Image recognition monitoring system and method based on Chinese cabbage growth state parameters

Abstract

The invention discloses an image recognition monitoring system and method based on Chinese cabbage growth state parameters. The system comprises an image acquisition module, an image preprocessing module, a feature extraction module and a growth state monitoring and user management module, and growth state parameters and growth environment parameter data of Chinese cabbages are acquired through cooperative work of a camera array and an environment sensor group; an improved YOLOv8 neural network is adopted, growth state parameters are fused into a feature extraction process, and the image feature extraction capability in a complex field environment is improved. The growth state monitoring module constructs a multi-modal fusion neural network evaluation model, accurately judges the growth stage, the health condition and the nutrition condition, and generates targeted suggestions. The system also has model dynamic optimization and early warning functions, realizes accurate monitoring and dynamic analysis of the growth state of the Chinese cabbage, and provides scientific and effective decision support for planting management.

Description

Technical Field

[0001] The present invention relates to the technical field of agricultural intelligent monitoring, and in particular to an image recognition monitoring system and method based on Chinese cabbage growth status parameters. Background Art

[0002] With the development of information technology, some agricultural monitoring systems have begun to incorporate sensor technology. These include soil moisture sensors and light sensors to obtain environmental parameters, and chlorophyll sensors and leaf area meters to measure plant physiological parameters. However, these sensor-based monitoring methods have significant limitations. Firstly, sensors can only capture a single or limited set of parameters, failing to fully reflect the overall growth status of Chinese cabbage. Secondly, sensors are expensive to install and maintain, and are susceptible to interference from field environmental factors, compromising data reliability.

[0003] In recent years, image recognition technology has been gradually applied to agriculture, enabling growth monitoring by analyzing crop images. However, existing image recognition-based monitoring systems still have shortcomings: most systems perform only simple feature extraction from images, failing to fully integrate key parameters of the crop growth process. At the algorithmic level, traditional neural network models have limited ability to extract image features in complex field environments, making it difficult to accurately identify subtle growth anomalies such as pests and diseases and nutrient deficiencies. Furthermore, existing systems lack the ability to dynamically analyze and predict growth parameters, making them unable to provide scientific and effective decision support for crop management.

[0004] Therefore, there is an urgent need for a cabbage growth status monitoring system and method that can integrate multi-source data and has accurate monitoring capabilities. Summary of the Invention

[0005] The main purpose of this application is to provide an image recognition monitoring system and method based on Chinese cabbage growth status parameters, aiming to address the shortcomings of existing technologies in monitoring Chinese cabbage growth status. To achieve the above purpose, this application provides the following solutions:

[0006] In a first aspect, the present application provides an image recognition and monitoring system based on Chinese cabbage growth status parameters, comprising an image acquisition module, an image preprocessing module, a feature extraction module, a growth status monitoring module, and a user management module. The image acquisition module comprises a camera array and an environmental sensor group. The camera array is used to acquire images of Chinese cabbage from different angles to obtain Chinese cabbage growth status parameters; the environmental sensor group is used to acquire Chinese cabbage growth environment parameters; the growth status parameters, growth environment parameters, and image acquisition time together constitute image metadata;

[0007] The image preprocessing module is used to perform labeling and data enhancement processing on the collected images, and quantify the growth state parameters into feature vectors as label data;

[0008] The feature extraction module adopts an improved deep learning network to integrate the growth state parameters into the network structure and calculation process for feature extraction and analysis of Chinese cabbage images;

[0009] The growth status monitoring module constructs an evaluation model based on the features extracted by the feature extraction module and the growth status parameters to determine the growth stage, health status and nutritional status of the Chinese cabbage;

[0010] The user management module is used to store image data, growth state parameters, feature extraction results and evaluation results.

[0011] Furthermore, when the image preprocessing module annotates the collected image, the growth state parameters of the Chinese cabbage are quantified into a multidimensional feature vector The growth state parameters include plant height H, number of leaves N, leaf area S, spread D and chlorophyll content C chl , and use the multidimensional feature vector as label data for supervised learning.

[0012] Furthermore, the improved deep learning network is an improved YOLOv8 neural network, including a backbone network, a neck network, and a head network; the backbone network adopts an improved C2f structure, and the growth state parameters are integrated into the feature extraction process through the following formula:

[0013]

[0014] Among them, n is the number of residual blocks, α is the weight coefficient, FC is the fully connected layer operation, F in is the input image feature map, F out is the output feature map.

[0015] Furthermore, the neck network adopts the BiFPN structure. During the feature fusion process, the growth state parameter is introduced to dynamically adjust the feature weight. The calculation formula for the m-th layer feature fusion is:

[0016]

[0017] Among them, the weight ω l According to the chlorophyll content chl calculate:

[0018]

[0019] β is the adjustment coefficient, L is the number of input features, f l is the lth input feature.

[0020] Furthermore, the evaluation model constructed by the growth status monitoring module is a multimodal fusion neural network, and the input includes the feature vector extracted by the improved YOLOv8 neural network. and growth state parameter vector The health status score S is calculated by the following formula health :

[0021]

[0022] in, is the weight matrix, ω1 and ω2 are weight coefficients, and ReLU is the linear rectification function.

[0023] Furthermore, the growth status monitoring module also calculates the nutritional supplement recommendation coefficient S according to the health status assessment results and environmental parameters through the following formula: n :

[0024] S n =δ1·Norm(S health )+δ2·Norm(I i -I s )+δ3·Norm(T i -T s )+δ4·Norm(H i -H s )

[0025] Among them, I i 、T i and H i are the growth environment parameters of light intensity, temperature and humidity, I s 、T s and H s are the optimal light intensity, temperature and humidity for the growth of Chinese cabbage, and δ1, δ2, δ3 and δ4 are weight coefficients.

[0026] Furthermore, the system further includes an early warning module, which issues an early warning when it determines that the Chinese cabbage is in an unhealthy state based on the evaluation result of the growth status monitoring module.

[0027] In a second aspect, the present application provides a method for monitoring the growth status of Chinese cabbage based on image recognition, comprising the following steps:

[0028] The image acquisition module is used to collect images of Chinese cabbage and simultaneously obtain environmental parameters and growth status parameters;

[0029] The image preprocessing module is used to annotate and enhance the data, and the growth state parameters are quantified into feature vectors as label data;

[0030] The pre-processed image is input into the improved deep learning network, and the growth state parameters are incorporated into the network operation process for feature extraction;

[0031] The growth status monitoring module uses the evaluation model constructed based on the extracted features and growth status parameters to determine the growth stage, health status and nutritional status of the Chinese cabbage;

[0032] The image data, growth status parameters, feature extraction results and evaluation results are stored in the user management module, and an early warning is issued through the early warning module based on the evaluation results.

[0033] Furthermore, when training the improved deep learning network, a multi-task joint training strategy is adopted to optimize the classification task loss function L C And the regression task loss function L R , the total loss function L T The calculation formula is:

[0034] L T =λ1·L C +λ2·L R

[0035] Among them, L C For health status classification tasks, L R It is used for growth state parameter prediction tasks, where λ1 and λ2 are weight coefficients. The total loss function is minimized by the back propagation algorithm, and the network parameters are iteratively updated.

[0036] Furthermore, the growth status monitoring module regularly updates and optimizes the evaluation model by collecting new image data and growth status parameters and fine-tuning the model using an incremental learning algorithm; when the amount of new data reaches a preset threshold, a transfer learning algorithm is used to retrain the original model.

[0037] Through the above technical solutions, the beneficial effects of the present invention are as follows: This application integrates multi-source data to overcome the defects of traditional sensor monitoring and existing image recognition technology. Through image acquisition and collaborative work with the environmental sensor group, comprehensive growth status parameters and growth environment parameter data are obtained; the image preprocessing module combines growth status parameters for annotation and data enhancement to improve data quality; the improved YOLOv8 neural network incorporates growth parameters in the feature extraction process to enhance the feature extraction capability of complex field images; the growth status monitoring module constructs a multimodal fusion neural network evaluation model to accurately judge the growth stage, health status and nutritional status, and generate targeted suggestions; at the same time, the system also has the function of dynamic optimization and early warning of the model. Overall, accurate monitoring and dynamic analysis of the growth status of Chinese cabbage are achieved, providing scientific and effective decision-making support for planting management, reducing interference from environmental factors, and improving monitoring accuracy and reliability. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention, and the embodiments in the drawings do not constitute any limitation to the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0039] Figure 1 A schematic diagram of the structure of a Chinese cabbage growth status monitoring system based on image recognition provided in one embodiment of the present application;

[0040] Figure 2 A schematic diagram of the improved YOLOv8 neural network structure provided in one embodiment of the present application;

[0041] Figure 3 A schematic diagram of a process for monitoring the growth status of Chinese cabbage based on image recognition according to an embodiment of the present application;

[0042] The realization of the objectives, functional features and advantages of this application will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0043] It should be understood that the specific embodiments described herein are merely for explaining the present application and are not intended to limit the present application. Rather, these embodiments are provided to make the present disclosure more thorough and complete and to fully convey the scope of the present disclosure to those skilled in the art.

[0044] The above and other technical contents, features and effects of the present invention are described below with reference to the attached Figure 1-3 The detailed description of the embodiments will clearly show that the structural contents mentioned in the following embodiments are all based on the accompanying drawings.

[0045] Various exemplary embodiments of the present invention will be described below with reference to the accompanying drawings.

[0046] The main solutions of the embodiments of this application are:

[0047] like Figure 1 As shown, a Chinese cabbage growth status monitoring system based on image recognition provided by an embodiment of the present application includes an image acquisition module, an image preprocessing module, a feature extraction module, a growth status monitoring module and a user management module.

[0048] Among them, the image acquisition module is composed of a high-definition camera array and an environmental sensor group. The high-definition camera array is arranged above and on the side of the Chinese cabbage planting area, and can collect high-definition images from different angles to ensure that comprehensive plant morphological information is obtained; qualitative monitoring data include cotyledon size, cotyledon color, outer leaf color, outer leaf anthocyanin color, outer leaf gloss, outer leaf back hair, outer leaf longitudinal section shape, outer leaf edge wavy, outer leaf edge notch, outer leaf bubble size, outer leaf bubble number, outer leaf vein thickness, plant growth habit, plant height, plant spread, plant side buds, leaf ball top closure type, leaf ball upper green color, leaf ball shape, leaf ball height and leaf ball width. The environmental sensor group includes a light intensity sensor, a temperature sensor and a humidity sensor, which collect light intensity I in real time. i , temperature T i and humidity H i It, together with the image acquisition time t and the Chinese cabbage growth status parameters, constitutes the image metadata.

[0049] In addition, the module features an intelligent shooting mode that dynamically adjusts the shooting frequency based on the Chinese cabbage growth cycle. For example, it can shoot every two hours during the seedling stage, every four hours during the rosette stage, and every six hours during the head stage, adapting to the monitoring needs of different growth stages.

[0050] When the image preprocessing module annotates the collected images, it adds plant height H, number of leaves N, leaf area S, spread D and chlorophyll content C chl Quantify the parameters of the isogeny state into multidimensional feature vectors As label data for supervised learning. At the same time, an adaptive data enhancement algorithm is used to adjust the enhancement strategy according to the environmental parameters and growth state parameters in the image metadata. For example, when the ambient temperature T i When the leaf chlorophyll content C chl When the value is below the threshold, a color change data augmentation operation simulating nutrient deficiency is added to improve data diversity and model generalization ability.

[0051] The feature extraction module uses an improved deep learning network as the improved YOLOv8 neural network, such as Figure 2 As shown, it includes a backbone network, a neck network and a head network; the backbone network adopts an improved C2f structure and incorporates the growth state parameters into the feature extraction process through the following formula:

[0052]

[0053] Among them, n is the number of residual blocks, α is the weight coefficient, FC is the fully connected layer operation, F in is the input image feature map, F outTo output the feature map, this structure is used to enhance the network's ability to express the growth characteristics of Chinese cabbage.

[0054] The neck network adopts the BiFPN structure. During the feature fusion process, the growth state parameter is introduced to dynamically adjust the feature weight. The calculation formula for the m-th layer feature fusion is:

[0055]

[0056] Among them, the weight ω l According to the chlorophyll content chl calculate:

[0057]

[0058] β is the adjustment coefficient, L is the number of input features, f l is the lth input feature, and this method is used to highlight the feature fusion related to health status.

[0059] The evaluation model constructed by the growth status monitoring module is a multimodal fusion neural network, and the input includes the feature vector extracted by the improved YOLOv8 neural network and growth state parameter vector The health status score S is calculated by the following formula health :

[0060]

[0061] in, is the weight matrix, ω1 and ω2 are weight coefficients, and ReLU is the linear rectification function; the growth status monitoring module is based on the health status score S health With the preset threshold S high 、S low The comparison results are used to determine whether the Chinese cabbage is in a healthy, sub-healthy or unhealthy state.

[0062] Furthermore, the growth status monitoring module also calculates the nutritional supplement recommendation coefficient S according to the health status assessment results and environmental parameters using the following formula: n :

[0063] S n =δ1·Norm(S health )+δ2·Norm(I i -I s )+δ3·Norm(T i -T s )+δ4·Norm(H i -H s )

[0064] Among them, I i 、Ti and H i are the growth environment parameters of light intensity, temperature and humidity, I s 、T s and H s are the optimal light intensity, temperature and humidity for the growth of Chinese cabbage, respectively; δ1, δ2, δ3 and δ4 are weight coefficients; and targeted fertilization and irrigation recommendations are generated based on the nutritional supplement recommendation coefficients.

[0065] The demonstration application was carried out at a 50-acre Chinese cabbage cultivation base. As for the image acquisition module, five high-definition cameras, specifically the Hikvision DS-2CD3T86FWDV2-I5S model, were evenly distributed above the base. They offer 4K resolution, a 30fps frame rate, and infrared night vision capabilities, meeting all-weather image acquisition requirements and capturing overall images of the Chinese cabbage growth from above. Eight cameras were set up at varying heights around the base to capture close-up images from various angles, ensuring clear capture of details such as leaf morphology and pest and disease spots.

[0066] Environmental sensors are evenly distributed throughout each sub-area. These sensors include a light intensity sensor (TSL2561, measuring range 0-65535 Lux, accuracy ±5%) for real-time monitoring of light intensity changes; a temperature sensor (DS18B20, measuring range -55°C-125°C, accuracy ±0.5°C) for accurate ambient temperature; and a humidity sensor (HIH-4000-3-S, measuring range 0-100% RH, accuracy ±2% RH) for monitoring air humidity. These environmental sensors collect data once per minute, precisely synchronized with image acquisition time, to form image metadata.

[0067] The image acquisition module, image preprocessing module, feature extraction module, growth status monitoring module, user management module, and early warning module are deployed on a high-performance server cluster in the base's data center. The server cluster consists of eight Dell PowerEdge R740xd servers, each equipped with an Intel Xeon Gold 6248R processor (2.5GHz, 20 cores), 128GB of memory, and a 2TB solid-state drive. A 10 Gigabit Ethernet switch enables high-speed data transmission and interaction, ensuring the system's efficient and stable operation despite large data volumes.

[0068] During the Chinese cabbage seedling stage (days 1-20 after planting), the image acquisition module automatically triggers an image acquisition task every two hours, strictly following the programmed intelligent capture mode. During each acquisition, the camera in each sub-area captures at least 20 high-definition images from different angles. This totals 240 images per day per sub-area, bringing the total daily image collection for the entire base to 9,600. The environmental sensor group simultaneously collects light intensity, temperature, and humidity data, and associates and stores this data with the corresponding images.

[0069] The image preprocessing module utilizes the professional image annotation tool LabelMe. A team of agricultural experts, botanical researchers, and technicians meticulously annotated the captured images. During the annotation process, specialized measurement tools and methods were employed to obtain various growth parameters of the Chinese cabbage: plant height was measured using a high-precision laser rangefinder (±1mm accuracy); leaf count was performed manually with the aid of image analysis software; leaf area was determined using ImageJ software through threshold segmentation and area calculation; and spread was measured by converting image pixels into actual distance.

[0070] At the same time, the chlorophyll content was measured by a portable chlorophyll meter (SPAD-502Plus model) on randomly selected sample leaves in the field. A regression model was then established based on the image color features to estimate the overall chlorophyll content. Texture features such as leaf roughness, contrast, and directionality were calculated and extracted using the gray-level co-occurrence matrix (GLCM) algorithm in Matlab. Finally, these growth state parameters were quantified into a multidimensional feature vector. As labeled data for supervised learning.

[0071] The feature extraction module uses an improved YOLOv8 neural network. In the improved C2f architecture of the backbone network, the number of residual blocks, n, is set to 5, and the weight coefficient, α, is set to 0.4. The neck network uses a BiFPN architecture with an adjustment coefficient, β, of 0.08. This dynamically adjusts feature weights based on chlorophyll content, highlighting the integration of features related to health status.

[0072] The multimodal fusion neural network evaluation model constructed by the growth status monitoring module sets the weight matrix Weight coefficients ω1 = 0.6, ω2 = 0.4. By calculating the health status score S health and the preset threshold S high =80, S how =65, and judge the health status of Chinese cabbage. For example, when the health status score of Chinese cabbage in a certain area is detected, health =60, in an unhealthy state, according to the nutritional supplementation recommendation coefficient formula, combined with the current light intensity I i=12000Lux, temperature T i =28℃, humidity H i =70%, and the optimal growth parameter I s =15000Lux, T s =25℃、H s =60%, calculate the nutritional supplement recommendation coefficient S n , which in turn generates fertilization and irrigation recommendations.

[0073] In one specific example, the Chinese cabbage variety "Qingza No. 3" was selected, and field management operations such as sowing, fertilization, and irrigation were uniformly performed to ensure consistency and comparability of the test samples. A total of 5,000 Chinese cabbage plants were planted for monitoring. Traditional manual inspection methods and existing monitoring technology based on YOLOv5s image recognition were used for comparison.

[0074] During the experiment, different growth cycles and environmental conditions were artificially simulated. This included setting different combinations of temperature (10°C, 15°C, 20°C, 25°C), humidity (40% RH, 60% RH, 80% RH), and light intensity (5000 Lux, 10000 Lux, 15000 Lux), simulating different weather conditions such as sunny, cloudy, and rainy days, as well as simulating environments with high incidence of pests and diseases.

[0075] After monitoring a growth cycle, the test data was recorded and analyzed in detail. The comparison of the core indicators of monitoring performance is shown in the following table:

[0076]

[0077] Compared with traditional manual inspections and existing YOLOv5s image recognition technology, the proposed system significantly improves accuracy across all monitoring indicators. The accuracy of growth stage determination increased by an average of 32.9%, health status diagnosis increased by an average of 23.7%, and nutritional status assessment increased by an average of 27.3%, demonstrating its superior precision monitoring capabilities.

[0078] The dynamic adaptability test results of the system of the present invention, as shown in the table below, demonstrate its excellent performance and high adaptability under different environments and conditions.

[0079]

[0080] In the aforementioned simulated tests under various environmental conditions, the accuracy of the system fluctuated narrowly, remaining within ±3.2%. For example, in a low-temperature environment, while the accuracy declined, it remained at 92.0%, demonstrating that the system can effectively adapt to different growth cycles and environmental conditions, with minimal interference from environmental factors on the monitoring results.

[0081] Based on the same inventive concept, the present invention proposes a method for monitoring the growth status of Chinese cabbage based on image recognition, such as Figure 3 As shown, the following steps are included:

[0082] Step S1, using an image acquisition module to acquire a Chinese cabbage image, and synchronously acquiring environmental parameters and growth status parameters;

[0083] Step S2: annotate and perform data enhancement on the image through an image preprocessing module, and quantify the growth state parameters into feature vectors as label data;

[0084] Step S3: input the pre-processed image into the improved deep learning network, and integrate the growth state parameters into the network operation process to perform feature extraction;

[0085] Step S4: using the evaluation model constructed by the growth status monitoring module based on the extracted features and growth status parameters to determine the growth stage, health status, and nutritional status of the Chinese cabbage;

[0086] Step S5: storing the image data, growth state parameters, feature extraction results and evaluation results in the user management module, and issuing an early warning through the early warning module according to the evaluation results.

[0087] When training the improved deep learning network, a multi-task joint training strategy is adopted to optimize the classification task loss function L C And the regression task loss function L R , the total loss function L T The calculation formula is:

[0088] L T =λ1·L C +λ2·L R

[0089] Among them, L C For health status classification tasks, L R It is used for the growth status parameter prediction task, where λ1 and λ2 are weight coefficients; the classification task is mainly used to judge the health status of Chinese cabbage, and the cross entropy loss function is used to calculate the difference between the predicted result and the true label; the regression task is used to predict the growth status parameters, and the mean square error loss function is used to measure the error between the predicted value and the actual value.

[0090] Through the back propagation algorithm, the total loss function L TThe gradients of the network are propagated to each parameter of the network, and the network parameters are iteratively updated to minimize the overall loss function. During training, a large-scale annotated dataset is used. The dataset contains images of Chinese cabbage at different growth stages and under different environmental conditions, along with their corresponding accurate growth parameters and health status labels. After multiple rounds of training, the network continuously learns the mapping between image features and growth parameters, improving the accuracy of monitoring the growth status of Chinese cabbage.

[0091] The growth status monitoring module regularly updates and optimizes the evaluation model. Using an incremental learning algorithm, the model is fine-tuned as new labeled data is generated, enabling it to quickly adapt to new growth status characteristics and environmental changes. When the amount of new data reaches a preset threshold, a transfer learning algorithm is used to retrain the original model, using the parameters of the pre-trained model as initialization parameters for further training on the new dataset. This improves the model's adaptability to environmental changes and dynamic evolution of growth status, ensuring that the monitoring system maintains efficient and accurate performance.

[0092] In the intelligent decision support test, this application verifies the effectiveness of the decision-making plan by setting different growth abnormalities for the nutritional supplement recommendations and warning information generated by the system, as shown in the following table:

[0093]

[0094]

[0095] During the experiment, the abnormal growth of Chinese cabbage due to nitrogen deficiency was set. In response to this situation, the system generated 120 nutritional supplement recommendations and warning messages, of which 112 were accurate, with an accuracy rate of 93.3%. After fertilizing according to the system's recommendations, the growth of the Chinese cabbage was significantly improved. The plant height growth rate increased by 25%, indicating that an adequate nitrogen supply promoted the vertical growth of the plant; the number of leaves increased by 20%, indicating that nitrogen has a positive effect on leaf differentiation and formation; the leaf area changed by 18%, reflecting that nitrogen contributes to leaf expansion and development; and the incidence of pests and diseases decreased by 42%, because when nitrogen is sufficient, the plant's resistance is enhanced and it is better able to resist the invasion of pests and diseases. This example fully demonstrates that the system can accurately diagnose and provide effective fertilization recommendations in the case of nitrogen deficiency, playing a key role in the growth and health of Chinese cabbage.

[0096] In a scenario simulating the onset of pests and diseases on Chinese cabbage leaves, the system generated 100 recommendations and warnings, with 92 correct responses, for a 92.0% accuracy rate. After the growers implemented the corresponding measures based on the system's prevention and control recommendations, the cabbage showed significant changes in growth indicators. Plant height growth increased by 20%, indicating that plant growth had returned to normal after pests and diseases were controlled; leaf count increased by 15%, indicating that plant vitality had been restored; and leaf area changed by 15%, demonstrating that leaves continued to grow normally after pest and disease control. The incidence of pests and diseases decreased by 38%, directly demonstrating that the system's warnings and prevention recommendations effectively reduced pest and disease damage to the plants and ensured the healthy growth of the cabbage.

[0097] In response to the problem of excessive water in Chinese cabbage, the system generated 80 decision recommendations, 75 of which were accurate, with an accuracy rate of 93.8%. After adjusting the irrigation strategy according to the system's recommendations, the growth of the Chinese cabbage improved. The plant height growth rate increased by 18%, indicating that after reasonable water control, plant growth was no longer inhibited; the number of leaves increased by 12%, indicating that plant growth gradually returned to normal; the leaf area changed by 12%, indicating that the leaves were able to grow and develop normally; the incidence of pests and diseases decreased by 35%, because after the water was appropriate, the plant growth environment was improved, reducing the possibility of pests and diseases breeding. This example verifies the system's decision support capabilities in water management, which can help growers adjust irrigation measures in a timely manner to ensure the growth of Chinese cabbage.

[0098] The abnormal growth condition of Chinese cabbage with too little water was set. The system generated 100 suggestions and warnings, 93 of which were accurate, with an accuracy rate of 93.0%. After increasing the irrigation amount according to the system's suggestions, the growth condition of the Chinese cabbage improved significantly. The plant height growth rate increased by 22%, indicating that the plant grew rapidly after sufficient water; the increase in the number of leaves was 16%, reflecting the enhanced growth vitality of the plant; the change in leaf area was 16%, reflecting that the leaves can grow well under sufficient water supply; the incidence of pests and diseases decreased by 37%, indicating that suitable water conditions can help improve the resistance of the plant and reduce the occurrence of pests and diseases. This embodiment shows that the system can accurately judge and give effective water replenishment suggestions when there is too little water, thereby promoting the healthy growth of Chinese cabbage.

[0099] Based on the average data from various examples, the system's accuracy in generating nutritional supplement recommendations and warning messages for different growth anomalies reached 92.6%. After implementing these recommendations, various growth indicators of Chinese cabbage, such as plant height growth rate, average increase of 21.25%, average increase of leaf number by 15.75%, and average change of leaf area by 15.25%, were improved, and the incidence of pests and diseases decreased by an average of 38%. This fully demonstrates the scientific effectiveness of the decision-making solutions generated by the image recognition monitoring system based on Chinese cabbage growth status parameters, providing strong support for planting management and effectively improving the growth quality and yield of Chinese cabbage.

[0100] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0101] This document uses specific examples to illustrate the principles and implementation methods of this application. The description of the above examples is only intended to help understand the method and core concept of this application. At the same time, for those skilled in the art, based on the concept of this application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. An image recognition monitoring system based on Chinese cabbage growth status parameters, comprising an image acquisition module, an image preprocessing module, a feature extraction module, a growth status monitoring module and a user management module, characterized in that: The image acquisition module includes a camera array and an environmental sensor group. The camera array is used to capture images of Chinese cabbage from different angles to obtain Chinese cabbage growth status parameters; the environmental sensor group is used to capture Chinese cabbage growth environment parameters. The growth status parameters, growth environment parameters, and image acquisition time together constitute image metadata. The image preprocessing module is used to perform labeling and data enhancement processing on the collected images, and quantify the growth state parameters into feature vectors as label data; The feature extraction module adopts an improved deep learning network to integrate the growth state parameters into the network structure and calculation process for feature extraction and analysis of Chinese cabbage images; The growth status monitoring module constructs an evaluation model based on the features extracted by the feature extraction module and the growth status parameters to determine the growth stage, health status and nutritional status of the Chinese cabbage; The user management module is used to store image data, growth state parameters, feature extraction results and evaluation results.

2. The image recognition monitoring system based on Chinese cabbage growth status parameters according to claim 1, characterized in that: When the image preprocessing module annotates the collected image, the growth state parameters of the Chinese cabbage are quantified into a multidimensional feature vector The growth state parameters include plant height H, number of leaves N, leaf area S, spread D and chlorophyll content C chl , and use the multidimensional feature vector as label data for supervised learning.

3. The image recognition monitoring system based on Chinese cabbage growth status parameters according to claim 2, characterized in that: The improved deep learning network is an improved YOLOv8 neural network, including a backbone network, a neck network, and a head network. The backbone network adopts an improved C2f structure and incorporates growth state parameters into the feature extraction process through the following formula: Among them, n is the number of residual blocks, α is the weight coefficient, FC is the fully connected layer operation, F in is the input image feature map, F out is the output feature map.

4. The image recognition monitoring system based on Chinese cabbage growth status parameters according to claim 3, characterized in that: The neck network adopts the BiFPN structure. During the feature fusion process, the growth state parameter is introduced to dynamically adjust the feature weight. The calculation formula for the m-th layer feature fusion is: Among them, the weight ω l According to the chlorophyll content chl calculate: β is the adjustment coefficient, L is the number of input features, f l is the lth input feature.

5. The image recognition monitoring system based on Chinese cabbage growth status parameters according to claim 4, characterized in that: The evaluation model constructed by the growth status monitoring module is a multimodal fusion neural network, and the input includes the feature vector extracted by the improved YOLOv8 neural network and growth state parameter vector The health status score S is calculated by the following formula health : in, is the weight matrix, ω1 and ω2 are weight coefficients, and ReLU is the linear rectification function.

6. The image recognition monitoring system based on Chinese cabbage growth status parameters according to claim 5, characterized in that: The growth status monitoring module also calculates the nutritional supplement recommendation coefficient S according to the health status assessment results and environmental parameters through the following formula: n : S n =δ1·Norm(S health )+δ2·Norm(I i -I s )+δ3·Norm(T i -T s )+δ4·Norm(H i -H s ) Among them, I i 、T i and H i are the growth environment parameters of light intensity, temperature and humidity, I s 、T s and H s are the optimal light intensity, temperature and humidity for the growth of Chinese cabbage, and δ1, δ2, δ3 and δ4 are weight coefficients.

7. The image recognition monitoring system based on Chinese cabbage growth status parameters according to claim 6, characterized in that: The system also includes an early warning module, which issues an early warning when it determines that the Chinese cabbage is in an unhealthy state based on the evaluation result of the growth status monitoring module.

8. A method for monitoring the growth status of Chinese cabbage based on image recognition, characterized in that: The monitoring system according to any one of claims 1 to 7 comprises the following steps: The image acquisition module is used to collect images of Chinese cabbage and simultaneously obtain environmental parameters and growth status parameters; The image preprocessing module is used to annotate and enhance the data, and the growth state parameters are quantified into feature vectors as label data; The pre-processed image is input into the improved deep learning network, and the growth state parameters are incorporated into the network operation process for feature extraction; The growth status monitoring module uses the evaluation model constructed based on the extracted features and growth status parameters to determine the growth stage, health status and nutritional status of the Chinese cabbage; The image data, growth status parameters, feature extraction results and evaluation results are stored in the user management module, and an early warning is issued through the early warning module based on the evaluation results.

9. The method for monitoring the growth status of Chinese cabbage based on image recognition according to claim 8, characterized in that: When training the improved deep learning network, a multi-task joint training strategy is adopted to optimize the classification task loss function L C And the regression task loss function L R , the total loss function L T The calculation formula is: L T =λ1·L C +λ2·L R Among them, L C For health status classification tasks, L R It is used for growth state parameter prediction tasks, where λ1 and λ2 are weight coefficients. The total loss function is minimized by the back propagation algorithm, and the network parameters are iteratively updated.

10. The method for monitoring the growth status of Chinese cabbage based on image recognition according to claim 9, characterized in that: The growth status monitoring module regularly updates and optimizes the evaluation model by collecting new image data and growth status parameters and fine-tuning the model using an incremental learning algorithm. When the amount of new data reaches a preset threshold, a transfer learning algorithm is used to retrain the original model.

Images