Tea bud yield estimation method, device, equipment and medium based on deep learning

Abstract

The present application relates to a method, apparatus, device, and medium for estimating tea bud yield based on deep learning. The method comprises: responding to a tea bud yield estimation instruction, obtaining an image of tea leaves to be detected containing tea buds, taken by a drone; introducing a super-resolution reconstruction module after the CSPDarknet network in a preset first tea bud recognition model, and introducing a multi-scale channel attention mechanism into the prediction head of the feature fusion network in the first tea bud recognition model to construct a second tea bud recognition model; inputting the image of the tea leaves to be detected into the second tea bud recognition model that has been trained to a convergence state to determine the number of tea buds in the image of the tea leaves to be detected; and using a preset tea bud estimation model to determine the quality of the tea buds in the tea garden to be estimated based on the number of tea buds. The present application can significantly improve the detection accuracy of the tea bud recognition model and greatly improve the accuracy of tea bud yield prediction.

Description

Technical Field

[0001] The present application relates to the field of agricultural production, and in particular to a tea bud yield estimation method based on deep learning, corresponding devices, electronic equipment and computer-readable storage media. Background Art

[0002] As one of the world's three major beverages, tea contains a variety of nutrients, including tea polyphenols, catechins, and trace elements. Its unique flavor is deeply loved by consumers. According to statistics, global tea production in 2023 was approximately 660.4 tons. China, as one of the world's largest tea producers, produced approximately 3.55 million tons annually. Crop yield estimation is a crucial component of modern precision agriculture and is crucial for national agricultural decision-making, farmland production management, and crop storage safety. As a traditionally advantageous and distinctive industry in China, timely and accurate tea yield estimation is crucial for tea garden field management, market stability, and the sustainable development of the tea industry.

[0003] Traditional tea yield estimation involves measuring one acre of land, manually plucking tea buds at the same time, and then weighing the buds to obtain the yield per acre. This method is easily affected by factors such as species, climate, and environment in actual operation, resulting in the estimated data lacking scientific reference value. Multispectral and hyperspectral remote sensing yield estimation methods construct yield estimation models based on vegetation indices and characteristic bands in different bands and biomass composition. While such methods are simple to construct, their implementation requires a large amount of sample data, and their application to other regions or crops is also limited. Crop growth models based on multidisciplinary cross-disciplinary approaches such as agronomy, meteorology, statistics, and soil science have strong mechanistic rationality and high estimation accuracy, but the relevant parameters of this method are difficult to obtain, and the interaction between multiple factors is difficult to determine during the experiment. A more reasonable quantitative description is still needed.

[0004] With the rapid development of computer vision, deep learning, leveraging its deep feature data mining and learning capabilities, has been widely applied in agricultural fields such as crop classification, nutrition monitoring, and pest and disease detection. Tea-related research based on deep learning is also emerging. Existing deep learning-based tea bud yield estimation methods often suffer from inaccurate bud counts and difficulty extracting tea bud feature information in complex environments, resulting in poor yield estimation. These methods also suffer from numerous deficiencies, such as severe occlusion of tea leaves from drones, diverse morphologies, difficulty extracting target features, and the impact of flight altitude on tea image resolution. These issues lead to low model detection accuracy and poor robustness.

[0005] To sum up, in order to adapt to the problems in the existing technology such as severe occlusion of tea leaves from the perspective of drones, diverse shapes, difficulty in extracting target features, and the impact of flight altitude on tea image resolution, which lead to low model detection accuracy and poor robustness, the applicant has made corresponding explorations to solve these problems. Summary of the Invention

[0006] The purpose of this application is to solve the above problems and provide a tea bud yield estimation method based on deep learning, corresponding devices, electronic devices and computer-readable storage media.

[0007] In order to meet the various objectives of this application, this application adopts the following technical solutions:

[0008] A tea bud yield estimation method based on deep learning is proposed to meet one of the purposes of this application, including:

[0009] In response to the tea bud yield estimation instruction, an image of tea leaves to be inspected containing tea buds, taken by the drone, is obtained;

[0010] A super-resolution reconstruction module is introduced after the CSPDarknet network in the preset first tea bud recognition model, and a multi-scale channel attention mechanism is introduced into the prediction head of the feature fusion network in the first tea bud recognition model to construct a second tea bud recognition model;

[0011] Inputting the tea leaf image to be detected into a second tea bud recognition model that has been trained to a convergence state to determine the number of tea buds in the tea leaf image to be detected;

[0012] A preset tea bud estimation model is used to determine the quality of the tea buds in the tea garden to be estimated based on the number of the tea buds, so as to complete the yield estimation of the tea buds.

[0013] Optionally, the steps of introducing a super-resolution reconstruction module after the CSPDarknet network in the preset first tea bud recognition model and introducing a multi-scale channel attention mechanism in the prediction head of the feature fusion network in the first tea bud recognition model to construct a second tea bud recognition model include:

[0014] The super-resolution reconstruction module includes a residual dense block and a residual channel attention module.

[0015] Optionally, the step of inputting the image of tea leaves to be detected into a second tea bud recognition model that has been trained to a convergent state to determine the number of tea buds in the image of tea leaves to be detected includes:

[0016] Inputting the image of tea leaves to be detected containing tea buds into a second tea bud recognition model that has been trained to a convergent state, performing shallow feature extraction using a CSPDarknet network in the second tea bud recognition model to obtain basic low-level features, and inputting the low-level features into the super-resolution reconstruction module;

[0017] In the residual dense block of the super-resolution reconstruction module, the residual dense block adopts multi-layer convolution features and fuses global features to capture detailed information of the image;

[0018] Through global feature fusion, the features of each convolutional layer are combined to extract more expressive global dense features;

[0019] The global dense features are input into the residual channel attention module of the super-resolution reconstruction module. The residual channel attention module uses the attention mechanism to dynamically adjust the importance of feature channels to enhance the network's ability to distinguish features of different channels and improve the quality and details of the reconstructed image.

[0020] Optionally, the step of inputting the image of tea leaves to be detected into a second tea bud recognition model that has been trained to a convergent state to determine the number of tea buds in the image of tea leaves to be detected includes:

[0021] The multi-scale channel attention mechanism of the prediction head of the feature fusion network of the second tea bud recognition model is used to extract feature information from feature maps of different scales to capture various levels of details in the image;

[0022] Perform channel attention calculation on the feature map of each scale, and summarize the features of each channel into a global description through global average pooling or global maximum pooling;

[0023] Generate channel attention weights using a fully connected layer or a convolutional layer, where the channel attention weights represent the importance of each channel;

[0024] Apply the generated channel attention weights to the original feature map to weight the features of each channel to amplify channel features of high importance and suppress channel features of low importance;

[0025] The weighted feature maps are fused to combine information at different scales to generate a comprehensive feature map with multi-scale context;

[0026] The fused feature map is input into the prediction head of the feature fusion network of the second tea bud recognition model, and the tea buds in the tea leaf image to be detected are output.

[0027] Optionally, the steps of constructing a tea bud estimation model include:

[0028] Determining the number of tea buds corresponding to each tea bud mass, and calculating an average of the number of tea buds corresponding to each tea bud mass, wherein the tea bud mass can be any value between 10 g and 500 g;

[0029] The least square method is used to fit the masses of the tea buds and the corresponding numbers of the tea buds to determine the parameters of the linear model, so as to complete the construction of the tea bud estimation model.

[0030] Optionally, the tea bud estimation model is expressed as:

[0031] n=a.m+b;

[0032] Among them, n is the predicted number of tea buds, m is the quality of tea buds, and a and b are the parameters of the tea bud estimation model.

[0033] Optionally, the basic network architecture of the first tea bud recognition model is a YOLOv8 model, and the basic network architecture of the second tea bud recognition model is an improved YOLOv8 model.

[0034] A tea bud yield estimation device based on deep learning is provided to meet another purpose of the present application, comprising:

[0035] a tea leaf image acquisition module configured to respond to a tea bud yield estimation instruction and acquire an image of tea leaves to be inspected containing tea buds, taken by a drone;

[0036] a recognition model construction module configured to introduce a super-resolution reconstruction module after the CSPDarknet network in the preset first tea bud recognition model, and introduce a multi-scale channel attention mechanism into the prediction head of the feature fusion network in the first tea bud recognition model to construct a second tea bud recognition model;

[0037] a bud quantity determination module configured to input the tea leaf image to be detected into a second tea bud recognition model that has been trained to a convergence state, so as to determine the number of tea buds in the tea leaf image to be detected;

[0038] The bud yield estimation module is configured to use a preset tea bud estimation model to determine the quality of the tea buds in the tea garden to be estimated based on the number of tea buds, so as to complete the tea bud yield estimation.

[0039] An electronic device provided to meet another purpose of the present application includes a central processing unit and a memory, wherein the central processing unit is used to call and run a computer program stored in the memory to execute the steps of the deep learning-based tea bud yield estimation method described in the present application.

[0040] A computer-readable storage medium is provided to meet another purpose of the present application, which stores a computer program implemented according to the deep learning-based tea bud yield estimation method in the form of computer-readable instructions. When the computer program is called and executed by a computer, the steps included in the corresponding method are executed.

[0041] Compared with the existing technology, this application addresses the problems in the existing technology such as severe occlusion of tea leaves from the perspective of drones, diverse shapes, difficulty in extracting target features, and the impact of flight altitude on tea image resolution, which lead to low model detection accuracy and poor robustness. This application includes but is not limited to the following beneficial effects:

[0042] First, this application introduces a super-resolution reconstruction module after the CSPDarknet network in the YOLOv8 network model, which can help the model better capture and utilize features at different scales, thereby improving the accuracy of tea bud recognition;

[0043] Second, by introducing a multi-scale channel attention mechanism into the prediction head of the feature fusion network of the YOLOv8 network model, the improved YOLOv8 network model can effectively process and fuse feature maps from different scales, capturing various details in the image, thereby improving the recognition accuracy of tea buds.

[0044] Furthermore, in view of the fact that the tea images from the perspective of drones show global clarity and local blur, and the tea shows characteristics such as high density, multi-scale and complex texture, this application introduces a super-resolution reconstruction module and a multi-scale channel attention mechanism to construct an improved YOLOv8 network model, which realizes the detection and counting of tea buds in complex natural environments, significantly improves the detection accuracy of the tea bud recognition model, and at the same time greatly improves the accuracy of tea bud yield prediction, making agricultural producers' prediction of tea bud yield more scientific, greatly saving manpower and material resources, and laying a solid theoretical foundation for the sustainable development of smart agriculture. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0046] Figure 1 Schematic diagram of the process of the tea bud yield estimation method based on deep learning in the embodiment of the present application;

[0047] Figure 2 Schematic diagram of the structure of the improved YOLOv8 model in the embodiment of the present application;

[0048] Figure 3 Schematic diagram of the structure of the residual module in the embodiment of the present application;

[0049] Figure 4 This is a schematic diagram of the structure of the dense module in the embodiment of the present application;

[0050] Figure 5 This is a structural diagram of the residual dense module in an embodiment of the present application;

[0051] Figure 6 Schematic diagram of the structure of the residual channel attention module in an embodiment of the present application;

[0052] Figure 7 This is a schematic diagram of the structure of multi-scale channel attention in an embodiment of the present application;

[0053] Figure 8 This is a schematic diagram of the structure of the improved YOLOv8 model trained in the embodiment of the present application;

[0054] Figure 9 This is a principle block diagram of a tea bud yield estimation device based on deep learning in an embodiment of the present application;

[0055] Figure 10 Schematic diagram of the structure of the computer device in the embodiment of the present application. DETAILED DESCRIPTION

[0056] The following describes in detail embodiments of the present application, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present application, and are not to be construed as limiting the present application.

[0057] It will be understood by those skilled in the art that, unless expressly stated otherwise, the singular forms "a", "an", "said" and "the" used herein may also include the plural forms. It should be further understood that the term "comprising" used in the specification of the present application refers to the presence of the features, integers, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or groups thereof. It should be understood that when we refer to an element as being "connected" or "coupled" to another element, it may be directly connected or coupled to the other element, or there may be intermediate elements. In addition, "connected" or "coupled" as used herein may include wireless connections or wireless couplings. The term "and / or" used herein includes all or any units and all combinations of one or more associated listed items.

[0058] It will be understood by those skilled in the art that, unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by those skilled in the art to which this application belongs. It should also be understood that terms such as those defined in common dictionaries should be understood to have meanings consistent with their meanings in the context of the prior art and will not be interpreted in an idealized or overly formal sense unless specifically defined as herein.

[0059] Those skilled in the art will appreciate that the terms "client," "terminal," and "terminal device" as used herein include both devices that are wireless signal receivers, i.e., devices that only have wireless signal receivers without transmission capabilities, and devices that have receiving and transmitting hardware capable of two-way communication over a two-way communication link. Such devices may include: cellular or other communication devices such as personal computers and tablet computers, which have single-line displays, multi-line displays, or cellular or other communication devices without multi-line displays; PCS (Personal Communications Services), which may combine voice, data processing, fax, and / or data communication capabilities; PDA (Personal Digital Assistant), which may include a radio frequency receiver, a pager, Internet / Intranet access, a web browser, a notepad, a calendar, and / or a GPS (Global Positioning System) receiver; and conventional laptop and / or palmtop computers or other devices, which have and / or include a radio frequency receiver. As used herein, the terms "client," "terminal," or "terminal device" may be portable, transportable, or installed in a vehicle (air, sea, and / or land), or may be adapted and / or configured to operate locally, and / or in a distributed manner, at any other location on Earth and / or in space. As used herein, the terms "client," "terminal," or "terminal device" may also refer to a communication terminal, an Internet access terminal, or a music / video playback terminal, such as a PDA, a MID (Mobile Internet Device), and / or a mobile phone with music / video playback capabilities, or may refer to a smart TV, a set-top box, or other device.

[0060] The hardware referred to by names such as "server", "client", and "service node" in this application is essentially an electronic device with capabilities equivalent to those of a personal computer. It is a hardware device that has the necessary components revealed by the von Neumann principle, such as a central processing unit (including an arithmetic unit and a controller), a memory, an input device, and an output device. Computer programs are stored in its memory, and the central processing unit loads the program stored in the external memory into the internal memory for execution, executes the instructions in the program, and interacts with the input and output devices to complete specific functions.

[0061] It should be noted that the concept of "server" referred to in this application can also be extended to server clusters. Based on the network deployment principles understood by those skilled in the art, the servers described should be logically divided. In physical space, these servers can be independent of each other but callable through interfaces, or integrated into a single physical computer or a computer cluster. Those skilled in the art should understand this flexibility and should not use it to constrain the implementation of the network deployment method of this application.

[0062] Unless expressly specified, one or more technical features of the present application can be deployed on a server for implementation and accessed by a client through a remote call to obtain an online service interface provided by the server, or can be directly deployed and run on a client for implementation.

[0063] Unless expressly specified otherwise, the neural network models referenced or may be referenced in this application may be deployed on a remote server and remotely called on the client, or may be deployed and directly called on a client with sufficient device capabilities. In some embodiments, when it runs on the client, its corresponding intelligence may be obtained through transfer learning to reduce the requirements for the client's hardware operating resources and avoid excessive occupation of the client's hardware operating resources.

[0064] Unless explicitly specified, the various data involved in this application can be stored remotely on a server or on a local terminal device, as long as they are suitable for being called by the technical solution of this application.

[0065] Those skilled in the art should be aware that although the various methods of this application are described based on the same concept and thus exhibit commonality, unless otherwise specified, these methods can be independently executed. Similarly, the various embodiments disclosed in this application are all based on the same inventive concept. Therefore, concepts with the same expression, as well as concepts that are appropriately transformed for convenience despite different expression, should be understood as equivalent.

[0066] Unless expressly stated to be mutually exclusive, the various embodiments disclosed in this application may be cross-combined with the relevant technical features of the various embodiments to flexibly construct new embodiments, as long as such combination does not deviate from the creative spirit of this application and can meet the needs of the prior art or resolve certain deficiencies in the prior art. Those skilled in the art should be aware of such flexibility.

[0067] See also Figure 1 In one embodiment, the tea bud yield estimation method based on deep learning of the present application includes:

[0068] Step S10: responding to the tea bud yield estimation instruction, obtaining an image of the tea leaves to be tested containing the tea buds, taken by the drone;

[0069] The tea bud yield estimation system can respond to the tea bud yield estimation instruction and obtain the image of the tea leaves to be detected containing the tea buds taken by the drone;

[0070] Specifically, the tea garden under study can be divided into multiple sampling areas, and 9 data sampling points are randomly selected. The data are acquired and processed to generate a dataset for training the model. A drone is used to capture RGB images of several tea buds at the picking stage (one bud and two leaves) at the 9 data sampling points at different flight altitudes. The drone RGB image resolution is 5472x3648. The drone flies according to the planned path trajectory, and the camera unit on the drone system is installed with set parameters to collect data while flying. The heading overlap rate is set to 90% and the lateral overlap rate is set to 80%. In addition, the camera unit on the drone system is in a 90° vertical shooting state throughout the flight. Based on the image stitching software Pi x4DMapper, the collected tea bud images are subjected to radiation correction and image stitching to obtain a digital orthophoto of the tea garden. To meet the needs of subsequent models and the needs of dataset diversification, the orthophoto images are cropped into small images of 640 pixels x 640 pixels, which ultimately constitute the tea bud dataset.

[0071] The tea buds in the samples were annotated using Label Img software. The criteria for determining the tea buds were mainly based on the following requirements: clear and complete tea buds; with the relatively complete one bud and two leaves picking standard within the image range as a reference, the target tea buds were annotated with horizontal frames, and the training set and validation set were divided into two parts according to the ratio of 7:1.5:1.5. Since the tea bud images and old leaves are of the same color, in order to reduce the influence of the background on the image, the brightness and contrast of the tea bud images were enhanced to increase the difference between the tea buds and the old leaves and other backgrounds, so that the model can better extract the target features.

[0072] Step S20: introducing a super-resolution reconstruction module after the CSPDarknet network in the preset first tea bud recognition model, and introducing a multi-scale channel attention mechanism in the prediction head of the feature fusion network in the first tea bud recognition model to construct a second tea bud recognition model;

[0073] After obtaining the image of tea leaves to be detected containing tea buds taken by a drone, a super-resolution reconstruction module is introduced after the CSPDarknet network in the preset first tea bud recognition model, and a multi-scale channel attention mechanism is introduced in the prediction head of the feature fusion network in the first tea bud recognition model to construct a second tea bud recognition model, wherein the basic network architecture of the first tea bud recognition model is the YOLOv8 model, and the basic network architecture of the second tea bud recognition model is the improved YOLOv8 model, and the super-resolution reconstruction module includes a residual dense block and a residual channel attention module.

[0074] In some embodiments, the CSPDarknet network in the first tea bud recognition model is the backbone network used for feature extraction in the YOLOv8 target detection model. The CSPDarknet network is an improvement on the traditional Darknet network, and improves the performance and computational efficiency of the network by introducing the Cross-Stage Partial (CSP) structure.

[0075] Specifically, see Figure 2 , constructing the second tea bud recognition model, specifically including: introducing a super-resolution reconstruction module and a multi-scale attention module into the YOLOv8 model that can refine features and realize multi-scale information fusion, wherein the super-resolution reconstruction module is composed of a residual dense block and a residual channel attention module. The residual dense block can make full use of the hierarchical features of each convolutional layer in the image to be reconstructed, and use the global feature fusion method to obtain the global dense features in the image to be reconstructed, thereby improving the image reconstruction performance; the residual channel attention module uses the attention mechanism to treat the feature information of different channels differently, thereby improving the network's super-resolution reconstruction capability for drone tea images and restoring the complex texture information of the image as much as possible. The residual dense block is an integration of the residual module and the dense module, combining the two to form a residual dense module.

[0076] See also Figures 3 to 5 ,in, Figure 3 is the residual module, Figure 4 For dense modules, Figure 5 The residual dense block model realizes the feature extraction of tea buds through a shallow feature extraction network. After the residual module is processed, the dense module establishes multi-layer skip links to map deeper features between feature maps. The residual channel attention module combines attention with residual ideas and integrates channel attention into the residual block to form a residual channel attention module. Figure 6 As shown, for the b-th residual block:

[0077] F g,b =Fg,b-1 +R g,b (X g,b )X g,b ,

[0078] Where, F g,b-1 represents the input of the residual channel attention module, F g,b ——The output of the residual channel attention module, R g,b represents channel attention, X g,b Indicates that it has undergone convolution and ReLU operations.

[0079] Specifically, first the input is subjected to convolution and ReLU operations to obtain X g,b , then X g,b After a channel attention mechanism, the result is added to the input to obtain the output of the residual channel attention module. The convolution operation uses a 3×3 convolution kernel. In addition, in view of the characteristics of the tea bud images collected by drones, such as complex textures, different shapes, and inconsistent sizes, a multi-scale channel attention module is introduced. While achieving multi-scale feature fusion of high-resolution dense images, the information on each channel is retained as the goal, the channel dimensions are grouped into multiple sub-features, focusing on the multi-scale contextual information of tea buds, and optimizing the high-resolution dense target prediction performance. The multi-scale channel attention module is as follows: Figure 7 shown.

[0080] In some embodiments, this application uses an end-to-end approach to train the improved YOLOv8 network described above, using the PyTorch deep learning framework for model improvement and algorithm training. The computer is equipped with an NVIDIA GeForce RTX3060 graphics processing unit (GPU) and a 12th Gen Intel(R) Core(TM) i5-12500 CPU. The experimental environment is Windows 10 Version 22H2 64-bit system and Python version 3.8.

[0081] In some embodiments, the model training parameters are set as follows: the number of training rounds (epochs) is determined to be 400 and the batch size (batch) is 16 through preliminary experiments; the training optimizer is Adam, the learning rate is 0.01, the momentum is 0.937, and the weight decay coefficient is 0.000 5. The training set images are input into the network for model training, and a validation is performed after every 5 rounds of training, and the best model weights are saved. The saved model weights are used to perform an accuracy test on the test set images and predict the number of tea buds. The model training process is as follows: Figure 8 shown.

[0082] Step S30: inputting the tea leaf image to be detected into a second tea bud recognition model that has been trained to a convergence state to determine the number of tea buds in the tea leaf image to be detected;

[0083] After the second tea bud recognition model is constructed, the image of the tea leaves to be detected is input into the second tea bud recognition model that has been trained to a convergence state to determine the number of tea buds in the image of the tea leaves to be detected;

[0084] Furthermore, the step of inputting the tea leaf image to be detected into a second tea bud recognition model that has been trained to a convergent state to determine the number of tea buds in the tea leaf image to be detected includes:

[0085] Step S301: inputting the image of tea leaves to be detected containing tea buds into a second tea bud recognition model that has been trained to a convergent state, performing shallow feature extraction using the CSPDarknet network in the second tea bud recognition model to obtain basic low-level features, and inputting the low-level features into the super-resolution reconstruction module;

[0086] Step S302: In the residual dense block of the super-resolution reconstruction module, the residual dense block adopts multi-layer convolution features and fuses global features to capture detailed information of the image;

[0087] Step S303: Combining the features of each convolutional layer through global feature fusion to extract more expressive global dense features;

[0088] Step S304: input the global dense features into the residual channel attention module of the super-resolution reconstruction module. The residual channel attention module uses the attention mechanism to dynamically adjust the importance of feature channels to enhance the network's ability to distinguish different channel features and improve the quality and details of the reconstructed image.

[0089] From the above steps, it can be seen that introducing the super-resolution reconstruction module after the CSPDarknet network in the preset first tea bud recognition model can help the model better capture and utilize features of different scales, thereby improving the accuracy of bud recognition.

[0090] Furthermore, the step of inputting the image of tea leaves to be detected into a second tea bud recognition model that has been trained to a convergent state to determine the number of tea buds in the image of tea leaves to be detected includes:

[0091] Step S3001: using the multi-scale channel attention mechanism of the prediction head of the feature fusion network of the second tea bud recognition model to extract feature information from feature maps of different scales to capture various levels of details in the image;

[0092] Step S3002: perform channel attention calculation on the feature map of each scale, and aggregate the features of each channel into a global description through global average pooling or global maximum pooling;

[0093] Step S3003: Generate channel attention weights using a fully connected layer or a convolutional layer, wherein the channel attention weights represent the importance of each channel;

[0094] Step S3004: Apply the generated channel attention weights to the original feature map to weight the features of each channel to amplify channel features of high importance and suppress channel features of low importance;

[0095] Step S3005: Fusing the weighted feature maps, combining information at different scales, and generating a comprehensive feature map with multi-scale context;

[0096] Step S3006: input the fused feature map into the prediction head of the feature fusion network of the second tea bud recognition model, and output the tea buds in the tea leaf image to be detected.

[0097] From the above steps, it can be seen that by introducing the multi-scale channel attention mechanism into the prediction head of the feature fusion network of the first tea bud recognition model, the second tea bud recognition model can effectively process and fuse feature maps from different scales, capture various details in the image, and thus improve the recognition accuracy of tea buds.

[0098] Step S40: using a preset tea bud estimation model to determine the quality of the tea buds in the tea garden to be estimated based on the number of the tea buds, so as to complete the yield estimation of the tea buds.

[0099] After determining the number of tea buds in the tea leaf image to be detected, a preset tea bud estimation model is used to determine the quality of the tea buds in the tea garden to be estimated based on the number of tea buds, so as to complete the tea bud yield estimation.

[0100] In some embodiments, the step of constructing a tea bud estimation model includes:

[0101] Step S401: determining the number of tea buds corresponding to each tea bud mass, and calculating an average value of the number of tea buds corresponding to each tea bud mass, wherein the tea bud mass can be any value between 10 g and 500 g;

[0102] Step S402: using the least squares method to perform fitting based on the masses of the tea buds and the corresponding quantities of the tea buds to determine the parameters of the linear model, thereby completing the construction of the tea bud estimation model.

[0103] Specifically, the tea bud yield is estimated by designing an experiment to establish a linear relationship between the number of tea buds and the quality of tea buds, so as to estimate the quality of tea buds by quantity; 30 portions of tea buds, 10, 20, …, 90, 100, 110, 120, …, 190, 200, 210, 220, …, 290, 300 g, can be weighed from the manually harvested tea buds, and each mass is weighed independently 3 times (the buds need to be put back after weighing, and the buds of that mass are weighed again from the harvested buds). The experiment is carried out 3 times independently at different sampling points, and the number of tea buds under each mass is recorded, and the average number of tea buds is calculated. Finally, the relationship between the quantity and quality of tea buds is fitted using the least squares method to construct a tea bud estimation model.

[0104] Furthermore, the expression of the tea bud estimation model is:

[0105] n=a.m+b;

[0106] Among them, n is the predicted number of tea buds, m is the quality of tea buds, and a and b are the parameters of the tea bud estimation model.

[0107] As can be seen from the above embodiments, compared with the prior art, the present application addresses the problems in the prior art such as severe occlusion of tea leaves from the perspective of drones, diverse shapes, difficulty in extracting target features, and the impact of flight altitude on tea image resolution, which lead to low model detection accuracy and poor robustness. The present application includes but is not limited to the following beneficial effects:

[0108] First, this application introduces a super-resolution reconstruction module after the CSPDarknet network in the YOLOv8 network model, which can help the model better capture and utilize features at different scales, thereby improving the accuracy of tea bud recognition;

[0109] Second, by introducing a multi-scale channel attention mechanism into the prediction head of the feature fusion network of the YOLOv8 network model, the improved YOLOv8 network model can effectively process and fuse feature maps from different scales, capturing various details in the image, thereby improving the recognition accuracy of tea buds.

[0110] Furthermore, in view of the fact that the tea images from the perspective of drones show global clarity and local blur, and the tea shows characteristics such as high density, multi-scale and complex texture, this application introduces a super-resolution reconstruction module and a multi-scale channel attention mechanism to construct an improved YOLOv8 network model, which realizes the detection and counting of tea buds in complex natural environments, significantly improves the detection accuracy of the tea bud recognition model, and at the same time greatly improves the accuracy of tea bud yield prediction, making agricultural producers' prediction of tea bud yield more scientific, greatly saving manpower and material resources, and laying a solid theoretical foundation for the sustainable development of smart agriculture.

[0111] See also Figure 9 A tea bud yield estimation device based on deep learning is provided to meet one of the purposes of this application, including a tea image acquisition module 1100, a recognition model construction module 1200, a bud quantity determination module 1300 and a bud yield estimation module 1400. Among them, the tea image acquisition module 1100 is configured to respond to the tea bud yield estimation instruction and obtain the image of the tea leaves to be detected containing tea buds taken by the drone; the recognition model construction module 1200 is configured to introduce a super-resolution reconstruction module after the CSPDarknet network in the preset first tea bud recognition model, and introduce a multi-scale channel attention mechanism in the prediction head of the feature fusion network in the first tea bud recognition model to construct a second tea bud recognition model; the bud quantity determination module 1300 is configured to input the tea leaf image to be detected into the second tea bud recognition model that has been trained to a convergence state to determine the number of tea buds in the tea leaf image to be detected; the bud yield estimation module 1400 is configured to use the preset tea bud estimation model to determine the quality of the tea buds in the tea garden to be estimated based on the number of tea buds, so as to complete the yield estimation of the tea buds.

[0112] Based on any embodiment of this application, please refer to Figure 10 Another embodiment of the present application further provides an electronic device, which can be implemented by a computer device, such as Figure 10As shown, a schematic diagram of the internal structure of a computer device. The computer device includes a processor, a computer-readable storage medium, a memory, and a network interface connected via a system bus. Among them, the computer-readable storage medium of the computer device stores an operating system, a database, and computer-readable instructions, and the database may store a control information sequence. When the computer-readable instructions are executed by the processor, the processor may implement a tea bud yield estimation method based on deep learning. The processor of the computer device is used to provide computing and control capabilities to support the operation of the entire computer device. The memory of the computer device may store computer-readable instructions. When the computer-readable instructions are executed by the processor, the processor may execute the tea bud yield estimation method based on deep learning of the present application. The network interface of the computer device is used to connect and communicate with the terminal. Those skilled in the art will understand that Figure 10 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0113] In this embodiment, the processor is used to execute Figure 9 The memory stores the program code and various data required to execute the specific functions of each module and its submodule. The network interface is used to transmit data between user terminals or servers. The memory in this embodiment stores the program code and data required to execute all modules / submodules in the deep learning-based tea bud yield estimation device of this application. The server can call the server's program code and data to execute the functions of all submodules.

[0114] The present application also provides a storage medium storing computer-readable instructions. When the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the deep learning-based tea bud yield estimation method described in any embodiment of the present application.

[0115] The present application also provides a computer program product, including a computer program / instruction, which, when executed by one or more processors, implements the steps of the deep learning-based tea bud yield estimation method described in any embodiment of the present application.

[0116] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments of the present application can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. The aforementioned storage medium can be a computer-readable storage medium such as a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0117] The above description is only part of the implementation methods of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

[0118] To sum up, this application aims at the tea images from the perspective of drones, which show global clarity and local blur, and the characteristics of tea leaves such as high density, multi-scale and complex texture. This application introduces a super-resolution reconstruction module and a multi-scale channel attention mechanism to construct an improved YOLOv8 network model, which realizes the detection and counting of tea buds in natural environments, and significantly improves the detection accuracy of the tea bud recognition model. On the basis of improving the detection accuracy of the tea bud recognition model, it enables agricultural producers to make more scientific predictions on the yield of tea buds, greatly saves manpower and material resources, and lays a solid theoretical foundation for the sustainable development of smart agriculture.

Claims

1. A tea bud yield estimation method based on deep learning, characterized in that: include: In response to the tea bud yield estimation instruction, an image of tea leaves to be inspected containing tea buds, taken by the drone, is obtained; A super-resolution reconstruction module is introduced after the CSPDarknet network in the preset first tea bud recognition model, and a multi-scale channel attention mechanism is introduced into the prediction head of the feature fusion network in the first tea bud recognition model to construct a second tea bud recognition model, wherein the basic network architecture of the first tea bud recognition model is a YOLOv8 model, and the basic network architecture of the second tea bud recognition model is an improved YOLOv8 model, and the super-resolution reconstruction module includes a residual dense block and a residual channel attention module; Inputting the tea leaf image to be detected into a second tea bud recognition model that has been trained to a convergent state to determine the number of tea buds in the tea leaf image to be detected, which includes: Inputting the image of tea leaves to be detected containing tea buds into a second tea bud recognition model that has been trained to a convergent state, performing shallow feature extraction using a CSPDarknet network in the second tea bud recognition model to obtain basic low-level features, and inputting the low-level features into the super-resolution reconstruction module; In the residual dense block of the super-resolution reconstruction module, the residual dense block adopts multi-layer convolution features and fuses global features to capture detailed information of the image; Through global feature fusion, the features of each convolutional layer are combined to extract more expressive global dense features; Inputting the global dense features into the residual channel attention module of the super-resolution reconstruction module, the residual channel attention module uses the attention mechanism to dynamically adjust the importance of feature channels to enhance the network's ability to distinguish features of different channels and improve the quality and details of the reconstructed image; A preset tea bud estimation model is used to determine the quality of the tea buds in the tea garden to be estimated based on the number of tea buds, so as to complete the yield estimation of the tea buds.

2. The tea bud yield estimation method based on deep learning according to claim 1, characterized in that: The step of inputting the tea leaf image to be detected into a second tea bud recognition model that has been trained to a convergent state to determine the number of tea buds in the tea leaf image to be detected comprises: The multi-scale channel attention mechanism of the prediction head of the feature fusion network of the second tea bud recognition model is used to extract feature information from feature maps of different scales to capture various levels of details in the image; Perform channel attention calculation on the feature map of each scale, and summarize the features of each channel into a global description through global average pooling or global maximum pooling; Generate channel attention weights using a fully connected layer or a convolutional layer, where the channel attention weights represent the importance of each channel; Apply the generated channel attention weights to the original feature map to weight the features of each channel to amplify channel features of high importance and suppress channel features of low importance; The weighted feature maps are fused to combine information at different scales to generate a comprehensive feature map with multi-scale context; The fused feature map is input into the prediction head of the feature fusion network of the second tea bud recognition model, and the tea buds in the tea leaf image to be detected are output.

3. The tea bud yield estimation method based on deep learning according to claim 1, characterized in that: The steps to construct the tea bud estimation model include: Determining the number of tea buds corresponding to each tea bud mass, and calculating and determining an average value of the number of tea buds corresponding to each tea bud mass, wherein the tea bud mass can be any value between 10 g and 500 g; The least square method is used to fit the masses of the tea buds and the corresponding numbers of the tea buds to determine the parameters of the linear model, so as to complete the construction of the tea bud estimation model.

4. The tea bud yield estimation method based on deep learning according to claim 3, characterized in that: The expression of the tea bud estimation model is: n=a.m+b; Where n is the predicted number of tea buds, m is the quality of tea buds, and a and b are the parameters of the tea bud estimation model.

5. A tea bud yield estimation device based on deep learning, characterized in that: include: a tea leaf image acquisition module configured to respond to a tea bud yield estimation instruction and acquire an image of tea leaves to be inspected containing tea buds, taken by a drone; A recognition model construction module is configured to introduce a super-resolution reconstruction module after the CSPDarknet network in the preset first tea bud recognition model, and introduce a multi-scale channel attention mechanism in the prediction head of the feature fusion network in the first tea bud recognition model to construct a second tea bud recognition model, wherein the basic network architecture of the first tea bud recognition model is a YOLOv8 model, and the basic network architecture of the second tea bud recognition model is an improved YOLOv8 model, and the super-resolution reconstruction module includes a residual dense block and a residual channel attention module; The bud number determination module is configured to input the tea leaf image to be detected into a second tea bud recognition model that has been trained to a convergence state to determine the number of tea buds in the tea leaf image to be detected, and includes: Inputting the image of tea leaves to be detected containing tea buds into a second tea bud recognition model that has been trained to a convergent state, performing shallow feature extraction using a CSPDarknet network in the second tea bud recognition model to obtain basic low-level features, and inputting the low-level features into the super-resolution reconstruction module; In the residual dense block of the super-resolution reconstruction module, the residual dense block adopts multi-layer convolution features and fuses global features to capture detailed information of the image; Through global feature fusion, the features of each convolutional layer are combined to extract more expressive global dense features; Inputting the global dense features into the residual channel attention module of the super-resolution reconstruction module, the residual channel attention module uses the attention mechanism to dynamically adjust the importance of feature channels to enhance the network's ability to distinguish features of different channels and improve the quality and details of the reconstructed image; The bud yield estimation module is configured to use a preset tea bud estimation model to determine the quality of the tea buds in the tea garden to be estimated based on the number of tea buds, so as to complete the tea bud yield estimation.

6. An electronic device comprising a central processing unit and a memory, characterized in that: The central processing unit is configured to call and run a computer program stored in the memory to execute the steps of the method according to any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that It stores a computer program implemented according to the method described in any one of claims 1 to 4 in the form of computer-readable instructions, and when the computer program is called and executed by a computer, the steps included in the corresponding method are executed.

Images