networks to fit complex data distributions and solve problems that linear models cannot handle. For example, in the lightweight network MobileNet V2, the choice of activation function directly affects network performance.

[0056] ReLU activation function: As the most common activation function, it achieves sparse activation by truncating negative values, improving computational efficiency, but may lead to the "dead ReLU" problem (permanent inactivation of some neurons).

[0057] In the activation function design of MobileNet V2, combining depthwise separable convolution operations with ReLU can lead to "trained but unusable" convolution kernels (i.e., a large number of invalid weights) during training. This indicates that the choice of activation function needs to be optimized in conjunction with the network architecture: MobileNet V2 uses linear bottleneck layers instead of ReLU to alleviate this problem by reducing the loss of information in high-dimensional space. In practical applications, a balance needs to be struck between nonlinear requirements and network efficiency. For example, ReLU can be used in the feature extraction layer to ensure nonlinearity, while linear activation functions can be used in the output layer to reduce information loss.

[0058] In summary, lightweight network design requires selecting activation functions based on computational resources and accuracy requirements, and optimizing their compatibility with the network architecture.

[0059] This application utilizes the Tanh activation function, which differs from the conventional ReLU activation function, to improve the Tanh activation function. This allows for adaptive processing of pest identification features, improving the network's representational power while reducing computational load, thus meeting the goal of lightweighting large models for mobile wearable devices.

[0060] In this embodiment, the activation function value is modified by modifying the crop growth characteristics to achieve the goal of lightweight neural network pruning. This results in a reduction of the computational load of the improved lightweight neural network, decreasing the number of computational layers (i.e., hidden layers) to make the model more lightweight. In this embodiment, there are only 3 hidden layers.

[0061] A wearable intelligent identification device for crop diseases and pests is also provided. This system implements a wearable intelligent identification method for crop diseases and pests, including a crop environmental parameter acquisition module, a multispectral imaging module, an edge computing module, a wearable crop disease and pest identification model construction module, and an intelligent identification module for crop diseases and pests.

[0062] The crop environmental parameter acquisition module is used to collect and process crop environmental parameters.

[0063] The multispectral imaging module acquires and processes image features of pests and diseases in crops based on wearable devices.

[0064] The edge computing module includes a crop data storage module, which stores cloud-based pest and disease image data and crop growth characteristics, obtains corresponding pest classification and evaluation results, and performs low-light enhancement processing on the pest and disease crop image features based on the crop environmental parameters and the cloud-based pest and disease image data to obtain optimized pest and disease crop image features; furthermore, based on the optimized pest and disease crop image features and the crop growth characteristics, it obtains pest identification features using a data augmentation fusion method.

[0065] The wearable crop pest and disease identification model construction module: Based on the pest identification features and the pest classification evaluation results, the wearable crop pest and disease identification model is constructed by inputting them into a lightweight neural network model.

[0066] The intelligent crop disease and pest identification module: Based on the wearable crop disease and pest identification model, it performs crop disease and pest identification and evaluation on the real-time crop disease and pest image features captured by personnel who are wearing or re-wearing the wearable device.

[0067] Furthermore, the process of performing low-light enhancement processing on the crop pest and disease image features based on the crop environmental parameters and the cloud-based pest and disease image data to obtain optimized crop pest and disease image features specifically employs an improved light reflection decomposition formula for calculation: ;

[0068] In the formula, x and y are the coordinates of the image pixels. To optimize image features for crops affected by pests and diseases at corresponding coordinates, where m represents the amount of pest and disease image data in the cloud, low-light enhancement processing is performed by comparing the features of the collected pest and disease images with those in the cloud database. This allows for more accurate acquisition of the reflectance after low-light enhancement. This refers to the maximum crop environmental parameter obtained during the crop growth cycle. Based on the minimum crop environmental parameters obtained during the crop growth cycle and their impact on crop diseases and pests, this invention utilizes the deviation values ​​of these parameters to correct the reflectance under low-light enhancement. This is a creative improvement discovered in this application, which can make crop images of diseases and pests more closely resemble real-world image features under the low-light enhancement method. The image features of the diseased and pest-infested crops are obtained at the corresponding coordinates. The Gaussian filter kernel function is the general form of Gaussian filter kernel function adopted in this field. It is usually in the form of an exponential function and is a conventional image filtering method for those skilled in the art, so it will not be described in detail here.

[0069] The core of the illumination-reflection decomposition formula is based on the simulation of color constancy in the human visual system. Its mathematical model achieves image enhancement by separating the illumination component and the reflection component of the image.

[0070] Furthermore, the lightweight neural network model utilizes an activation function modified based on crop growth characteristics, and uses the activation function value to perform lightweight modification of the input value. The calculation formula for the activation function is as follows: ;

[0071] In the formula, P represents the activation function value, where P is the feature value of the input pest identification features after linear transformation through weights and biases. Leaf area index, For the thickness of the blade, The diameter is the stem diameter.

[0072] Activation functions are suitable for lightweight neural networks, but the appropriate activation function must be selected based on the characteristics of the network design.

[0073] Therefore, the beneficial effects of this invention are as follows: a lightweight neural network model is constructed based on crop growth characteristics combined with crop image features collected and processed by wearable devices. By pruning the neural networks of the hidden layers of the activation function, highly structured feature data is processed efficiently. Furthermore, by utilizing high-quality crop pest and disease data from the cloud to process and correct the timely collected pest and disease image data, more suitable pest identification features are obtained. This makes the trained model more lightweight and robust, resulting in more objective identification of crop pests and diseases. This facilitates the development of subsequent pest and disease control methods, improves agricultural production efficiency, and solves the problem of image blurring in low-light conditions for pest and disease images. The improved low-light enhancement image processing method solves the problem of image blurring in pest and disease images, enabling the pest identification features obtained by enhancing and fusing the data input to the model training to make the trained model more closely resemble mobile intelligent identification of crop pests and diseases.

[0074] A neural network is a computational model composed of a large number of interconnected neurons. Inspired by the human nervous system, it can achieve a highly adaptive and nonlinear mapping from input data to output results through the combination and training of multiple layers of neurons.

[0075] Neural networks typically consist of multiple layers, including an input layer, hidden layers, and an output layer. The input layer receives input data, while the hidden and output layers are responsible for calculating the output. Each neuron receives multiple inputs from the previous layer, which are weighted and calculated, then a bias is applied, followed by a nonlinear transformation through an activation function to produce the final output. Connections between neurons are usually represented by weights, where the weight values ​​represent the strength of the connection. These weights can be updated during training to adjust the connection strength between neurons. Neural networks possess strong adaptability and nonlinear mapping capabilities, allowing them to adapt to diverse input data and complex problems. In practical applications, neural networks have been widely used in image recognition, speech recognition, natural language processing, and intelligent control, achieving considerable success.

[0076] This invention is based on a general neural network model, and selects the activation function used in conventional principles to perform nonlinear transformation of input feature data. Therefore, it is an improvement on the activation function of general neural networks, transforming the conventionally used ReLU activation function into the lightweight activation function of this application.

[0077] Tanh Activation Function: The Tanh (hyperbolic tangent) activation function is a commonly used non-linear activation function in deep learning, with a shape similar to the Sigmoid activation function. This function maps input values ​​to the range of -1 to 1. This zero-centered characteristic makes Tanh more powerful than the Sigmoid function in terms of representation.

[0078] ReLU activation function: ReLU (Modified Linear Unit) is a very popular activation function in deep learning, mainly used in hidden layers of neural networks. ReLU is simple and efficient in design, effectively handling the vanishing gradient problem and enabling deep neural networks to be trained.

[0079] Of course, it is understood that each embodiment of the present invention can achieve one of the effects individually, and the combination of multiple embodiments of the present invention can achieve all the above effects. However, it is not required that each embodiment of the present invention achieve all the above advantages and effects, because each embodiment of the present invention can constitute a separate technical solution and make one or more contributions to the prior art. Each embodiment of the present invention is not affected, and the solution can still be implemented even if one embodiment is deleted.

[0080] For any module structures not specifically defined in this invention, the existing technical descriptions shall prevail. The prior art mentioned in the foregoing background and specific embodiments sections can be considered part of this invention and used to understand the meaning of certain technical features or parameters.

Claims

1. A wearable intelligent identification method for crop diseases and pests, characterized in that, The method includes: N1. Collect and process crop environmental parameters, and obtain crop image features of pests and diseases based on wearable devices. The crop data storage module stores cloud-based image data of pests and diseases and crop growth characteristics, and obtains corresponding pest classification and assessment results. N2. Based on the crop environmental parameters and the cloud-based pest and disease image data, perform low-light enhancement processing on the pest and disease crop image features to obtain optimized pest and disease crop image features; N3. Based on the optimized image features of the pest-infested crops and the crop growth features, pest identification features are obtained using a data augmentation fusion method. N4. Based on the pest identification features and the pest classification evaluation results, input them into a lightweight neural network model to construct a wearable crop pest identification model.

2. The wearable intelligent identification method for crop diseases and pests as described in claim 1, characterized in that: The crop environmental parameters are obtained by processing soil environmental parameters, meteorological environmental parameters, and water quality parameters during the actual growth cycle of the crops.

3. A wearable intelligent identification method for crop diseases and pests as described in claim 1 or 2, characterized in that: The optimized image features of pest-infested crops are obtained by performing low-light enhancement processing on the crop environmental parameters and cloud-based pest and disease image data, specifically using an improved light reflection decomposition formula.

4. A wearable intelligent identification method for crop diseases and pests as described in claim 1 or 2, characterized in that: The crop growth characteristics include leaf area index, leaf thickness, and stem diameter.

5. The wearable intelligent identification method for crop diseases and pests as described in claim 3, characterized in that: The data augmentation and fusion method specifically refers to obtaining the optimized image features of the diseased and pest-infested crops and the crop growth features by matrix vectorization using three-dimensional feature vectors.

6. The wearable intelligent identification method for crop diseases and pests as described in claim 4, characterized in that: The wearable crop disease and pest identification model adopts a lightweight neural network model based on improved crop growth characteristics.

7. The wearable intelligent identification method for crop diseases and pests as described in claim 6, characterized in that: The lightweight neural network model utilizes an activation function that is modified based on crop growth characteristics, and uses the activation function value to perform lightweight modification of the input value.

8. A wearable intelligent identification device for crop diseases and pests, the system implementing the method described in any one of claims 1-7, comprising a crop environmental parameter acquisition module, a multispectral imaging module, an edge computing module, a wearable crop disease and pest identification model construction module, and an intelligent identification module for crop diseases and pests, characterized in that: The crop environmental parameter acquisition module is used to collect and process crop environmental parameters. The multispectral imaging module acquires and processes image features of pests and diseases in crops based on wearable devices. The edge computing module includes a crop data storage module, which stores cloud-based pest and disease image data and crop growth characteristics, obtains corresponding pest classification and evaluation results, and performs low-light enhancement processing on the pest and disease crop image features based on the crop environmental parameters and the cloud-based pest and disease image data to obtain optimized pest and disease crop image features; furthermore, based on the optimized pest and disease crop image features and the crop growth characteristics, it obtains pest identification features using a data augmentation fusion method. The wearable crop pest and disease identification model construction module: Based on the pest identification features and the pest classification evaluation results, the wearable crop pest and disease identification model is constructed by inputting them into a lightweight neural network model. The intelligent crop disease and pest identification module: Based on the wearable crop disease and pest identification model, it performs crop disease and pest identification and evaluation on the real-time crop disease and pest image features captured by personnel who are wearing or re-wearing the wearable device.

9. A wearable intelligent identification device for crop diseases and pests as described in claim 8, characterized in that: The optimized image features of pest-infested crops are obtained by performing low-light enhancement processing on the crop environmental parameters and cloud-based pest and disease image data, specifically using an improved light reflection decomposition formula.

10. A wearable intelligent identification device for crop diseases and pests as described in claim 9, characterized in that: The lightweight neural network model utilizes an activation function that is modified based on crop growth characteristics, and uses the activation function value to perform lightweight modification of the input value.

Images