Sapling planting growth evaluation method based on image recognition parameter acquisition

Abstract

The invention discloses a sapling planting growth evaluation method based on image recognition parameter acquisition, relates to the technical field of sapling planting growth evaluation, and aims to solve the problem that the sapling survival rate is reduced due to inaccurate sapling growth cycle evaluation. According to the method, visual information is converted into quantitative data through single plant marking and feature binding, the objectivity and accuracy of growth state evaluation are greatly improved, a high-quality data foundation is laid for subsequent analysis, the leaf withering and yellowing risk is pre-judged through correlation of temperature fluctuation and soil humidity, a prospective management basis is provided for planters, and the method is suitable for large-scale popularization and application. According to the method, the timeliness and scientificity of growth regulation and control are remarkably improved, the withered and yellow risk of the leaves is pre-judged through correlation of temperature fluctuation and soil humidity, a prospective management basis is provided for planters, and the timeliness and scientificity of growth regulation and control are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of seedling planting growth assessment technology, specifically a seedling planting growth assessment method based on image recognition parameters. Background Technology

[0002] Traditional tree seedling growth assessments rely heavily on manual observation, which suffers from low efficiency, strong subjectivity, and limited coverage, making it difficult to meet the precision management needs of large-scale planting. Existing technologies sometimes employ single-sensor monitoring, but the data dimensions are limited, failing to comprehensively correlate seedling morphology, physiological characteristics, and environmental factors. When image recognition technology is applied, improper preprocessing and incomplete feature extraction often lead to significant biases in growth status analysis. Furthermore, current assessments tend to focus on current status judgments, lacking dynamic correlation with environmental parameters and forward-looking predictions, making it difficult to provide early warnings of growth risks and hindering timely optimization of planting plans. Therefore, a comprehensive assessment method that integrates multi-source data and combines accuracy and predictive capabilities is urgently needed. Summary of the Invention

[0003] The purpose of this invention is to provide a method for evaluating the growth of seedlings based on image recognition parameters. By marking individual plants and binding them with features, visual information is transformed into quantitative data, significantly improving the objectivity and accuracy of growth status assessment and laying a high-quality data foundation for subsequent analysis. By correlating temperature fluctuations with soil moisture to predict the risk of leaf yellowing, it provides growers with a forward-looking management basis and significantly improves the timeliness and scientific nature of growth regulation. This invention can solve the problems in existing technologies.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] A method for evaluating seedling growth based on image recognition parameters includes:

[0006] Sensors are deployed according to the planting species in the seedling planting area; the deployed sensors are functionally tested; images of the seedlings' growth status are acquired using the sensors; the acquired growth status images are preprocessed; features are extracted from the preprocessed growth status images; the physiological characteristics of the seedlings are analyzed based on the extracted features; a growth prediction model for the seedlings is established based on the analyzed physiological characteristics, and image parameters are correlated with actual growth indicators; the established growth prediction model is correlated with environmental parameters, and the growth status of the seedlings is evaluated based on the correlation results; and decision-making suggestions for the seedling planting plan are generated based on the evaluation results.

[0007] Preferably, sensors are deployed in the seedling planting area according to the planting type, including:

[0008] First, identify the type of sensor, including image acquisition sensors and environmental parameter sensors;

[0009] Among them, image acquisition sensors include: high-definition cameras, multispectral cameras, and infrared thermal imagers; environmental parameter sensors include: air temperature and humidity sensors, soil temperature and humidity sensors, light intensity sensors, soil nutrient sensors, and oxygen concentration sensors.

[0010] Based on the key characteristics of the planted seedlings, sensors were selected and matched in number. The key characteristics include the range of plant height, crown size, planting density, main root distribution depth, and monitoring stage in the growth cycle.

[0011] Install each sensor according to the selected sensor and the number of matching sensors;

[0012] After installation, a fixation check is performed. Once the check is completed and the sensor passes inspection, the sensor deployment is complete.

[0013] Preferably, functional testing of the deployed sensors includes:

[0014] Before acquiring images of the seedlings' growth status, the function of each sensor is first tested.

[0015] The functional testing process is as follows: First, check the power supply and hardware connection of image acquisition sensors and environmental parameter sensors respectively;

[0016] After power supply and hardware connection checks are completed, perform basic connection tests.

[0017] The basic connectivity tests include: testing the field of view and coverage, clarity and color, and stability of the high-definition camera; testing the band effectiveness and visible light alignment of the multispectral camera; testing the temperature accuracy and thermal image integrity of the infrared thermal imager; testing the reasonableness of readings and response speed of the air temperature and humidity sensor and oxygen concentration sensor; testing the soil temperature and humidity sensor and soil nutrient sensor; and testing the light intensity sensor and oxygen concentration sensor.

[0018] After all basic connections have been tested and passed, the sensor to be collected is obtained.

[0019] Preferably, the growth status images of the seedlings are acquired using sensors, including:

[0020] Before collecting images of the seedlings' growth status, a collection plan is first formulated. The collection plan is as follows: the collection frequency is determined according to the seedling growth cycle, including the seedling stage, the shoot emergence stage, and the dormant stage. Then, the collection time period is set. After the collection time period is set, the collection plan is finalized.

[0021] After the data acquisition plan was formulated, the acquisition parameters of the sensors were set. The resolution of the high-definition camera was set to 1920×1080, the frame rate was set to 25fps, and the white balance was set to automatic. The fixed bands of the multispectral camera were combined and set, and the integration time was adjusted according to the light intensity. The temperature measurement range of the infrared thermal imager was set to 0-40℃, and the emissivity was set to 0.95.

[0022] After the acquisition parameters are set, the automatic acquisition process is started. The automatic acquisition process is as follows: according to the acquisition plan, the data acquisition device is used to set a timed task. After the preset time is reached, the sensor is automatically woken up. First, the high-definition camera is started to take 3-5 static images of the planting area. Then, the multispectral camera is started to acquire multi-band images of the planting area. Finally, the infrared thermal imager takes 1 thermal image. The acquisition interval between the high-definition camera, the multispectral camera and the infrared thermal imager shall not exceed 30 seconds.

[0023] High-definition cameras, multispectral cameras, and infrared thermal imagers are used to collect images of the planting area and make preliminary quality assessments.

[0024] The preliminary quality assessment includes: scoring the clarity of images captured by high-definition cameras; confirming the band intensity of multispectral cameras; and randomly checking 10% of the captured images for motion blur, abnormal temperature jumps in thermal images, and whether the target seedling is completely included.

[0025] If the image quality fails to meet the standard for three consecutive acquisitions, image acquisition of the planting area will be suspended.

[0026] The acquired images are processed or adjusted based on the preliminary quality assessment results;

[0027] Finally, an image of the growth status is obtained.

[0028] Preferably, image preprocessing is performed on the acquired growth state images, including:

[0029] First, the growth status images are standardized in image format. After standardization, the size is set according to the characteristics of the seedlings.

[0030] The standardized growth state images are subjected to noise processing, including random noise removal and motion blur restoration for images acquired by high-definition cameras; band noise elimination for images acquired by multispectral cameras; and temperature fluctuation suppression for images acquired by infrared thermal imagers. Finally, the growth state images are restored by neighboring pixel interpolation, thus completing the noise processing of the growth state images.

[0031] Geometric correction is performed on the growth state image after noise processing, including correcting the shooting angle deviation and eliminating distortion of the growth state image, aligning the image acquired by the multispectral camera with the visible light image, and aligning the image acquired by the infrared thermal imager with the visible light image.

[0032] Image enhancement is performed on the growth state images after geometric correction. Specifically, contrast enhancement and color normalization are performed on images acquired by high-definition cameras; band normalization is performed on images acquired by multispectral cameras; and temperature range normalization is performed on images acquired by infrared thermal imagers.

[0033] Non-target regions are removed from the image-enhanced growing state image, and edge optimization is performed.

[0034] Image preprocessing of the growth state image is completed after edge optimization.

[0035] Preferably, feature extraction is performed on the preprocessed growth state image, including:

[0036] The planting area is located in the growth status image after image preprocessing. The planting area is located as follows: the growth status image after removing non-target areas is divided into individual plant boundaries by the row spacing and plant spacing features in the image. Then, the individual plant boundaries are marked with the main trunk as the core. After marking the individual plants, the located individual seedlings are obtained.

[0037] The core features of individual seedlings marked in the high-definition camera are extracted, including trunk features, leaf features, and crown features. The trunk features include plant height, stem diameter, and trunk uprightness; the leaf features include leaf number, leaf area, and average leaf color; and the crown feature is the crown diameter.

[0038] The spectral features of individual seedlings marked in the multispectral camera are extracted, including key band combination calculation, vegetation index quantification, and single-species spectral mean. Among them, the key band combination calculation includes chlorophyll-related features and water-related features; vegetation index quantification includes the calculation of normalized vegetation index and greenness index.

[0039] Temperature characteristics of individual seedlings marked in infrared thermal imagers are extracted, including basic leaf temperature parameters and identification of abnormal temperature regions. The basic leaf temperature parameters include average temperature and temperature standard deviation; the identification of abnormal temperature regions includes extraction of high-temperature areas and extraction of temperature gradients.

[0040] The extracted temperature features, spectral features and core features are integrated. The feature integration is to summarize the three extracted features to obtain the feature set of a single seedling. Then, the three features collected at the same time are bound according to the timestamp.

[0041] After binding is completed, the characteristic data of the planted seedlings are obtained.

[0042] Preferably, the physiological characteristics of the seedlings are analyzed based on the extracted features, including:

[0043] First, identify the physiological characteristics and their states, including growth state, nutritional state, water state, and health state.

[0044] The extracted features were associated with the confirmed state types, and the mapping relationship between the features and state types was confirmed. Specifically, growth status was associated with trunk features and crown features; nutrient status was associated with leaf features and spectral features; water status was associated with spectral features; and health status was associated with leaf features, infrared temperature features, and spectral anomaly data.

[0045] Based on the correlation results, a preliminary analysis was performed on each state type. The analysis of growth status was as follows: the growth rate was calculated using time-series data of plant height and stem diameter, and the balance was determined by the ratio of crown diameter to plant height. The analysis of nutrient status was as follows: the chlorophyll level was determined based on the numerical range of the normalized vegetation index. The analysis of water status was as follows: the leaf fullness was determined based on the reflectance threshold of water-related bands. The analysis of health status was as follows: the health level was determined by the average leaf temperature and leaf number obtained from infrared thermal imagers, whether they were lower than the average for the same growth stage, and whether there were any local abnormal colors.

[0046] Cross-validation is performed on the preliminary analysis results, and feature error correction is performed based on the cross-validation results;

[0047] The feature error correction results were combined with the physiological characteristics of different growth stages to obtain the analysis results for the seedling stage, the shoot emergence stage, and the dormancy stage.

[0048] Finally, the physiological characteristics of the seedlings were analyzed.

[0049] Preferably, a growth prediction model for seedlings is established based on the analyzed physiological characteristics, and image parameters are correlated with actual growth indicators, including:

[0050] Before establishing a growth prediction model, the prediction targets are first identified, including short-term prediction, medium-term prediction, and state prediction. Then, the parsed physiological characteristic data and the corresponding image parameters are used as input variables, and the actual growth indicators are used as output variables.

[0051] Retrieve historical data from the database and divide the retrieved historical data into sample sets;

[0052] A time series forecasting model is used for short-term and medium-term forecasting; a classification model is used for state forecasting, and then the adopted models are trained.

[0053] The input and output variables are correlated and mapped, and the correlation mapping results are embedded into the trained model to obtain a preliminary prediction model.

[0054] The accuracy of the initial prediction model is verified, and iterative optimization is performed based on the verification results;

[0055] After iterative optimization, a complete growth prediction model was obtained.

[0056] Preferably, the established growth prediction model is correlated with environmental parameters, and the growth status of the seedlings is evaluated based on the correlation results, including:

[0057] The correlation between the growth prediction model and environmental parameters is as follows: First, the environmental parameters are classified, including climate parameters, soil parameters and stress parameters. The correlation dimension is determined based on the environmental parameters. The environmental parameters are used as supplementary input variables of the model to correct the prediction results. Then, the suitability of the environmental parameters is inferred from the growth index output by the model.

[0058] After the growth prediction model is associated with environmental parameters, the environmental parameters are time-aligned with the input and output data of the growth prediction model according to the timestamp of data collection. After time alignment, a three-dimensional dataset of environmental parameters, physiological characteristics and predicted growth indicators is obtained.

[0059] The three-dimensional dataset is embedded into the growth prediction model, and the resulting logic for the linkage between seedling environment and growth is obtained.

[0060] The prediction logic is evaluated based on the set evaluation indicators, which are growth quality indicators, environmental adaptability indicators, and stress risk indicators.

[0061] Among them, the growth quality index is to compare the predicted growth index with the standard value under the same environment and calculate the compliance rate; the compliance rate is to analyze whether the environmental parameters are within the suitable range and to assess the matching degree between growth and environment; the stress risk index is to predict the probability of growth stagnation and leaf yellowing within the next 7 days based on the degree to which environmental parameters deviate from the threshold.

[0062] The scope of the assessment is confirmed based on the assessment indicators, including individual plant assessment and regional assessment.

[0063] Finally, the growth status of the seedlings within the assessment scope is classified into three levels based on the assessment indicators: high risk, medium risk, and low risk.

[0064] After the assessment and grading are completed, the final seedling growth status assessment data is obtained.

[0065] Preferably, decision-making suggestions for seedling planting plans are generated based on the evaluation results, including:

[0066] The obtained seedling growth status assessment data is broken down and the core data is identified. The core data includes the assessment level and problem type, and then the seedlings are prioritized according to the problem type and assessment level.

[0067] A decision-making framework is constructed based on the growth cycle of the seedlings and the types of problems, including nutrient deficiency, water stress, and uneven growth.

[0068] Based on the growth cycle and problem type in the decision-making recommendation framework, implementation measures are proposed, and the proposed implementation measures are integrated into a plan of action.

[0069] After the measures and plans are integrated, we obtain the types of problems encountered in each growth cycle of the seedlings and the corresponding measures and plans based on the types of problems.

[0070] The types of problems encountered in each growth cycle of the seedlings and the corresponding solutions are converted into reports and then transmitted to the display terminal for display.

[0071] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0072] 1. The seedling growth assessment method based on image recognition parameter acquisition provided by this invention achieves comprehensive data coverage from seedling morphology and physiological characteristics to environmental factors through customized deployment of image acquisition and environmental parameter sensors, combined with targeted image preprocessing and feature extraction. Compared with traditional manual observation or single sensor monitoring, this innovation can capture multi-dimensional indicators such as plant height, chlorophyll level, and soil nutrients. By marking individual plants and binding features, visual information is transformed into quantitative data, significantly improving the objectivity and accuracy of growth status assessment and laying a high-quality data foundation for subsequent analysis.

[0073] 2. The seedling planting growth assessment method based on image recognition parameter acquisition provided by this invention incorporates environmental parameters into the prediction model through bidirectional mapping. It corrects growth prediction results using climate and soil data, and infers environmental suitability through growth indicators, forming a "environment-growth" linkage logic. This innovation overcomes the limitations of traditional models that rely solely on biological characteristics, accurately predicting short-term growth trends and medium- to long-term stress risks. It also predicts the risk of leaf yellowing by correlating temperature fluctuations with soil moisture, providing growers with forward-looking management guidance and significantly improving the timeliness and scientific rigor of growth regulation.

[0074] 3. The seedling planting and growth assessment method based on image recognition parameters provided by this invention prioritizes seedlings according to assessment level and problem type, and constructs a decision-making framework based on the growth cycle, transforming complex data into targeted measures. For high-risk seedlings with insufficient nutrition during the shoot emergence period, a rapid fertilization plan is automatically generated and presented intuitively in the form of a report. This innovation simplifies the decision-making process, reduces the cost of manual intervention, and is particularly suitable for large-scale planting scenarios. It can accurately solve individual tree problems while coordinating regional management, achieving closed-loop optimization from data collection to practical guidance, and significantly improving planting management efficiency. Attached Figure Description

[0075] Figure 1 This is a schematic diagram of the seedling planting and growth assessment steps of the present invention. Detailed Implementation

[0076] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0077] To address the problems in existing technologies, such as mismatched sensor deployment with planting species, missing functional detection, and unreasonable data acquisition schemes, which lead to poor data quality, unreliability, and difficulty in effectively monitoring seedling growth, please refer to [link / reference needed]. Figure 1 This embodiment provides the following technical solution:

[0078] A method for evaluating seedling growth based on image recognition parameters includes:

[0079] Sensors are deployed according to the planting species in the seedling planting area; the deployed sensors are functionally tested; images of the seedlings' growth status are acquired using the sensors; the acquired growth status images are preprocessed; features are extracted from the preprocessed growth status images; the physiological characteristics of the seedlings are analyzed based on the extracted features; a growth prediction model for the seedlings is established based on the analyzed physiological characteristics, and image parameters are correlated with actual growth indicators; the established growth prediction model is correlated with environmental parameters, and the growth status of the seedlings is evaluated based on the correlation results; and decision-making suggestions for the seedling planting plan are generated based on the evaluation results.

[0080] Specifically, by customizing and deploying sensors according to planting types, the growth characteristics of different seedlings can be captured in a targeted manner, avoiding parameter deviations of general-purpose equipment. The functional testing process further ensures the stable operation of the sensors, improving data reliability from the source and laying a high-quality foundation for subsequent analysis. Image preprocessing removes noise and redundant information, allowing feature extraction to focus more on key growth indicators. Combining image recognition technology to analyze physiological characteristics, visual information such as leaf morphology and plant height changes is transformed into quantitative data, breaking through the subjective limitations of traditional manual observation and achieving objective assessment of growth status. The growth prediction model links image parameters with actual indicators, and can dynamically predict growth trends through real-time image data. After superimposing environmental parameter analysis, the influence weight of factors such as light and humidity on growth can be clarified, making the assessment results take into account both the seedling's own condition and external conditions, making them more valuable for reference. The assessment results are directly transformed into planting plan suggestions, realizing closed-loop management from data collection to decision output, reducing the cost of manual intervention, and improving the level of refined management of seedling planting, which is especially suitable for efficient monitoring and control in large-scale planting scenarios.

[0081] Sensors will be deployed in the seedling planting area according to the type of tree being planted, including:

[0082] First, identify the type of sensor, including image acquisition sensors and environmental parameter sensors;

[0083] Among them, image acquisition sensors include: high-definition cameras, multispectral cameras, and infrared thermal imagers; environmental parameter sensors include: air temperature and humidity sensors, soil temperature and humidity sensors, light intensity sensors, soil nutrient sensors, and oxygen concentration sensors.

[0084] Based on the key characteristics of the planted seedlings, sensors were selected and matched in number. The key characteristics include the range of plant height, crown size, planting density, main root distribution depth, and monitoring stage in the growth cycle.

[0085] Install each sensor according to the selected sensor and the number of matching sensors;

[0086] After installation, a fixation check is performed. Once the check is completed and the sensor passes inspection, the sensor deployment is complete.

[0087] Specifically, the system encompasses two main categories of sensors: image acquisition sensors and environmental parameter sensors. Image acquisition sensors include high-definition cameras (capturing details of appearance and morphology), multispectral cameras (analyzing physiological states), and infrared thermal imagers (monitoring energy metabolism), providing comprehensive recording of the seedlings' visual characteristics. Environmental parameter sensors cover key indicators such as air temperature and humidity, soil temperature and humidity, and light intensity, enabling three-dimensional monitoring of the growth environment. The combination of these two types of data forms a complete growth assessment data chain. Key characteristics of the seedlings (height range, crown size, etc.) guide sensor selection and quantity matching; for example, taller seedlings are suited to cameras installed at higher heights, and wider-crowned varieties require sensors with increased monitoring coverage. The deployment of sensors in deep soil for deep-rooted seedlings enhances the efficiency of data acquisition by ensuring a high degree of matching between equipment configuration and seedling characteristics. Post-installation inspection promptly identifies issues such as loosening and angular deviations, ensuring stable sensor operation in complex field environments and minimizing data loss due to equipment malfunctions. This provides continuous and reliable foundational data for subsequent growth assessments. A quantitative matching mechanism based on planting characteristics avoids sensor redundancy or insufficiency, significantly reducing deployment costs in large-scale planting scenarios. Furthermore, precise data acquisition lays a high-quality data foundation for subsequent growth assessments, maximizing monitoring effectiveness.

[0088] Perform functional testing on the deployed sensors, including:

[0089] Before acquiring images of the seedlings' growth status, the function of each sensor is first tested.

[0090] The functional testing process is as follows: First, check the power supply and hardware connection of image acquisition sensors and environmental parameter sensors respectively;

[0091] After power supply and hardware connection checks are completed, perform basic connection tests.

[0092] The basic connectivity tests include: testing the field of view and coverage, clarity and color, and stability of the high-definition camera; testing the band effectiveness and visible light alignment of the multispectral camera; testing the temperature accuracy and thermal image integrity of the infrared thermal imager; testing the reasonableness of readings and response speed of the air temperature and humidity sensor and oxygen concentration sensor; testing the soil temperature and humidity sensor and soil nutrient sensor; and testing the light intensity sensor and oxygen concentration sensor.

[0093] After all basic connections have been tested and passed, the sensor to be collected is obtained.

[0094] Specifically, completing the inspection before acquiring images of the growth state allows for the early detection of sensor malfunctions, preventing invalid data acquisition due to equipment problems. This ensures the reliability of subsequent evaluation data from the source, reducing the cost of later data re-acquisition or correction. Sensors are inspected separately for image acquisition and environmental parameter types, with customized inspection items designed based on the characteristics of each type of equipment. For example, the effectiveness of the focusing band for multispectral cameras is assessed, while soil sensors focus on soil parameter detection. This ensures that the inspection is more aligned with the essential functions of the equipment, accurately locating potential problems. From power supply and hardware connection checks to basic connection testing, a progressive inspection chain is formed. Basic connection issues are resolved first, followed by verification of core functions, avoiding omissions of critical links, ensuring comprehensive inspection, and improving overall inspection efficiency. Inspection content is designed for key indicators of various sensors, such as the field of view and clarity of high-definition cameras, and the temperature accuracy of infrared thermal imagers, comprehensively verifying whether the sensor functions meet standards and providing a solid guarantee for subsequent accurate data acquisition.

[0095] Using sensors to acquire images of the seedlings' growth status, including:

[0096] Before collecting images of the seedlings' growth status, a collection plan is first formulated. The collection plan is as follows: the collection frequency is determined according to the seedling growth cycle, including the seedling stage, the shoot emergence stage, and the dormant stage. Then, the collection time period is set. After the collection time period is set, the collection plan is finalized.

[0097] After the data acquisition plan was formulated, the acquisition parameters of the sensors were set. The resolution of the high-definition camera was set to 1920×1080, the frame rate was set to 25fps, and the white balance was set to automatic. The fixed bands of the multispectral camera were combined and set, and the integration time was adjusted according to the light intensity. The temperature measurement range of the infrared thermal imager was set to 0-40℃, and the emissivity was set to 0.95.

[0098] After the acquisition parameters are set, the automatic acquisition process is started. The automatic acquisition process is as follows: according to the acquisition plan, the data acquisition device is used to set a timed task. After the preset time is reached, the sensor is automatically woken up. First, the high-definition camera is started to take 3-5 static images of the planting area. Then, the multispectral camera is started to acquire multi-band images of the planting area. Finally, the infrared thermal imager takes 1 thermal image. The acquisition interval between the high-definition camera, the multispectral camera and the infrared thermal imager shall not exceed 30 seconds.

[0099] High-definition cameras, multispectral cameras, and infrared thermal imagers are used to collect images of the planting area and make preliminary quality assessments.

[0100] The preliminary quality assessment includes: scoring the clarity of images captured by high-definition cameras; confirming the band intensity of multispectral cameras; and randomly checking 10% of the captured images for motion blur, abnormal temperature jumps in thermal images, and whether the target seedling is completely included.

[0101] If the image quality fails to meet the standard for three consecutive acquisitions, image acquisition of the planting area will be suspended.

[0102] The acquired images are processed or adjusted based on the preliminary quality assessment results;

[0103] Finally, an image of the growth status is obtained.

[0104] Specifically, differentiated data collection frequencies are established based on different growth stages of the seedlings (seedling stage, shoot emergence stage, and dormancy stage) to meet the monitoring needs of each stage's growth characteristics. For example, high-frequency data collection is required during the active shoot emergence stage, while the frequency can be reduced during the dormancy stage to avoid redundant and invalid data. At the same time, precise data collection time periods are set to ensure that the acquired images accurately reflect the typical growth state of the seedlings, improving the effectiveness of the data in the temporal dimension. Parameters are customized for different image sensor characteristics: the high-definition camera uses a 1920×1080 resolution and a 25fps frame rate to balance detail capture and storage costs; the multispectral camera uses a fixed band combination and adaptive integration time based on illumination to ensure the stability of spectral data; the infrared thermal imager is set with a temperature measurement range of 0-40℃ and an emissivity of 0.95 to match the temperature range of seedling growth. The configuration is highly matched with the equipment functions and monitoring targets, providing high-quality raw data for feature extraction. Timed tasks enable orderly sensor wake-up, with high-definition cameras, multispectral cameras, and infrared thermal imagers collecting data sequentially at intervals of no more than 30 seconds, reducing environmental interference caused by time differences and ensuring the correlation of multi-dimensional data at the same time. At the same time, automated operation reduces labor costs and improves the collection efficiency in large-scale planting scenarios. Through multi-dimensional preliminary judgments such as clarity scoring, band intensity confirmation, and random sampling, problems such as motion blur, temperature jumps, and target missingness are promptly detected. Collection is paused and adjusted for areas that fail to meet the standards three times in a row. Invalid data is filtered from the collection end to avoid wasting subsequent processing resources, ensuring the availability and consistency of image data, and laying a solid data foundation for subsequent feature extraction and growth assessment.

[0105] To address the issues in existing seedling growth assessment technologies, such as messy image data formats, noise interference, unsystematic feature extraction, and one-sided physiological state analysis, which fail to accurately reflect the true state of growth and nutrition and thus lead to low assessment reliability, please refer to [the relevant documentation / reference]. Figure 1 This embodiment provides the following technical solution:

[0106] Image preprocessing is performed on the acquired growth status images, including:

[0107] First, the growth status images are standardized in image format. After standardization, the size is set according to the characteristics of the seedlings.

[0108] The standardized growth state images are subjected to noise processing, including random noise removal and motion blur restoration for images acquired by high-definition cameras; band noise elimination for images acquired by multispectral cameras; and temperature fluctuation suppression for images acquired by infrared thermal imagers. Finally, the growth state images are restored by neighboring pixel interpolation, thus completing the noise processing of the growth state images.

[0109] Geometric correction is performed on the growth state image after noise processing, including correcting the shooting angle deviation and eliminating distortion of the growth state image, aligning the image acquired by the multispectral camera with the visible light image, and aligning the image acquired by the infrared thermal imager with the visible light image.

[0110] Image enhancement is performed on the growth state images after geometric correction. Specifically, contrast enhancement and color normalization are performed on images acquired by high-definition cameras; band normalization is performed on images acquired by multispectral cameras; and temperature range normalization is performed on images acquired by infrared thermal imagers.

[0111] Non-target regions are removed from the image-enhanced growing state image, and edge optimization is performed.

[0112] Image preprocessing of the growth state image is completed after edge optimization.

[0113] Specifically, by standardizing image formats and sizes according to seedling characteristics, the differences in acquisition formats and image sizes across different devices are eliminated. This provides a consistent data benchmark for subsequent cross-device and cross-time period image comparison and analysis, avoiding feature extraction deviations caused by format or size inconsistencies and improving data compatibility. Differentiated noise reduction strategies are designed for the characteristics of images from different sensors: high-definition images focus on random noise and motion blur repair; multispectral images focus on band noise elimination; and infrared thermal images focus on suppressing temperature fluctuations. Combined with neighboring pixel interpolation repair, typical noise problems in various images are accurately solved, maximizing the preservation of effective growth information and reducing noise interference with feature analysis. By correcting shooting angle deviations, eliminating distortion, and ensuring precise alignment of multispectral, infrared, and visible light images, the problems caused by equipment installation angles are addressed. To address the spatial misalignment of images caused by changes in shooting position, this method ensures that the spatial position of the same seedling matches in different types of images, providing a foundation for spatial coordinate consistency in multimodal data fusion analysis. Optimization and enhancement methods are implemented for different image types: high-definition images improve contrast and color consistency; multispectral images achieve band normalization; and infrared images standardize the temperature range, effectively highlighting the morphological, physiological, and temperature characteristics of the seedling. This makes it easier to extract key information such as leaf texture, plant structure, and temperature distribution, improving the accuracy of subsequent feature analysis. Removing non-target areas and optimizing edges eliminates background interference from soil, weeds, and other elements, accurately defining the main body of the seedling. Edge optimization also makes the seedling outline clearer, reducing interference from irrelevant information in the assessment of growth status and providing focused, high-quality image data for subsequent physiological feature analysis.

[0114] Feature extraction is performed on the preprocessed growth state image, including:

[0115] The planting area is located in the growth status image after image preprocessing. The planting area is located as follows: the growth status image after removing non-target areas is divided into individual plant boundaries by the row spacing and plant spacing features in the image. Then, the individual plant boundaries are marked with the main trunk as the core. After marking the individual plants, the located individual seedlings are obtained.

[0116] The core features of individual seedlings marked in the high-definition camera are extracted, including trunk features, leaf features, and crown features. The trunk features include plant height, stem diameter, and trunk uprightness; the leaf features include leaf number, leaf area, and average leaf color; and the crown feature is the crown diameter.

[0117] The spectral features of individual seedlings marked in the multispectral camera are extracted, including key band combination calculation, vegetation index quantification, and single-species spectral mean. Among them, the key band combination calculation includes chlorophyll-related features and water-related features; vegetation index quantification includes the calculation of normalized vegetation index and greenness index.

[0118] Temperature characteristics of individual seedlings marked in infrared thermal imagers are extracted, including basic leaf temperature parameters and identification of abnormal temperature regions. The basic leaf temperature parameters include average temperature and temperature standard deviation; the identification of abnormal temperature regions includes extraction of high-temperature areas and extraction of temperature gradients.

[0119] The extracted temperature features, spectral features and core features are integrated. The feature integration is to summarize the three extracted features to obtain the feature set of a single seedling. Then, the three features collected at the same time are bound according to the timestamp.

[0120] After binding is completed, the characteristic data of the planted seedlings are obtained.

[0121] Specifically, individual tree boundaries are defined using row spacing and plant spacing features, with the trunk as the core for marking. This enables precise positioning of individual seedlings within the planting area, avoiding interference between trees and providing a clear analytical unit for subsequent single-tree-level feature extraction and growth assessment. This is particularly suitable for densely planted scenarios, enhancing the relevance of data analysis. Differentiated key features are extracted from images from different devices: high-definition images focus on morphological features such as the trunk (height, stem diameter, etc.), leaves (number, area, etc.), and crown width; multispectral images extract physiologically related spectral features such as chlorophyll and water content, as well as vegetation indices; and infrared images capture energy metabolism features such as average temperature and abnormal areas. These three types of features comprehensively depict the seedling growth status from morphological, physiological, and metabolic perspectives, providing rich evidence for comprehensive assessment. Each type of feature closely aligns with key seedling growth indicators, such as spectral features. The chlorophyll-related calculations directly reflect photosynthetic capacity; the identification of abnormal areas in temperature features can provide early warning of stress conditions; and stem uprightness is related to plant health stability. The extraction logic is highly consistent with plant physiological characteristics, ensuring the clear biological significance of the features and improving the scientific rigor of the assessment. By summarizing the three types of features to form a single-plant feature set and binding data from the same time period according to timestamps, a systematic integration of single-plant features is achieved, while ensuring the temporal matching of features from different dimensions and avoiding data fragmentation. This provides a structured, time-series-based, high-quality data foundation for subsequently establishing a correlation model between image parameters and actual growth indicators. It also facilitates tracing the feature change patterns at different growth stages, improving the accuracy of growth prediction. The sensor acquisition parameters and feature extraction numerical standards for different growth cycles are shown in the following table:

[0122]

[0123]

[0124]

[0125] The physiological characteristics of the seedlings are analyzed based on the extracted features, including:

[0126] First, identify the physiological characteristics and their states, including growth state, nutritional state, water state, and health state.

[0127] The extracted features were associated with the confirmed state types, and the mapping relationship between the features and state types was confirmed. Specifically, growth status was associated with trunk features and crown features; nutrient status was associated with leaf features and spectral features; water status was associated with spectral features; and health status was associated with leaf features, infrared temperature features, and spectral anomaly data.

[0128] Based on the correlation results, a preliminary analysis was performed on each state type. The analysis of growth status was as follows: the growth rate was calculated using time-series data of plant height and stem diameter, and the balance was determined by the ratio of crown diameter to plant height. The analysis of nutrient status was as follows: the chlorophyll level was determined based on the numerical range of the normalized vegetation index. The analysis of water status was as follows: the leaf fullness was determined based on the reflectance threshold of water-related bands. The analysis of health status was as follows: the health level was determined by the average leaf temperature and leaf number obtained from infrared thermal imagers, whether they were lower than the average for the same growth stage, and whether there were any local abnormal colors.

[0129] Cross-validation is performed on the preliminary analysis results, and feature error correction is performed based on the cross-validation results;

[0130] The feature error correction results were combined with the physiological characteristics of different growth stages to obtain the analysis results for the seedling stage, the shoot emergence stage, and the dormancy stage.

[0131] Finally, the physiological characteristics of the seedlings were analyzed.

[0132] Specifically, physiological characteristics are clearly divided into four major state types: growth, nutrition, water, and health, forming an assessment system covering the core dimensions of seedling growth. This clear classification avoids confusion and omissions in feature analysis, ensuring that the assessment objectives for each state type are clear and laying a framework for subsequent precise analysis. This ensures a clear and hierarchical analysis process, matching specific features to different state types. For example, growth status is associated with trunk and crown characteristics, and nutrition status is linked to leaf and spectral characteristics, with the mapping relationship highly consistent with plant physiological laws. This precise association avoids feature misuse. For instance, using water-related bands in spectral characteristics to analyze water status directly points to the core indicator of leaf water content, improving the reliability of the analysis results. The analysis of each state type employs scientific quantitative methods: growth rate is calculated using time-series data, chlorophyll level is determined based on the normalized vegetation index range, and health level is comprehensively assessed by combining temperature, leaf quantity, and color anomalies. The analytical method considers both dynamic changes (such as growth rate) and threshold judgments (such as reflectivity threshold), ensuring that the results reflect trends and have clear judgment criteria, thus enhancing the scientific rigor and operability of the analysis. The cross-validation stage after preliminary analysis identifies errors through logical consistency between different features (such as the correlation between moisture status and health status), and reduces the bias of single-feature analysis by combining feature error correction. Furthermore, by integrating with characteristics of different growth stages, the analytical criteria are adjusted for differences in stages such as seedling and shoot emergence, making the results more closely reflect the physiological characteristics of seedlings at each stage. The final staged analysis results accurately reflect the true physiological state of seedlings at different growth stages, providing a high-quality basis for subsequent growth prediction.

[0133] To address the problems in existing technologies where seedling planting assessment relies on manual measurement, resulting in low efficiency, high subjectivity, difficulty in real-time correlation between growth status and environmental parameters, and inability to accurately predict growth trends and potential risks, leading to lagging and insufficiently targeted planting management, please refer to [the relevant documentation / reference]. Figure 1 This embodiment provides the following technical solution:

[0134] A growth prediction model for seedlings was established based on the analyzed physiological characteristics, and image parameters were correlated with actual growth indicators, including:

[0135] Before establishing a growth prediction model, the prediction targets are first identified, including short-term prediction, medium-term prediction, and state prediction. Then, the parsed physiological characteristic data and the corresponding image parameters are used as input variables, and the actual growth indicators are used as output variables.

[0136] Retrieve historical data from the database and divide the retrieved historical data into sample sets;

[0137] A time series forecasting model is used for short-term and medium-term forecasting; a classification model is used for state forecasting, and then the adopted models are trained.

[0138] The input and output variables are correlated and mapped, and the correlation mapping results are embedded into the trained model to obtain a preliminary prediction model.

[0139] The accuracy of the initial prediction model is verified, and iterative optimization is performed based on the verification results;

[0140] After iterative optimization, a complete growth prediction model was obtained.

[0141] Specifically, the prediction objectives are divided into short-term, medium-term, and state predictions, specifically covering different time dimensions and assessment needs. Short-term predictions can reflect recent growth dynamics in a timely manner, medium-term predictions assist in phased management planning, and state predictions focus on key state changes such as health. This hierarchical setting makes the model's functions clearer, meets diverse prediction needs in planting management, and enhances the practicality of the model application. The parsed physiological characteristic data and corresponding image parameters are used as input variables, and actual growth indicators are used as output variables, achieving a direct correlation from image information to growth results. Input variables cover the physiological state of seedlings and original image parameters, while output variables are anchored to real growth indicators. The variable system is complete and logically closed-loop, providing a high-quality foundation of correlated data for model learning, ensuring the relevance and reliability of model predictions. Dedicated models are selected for different prediction objectives: time-series prediction models excel at capturing the dynamic trends of short-term and medium-term growth, aligning with the continuous changes in indicators such as plant height and stem diameter; classification models are suitable for determining discrete states such as health level and nutrient level in state prediction. The model type is highly matched with the characteristics of the prediction target, improving prediction accuracy and avoiding the performance limitations of general models in complex prediction scenarios. By mapping the correlation between input and output variables and embedding the results into the model, a quantitative correlation is established between image parameters and actual growth indicators. For example, a correspondence is established between the average leaf color and chlorophyll content, transforming abstract image data into interpretable growth indicators. This not only enhances the utilization value of image parameters but also provides a traceable logical basis for the model's prediction results. After the initial model is verified for accuracy, iterative optimization is performed. By continuously adjusting parameters and correcting deviations, the model's prediction accuracy is gradually improved. This dynamic optimization mechanism can adapt to the differences in different seedling varieties and growth environments, enabling the model to maintain stable and reliable prediction performance in practical applications and providing strong support for subsequent growth assessment and planting decisions.

[0142] The established growth prediction model is correlated with environmental parameters, and the growth status of the seedlings is evaluated based on the correlation results, including:

[0143] The correlation between the growth prediction model and environmental parameters is as follows: First, the environmental parameters are classified, including climate parameters, soil parameters and stress parameters. The correlation dimension is determined based on the environmental parameters. The environmental parameters are used as supplementary input variables of the model to correct the prediction results. Then, the suitability of the environmental parameters is inferred from the growth index output by the model.

[0144] After the growth prediction model is associated with environmental parameters, the environmental parameters are time-aligned with the input and output data of the growth prediction model according to the timestamp of data collection. After time alignment, a three-dimensional dataset of environmental parameters, physiological characteristics and predicted growth indicators is obtained.

[0145] The three-dimensional dataset is embedded into the growth prediction model, and the resulting logic for the linkage between seedling environment and growth is obtained.

[0146] The prediction logic is evaluated based on the set evaluation indicators, which are growth quality indicators, environmental adaptability indicators, and stress risk indicators.

[0147] Among them, the growth quality index is to compare the predicted growth index with the standard value under the same environment and calculate the compliance rate; the compliance rate is to analyze whether the environmental parameters are within the suitable range and to assess the matching degree between growth and environment; the stress risk index is to predict the probability of growth stagnation and leaf yellowing within the next 7 days based on the degree to which environmental parameters deviate from the threshold.

[0148] The scope of the assessment is confirmed based on the assessment indicators, including individual plant assessment and regional assessment.

[0149] Finally, the growth status of the seedlings within the assessment scope is classified into three levels based on the assessment indicators: high risk, medium risk, and low risk.

[0150] After the assessment and grading are completed, the final seedling growth status assessment data is obtained.

[0151] Specifically, environmental parameters are categorized into three types: climate, soil, and stress. A two-way correlation mechanism—"supplementing inputs to correct prediction results + using growth indicators to infer environmental suitability"—breaks the limitation of growth prediction models relying solely on physiological characteristics. This correlation allows environmental factors to participate in the dynamic adjustment of growth predictions and also enables the reverse verification of environmental suitability through growth results, forming a complete logical chain of "environment-growth" feedback, thus improving the scientific rigor of the assessment. Based on timestamps, environmental parameters are aligned with model input and output data, constructing a three-dimensional dataset of environment, physiological characteristics, and predicted growth indicators, ensuring accurate matching of multi-dimensional data at the same time point. To avoid correlational biases caused by time misalignment and to make the analysis of the impact of environmental factors on growth more accurate, this system provides a reliable data foundation for linked prediction logic. Growth quality indicators measure growth effectiveness by comparing with standard values, environmental adaptability indicators assess the degree of matching between growth and the environment, and stress risk indicators provide early warnings of future growth risks. These three types of indicators construct a complete assessment framework from three levels: current status, adaptability, and potential risks. This comprehensive framework covers the core dimensions of growth status, avoiding the one-sidedness of single-indicator assessments. Individual plant assessments can accurately pinpoint problematic individuals, while regional assessments facilitate understanding of the overall growth trend and meet different management needs. The classification of high, medium, and low risk levels transforms complex assessment data into intuitive management guidelines, enabling growers to quickly identify key areas of concern, develop differentiated intervention strategies, and improve management efficiency. Simultaneously, the probability prediction of stress risk within 7 days provides a time window for taking preventative measures in advance, reducing the risk of growth loss and strengthening the practical guiding significance of the assessment results.

[0152] Based on the evaluation results, decision-making recommendations for seedling planting plans are generated, including:

[0153] The obtained seedling growth status assessment data is broken down and the core data is identified. The core data includes the assessment level and problem type, and then the seedlings are prioritized according to the problem type and assessment level.

[0154] A decision-making framework is constructed based on the growth cycle of the seedlings and the types of problems, including nutrient deficiency, water stress, and uneven growth.

[0155] Based on the growth cycle and problem type in the decision-making recommendation framework, implementation measures are proposed, and the proposed implementation measures are integrated into a plan of action.

[0156] After the measures and plans are integrated, we obtain the types of problems encountered in each growth cycle of the seedlings and the corresponding measures and plans based on the types of problems.

[0157] The types of problems encountered in each growth cycle of the seedlings and the corresponding solutions are converted into reports and then transmitted to the display terminal for display.

[0158] Specifically, by breaking down assessment data and extracting assessment levels and problem types as core data, and prioritizing them, we ensure that decision-making recommendations directly address key issues. For example, high-risk nutrient deficiency issues are addressed first, avoiding wasting resources on low-priority matters. This allows decision-making resources to be tilted towards the most critical areas, improving management efficiency. A framework is constructed based on the seedling growth cycle and problem type, developing appropriate recommendations for different stages such as the seedling stage and the shoot emergence stage, addressing specific issues like nutrient deficiency and water stress. For instance, during the vigorous growth period of shoot emergence, recommendations for nutrient deficiency focus on rapid fertilization, while during dormancy, recommendations emphasize gentle conditioning. This ensures that recommendations are highly aligned with the seedling growth rhythm, enhancing the scientific validity and feasibility of the measures. Scattered implementation recommendations are integrated into a complete plan, clearly defining the problems and solutions for each growth cycle, forming a closed-loop system of "problem-cycle-measure." This avoids the execution chaos caused by fragmented recommendations; growers can directly refer to the integrated plan for operation, reducing the connection costs between decision-making and execution, improving management consistency, and converting the plan into a report format displayed on the terminal, making complex decision-making logic and measure content clear and readable. Growers can quickly understand problem priorities, corresponding measures, and implementation timing without needing professional data analysis skills, lowering the technical threshold and ensuring that recommendations are effectively implemented. Meanwhile, the report format facilitates archiving and traceability, providing a reference for subsequent planting plan optimization and forming a continuous improvement management loop. Measures are customized for specific problem types, such as adjusting irrigation plans for water stress and focusing on pruning and light control for uneven growth, avoiding the ambiguity of general recommendations. Precise targeted measures can quickly alleviate seedling growth problems, shorten the recovery period, and improve planting effectiveness.

[0159] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0160] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

Claims

1. A method for evaluating seedling planting and growth based on image recognition parameters, characterized in that, include: Sensors will be deployed in the seedling planting area according to the type of tree being planted; Perform functional testing on the deployed sensors; use the sensors to acquire images of the seedlings' growth status; perform image preprocessing on the acquired growth status images; Feature extraction is performed on the preprocessed growth status images; the physiological characteristics of the seedlings are analyzed based on the extracted features; a growth prediction model for the seedlings is established based on the analyzed physiological characteristics, and the image parameters are correlated with actual growth indicators; the established growth prediction model is correlated with environmental parameters, and the growth status of the seedlings is evaluated based on the correlation results; and decision suggestions for the seedling planting plan are generated based on the evaluation results.

2. The method for evaluating seedling planting and growth based on image recognition parameters according to claim 1, characterized in that, Sensors will be deployed in the seedling planting area according to the type of tree being planted, including: First, identify the type of sensor, including image acquisition sensors and environmental parameter sensors; Among them, image acquisition sensors include: high-definition cameras, multispectral cameras, and infrared thermal imagers; environmental parameter sensors include: air temperature and humidity sensors, soil temperature and humidity sensors, light intensity sensors, soil nutrient sensors, and oxygen concentration sensors. Based on the key characteristics of the planted seedlings, sensors were selected and matched in number. The key characteristics include the range of plant height, crown size, planting density, main root distribution depth, and monitoring stage in the growth cycle. Install each sensor according to the selected sensor and the number of matching sensors; After installation, a fixation check is performed. Once the check is completed and the sensor passes inspection, the sensor deployment is complete.

3. The method for evaluating seedling planting and growth based on image recognition parameters according to claim 2, characterized in that, Perform functional testing on the deployed sensors, including: Before acquiring images of the seedlings' growth status, the function of each sensor is first tested. The functional testing process is as follows: First, check the power supply and hardware connection of image acquisition sensors and environmental parameter sensors respectively; After power supply and hardware connection checks are completed, perform basic connection tests. The basic connectivity tests include: testing the field of view and coverage, clarity and color, and stability of the high-definition camera; testing the band effectiveness and visible light alignment of the multispectral camera; testing the temperature accuracy and thermal image integrity of the infrared thermal imager; testing the reasonableness of readings and response speed of the air temperature and humidity sensor and oxygen concentration sensor; testing the soil temperature and humidity sensor and soil nutrient sensor; and testing the light intensity sensor and oxygen concentration sensor. After all basic connections have been tested and passed, the sensor to be collected is obtained.

4. The method for evaluating seedling planting and growth based on image recognition parameters according to claim 3, characterized in that, Using sensors to acquire images of the seedlings' growth status, including: Before collecting images of the seedlings' growth status, a collection plan is first formulated. The collection plan is as follows: the collection frequency is determined according to the seedling growth cycle, including the seedling stage, the shoot emergence stage, and the dormant stage. Then, the collection time period is set. After the collection time period is set, the collection plan is finalized. After the data acquisition plan was formulated, the acquisition parameters of the sensors were set. The resolution of the high-definition camera was set to 1920×1080, the frame rate was set to 25fps, and the white balance was set to automatic. The fixed bands of the multispectral camera were combined and set, and the integration time was adjusted according to the light intensity. The temperature measurement range of the infrared thermal imager was set to 0-40℃, and the emissivity was set to 0.

95. After the acquisition parameters are set, the automatic acquisition process is started. The automatic acquisition process is as follows: according to the acquisition plan, the data acquisition device is used to set a timed task. After the preset time is reached, the sensor is automatically woken up. First, the high-definition camera is started to take 3-5 static images of the planting area. Then, the multispectral camera is started to acquire multi-band images of the planting area. Finally, the infrared thermal imager takes 1 thermal image. The acquisition interval between the high-definition camera, the multispectral camera and the infrared thermal imager shall not exceed 30 seconds. High-definition cameras, multispectral cameras, and infrared thermal imagers are used to collect images of the planting area and make preliminary quality assessments. The preliminary quality assessment includes: scoring the clarity of images captured by high-definition cameras; confirming the band intensity of multispectral cameras; and randomly checking 10% of the captured images for motion blur, abnormal temperature jumps in thermal images, and whether the target seedling is completely included. If the image quality fails to meet the standard for three consecutive acquisitions, image acquisition of the planting area will be suspended. The acquired images are processed or adjusted based on the preliminary quality assessment results; Finally, an image of the growth status is obtained.

5. The method for evaluating seedling planting and growth based on image recognition parameters according to claim 4, characterized in that, Image preprocessing is performed on the acquired growth status images, including: First, the growth status images are standardized in image format. After standardization, the size is set according to the characteristics of the seedlings. The standardized growth state images are subjected to noise processing, including random noise removal and motion blur restoration for images acquired by high-definition cameras; band noise elimination for images acquired by multispectral cameras; and temperature fluctuation suppression for images acquired by infrared thermal imagers. Finally, the growth state images are restored by neighboring pixel interpolation, thus completing the noise processing of the growth state images. Geometric correction is performed on the growth state image after noise processing, including correcting the shooting angle deviation and eliminating distortion of the growth state image, aligning the image acquired by the multispectral camera with the visible light image, and aligning the image acquired by the infrared thermal imager with the visible light image. Image enhancement is performed on the growth state images after geometric correction. Specifically, contrast enhancement and color normalization are performed on images acquired by high-definition cameras; band normalization is performed on images acquired by multispectral cameras; and temperature range normalization is performed on images acquired by infrared thermal imagers. Non-target regions are removed from the image-enhanced growing state image, and edge optimization is performed. Image preprocessing of the growth state image is completed after edge optimization.

6. The method for evaluating seedling planting and growth based on image recognition parameters according to claim 5, characterized in that, Feature extraction is performed on the preprocessed growth state image, including: The planting area is located in the growth status image after image preprocessing. The planting area is located as follows: the growth status image after removing non-target areas is divided into individual plant boundaries by the row spacing and plant spacing features in the image. Then, the individual plant boundaries are marked with the main trunk as the core. After marking the individual plants, the located individual seedlings are obtained. The core features of individual seedlings marked in the high-definition camera are extracted, including trunk features, leaf features, and crown features. The trunk features include plant height, stem diameter, and trunk uprightness; the leaf features include leaf number, leaf area, and average leaf color; and the crown feature is the crown diameter. The spectral features of individual seedlings marked in the multispectral camera are extracted, including key band combination calculation, vegetation index quantification, and single-species spectral mean. Among them, the key band combination calculation includes chlorophyll-related features and water-related features; vegetation index quantification includes the calculation of normalized vegetation index and greenness index. Temperature characteristics of individual seedlings marked in infrared thermal imagers are extracted, including basic leaf temperature parameters and identification of abnormal temperature regions. The basic leaf temperature parameters include average temperature and temperature standard deviation; the identification of abnormal temperature regions includes extraction of high-temperature areas and extraction of temperature gradients. The extracted temperature features, spectral features and core features are integrated. The feature integration is to summarize the three extracted features to obtain the feature set of a single seedling. Then, the three features collected at the same time are bound according to the timestamp. After binding is completed, the characteristic data of the planted seedlings are obtained.

7. The method for evaluating seedling planting and growth based on image recognition parameters according to claim 6, characterized in that, The physiological characteristics of the seedlings are analyzed based on the extracted features, including: First, identify the physiological characteristics and their states, including growth state, nutritional state, water state, and health state. The extracted features were associated with the confirmed state types, and the mapping relationship between the features and state types was confirmed. Specifically, growth status was associated with trunk features and crown features; nutrient status was associated with leaf features and spectral features; water status was associated with spectral features; and health status was associated with leaf features, infrared temperature features, and spectral anomaly data. Based on the correlation results, a preliminary analysis was performed on each state type. The analysis of growth status was as follows: the growth rate was calculated using time-series data of plant height and stem diameter, and the balance was determined by the ratio of crown diameter to plant height. The analysis of nutrient status was as follows: the chlorophyll level was determined based on the numerical range of the normalized vegetation index. The analysis of water status was as follows: the leaf fullness was determined based on the reflectance threshold of water-related bands. The analysis of health status was as follows: the health level was determined by the average leaf temperature and leaf number obtained from infrared thermal imagers, whether they were lower than the average for the same growth stage, and whether there were any local abnormal colors. Cross-validation is performed on the preliminary analysis results, and feature error correction is performed based on the cross-validation results; The feature error correction results were combined with the physiological characteristics of different growth stages to obtain the analysis results for the seedling stage, the shoot emergence stage, and the dormancy stage. Finally, the physiological characteristics of the seedlings were analyzed.

8. The method for evaluating seedling planting and growth based on image recognition parameters according to claim 7, characterized in that, A growth prediction model for seedlings was established based on the analyzed physiological characteristics, and image parameters were correlated with actual growth indicators, including: Before establishing a growth prediction model, the prediction targets are first identified, including short-term prediction, medium-term prediction, and state prediction. Then, the parsed physiological characteristic data and the corresponding image parameters are used as input variables, and the actual growth indicators are used as output variables. Retrieve historical data from the database and divide the retrieved historical data into sample sets; A time series forecasting model is used for short-term and medium-term forecasting; a classification model is used for state forecasting, and then the adopted models are trained. The input and output variables are correlated and mapped, and the correlation mapping results are embedded into the trained model to obtain a preliminary prediction model. The accuracy of the initial prediction model is verified, and iterative optimization is performed based on the verification results; After iterative optimization, a complete growth prediction model was obtained.

9. The method for evaluating seedling planting and growth based on image recognition parameters according to claim 8, characterized in that, The established growth prediction model is correlated with environmental parameters, and the growth status of the seedlings is evaluated based on the correlation results, including: The correlation between the growth prediction model and environmental parameters is as follows: First, the environmental parameters are classified, including climate parameters, soil parameters and stress parameters. The correlation dimension is determined based on the environmental parameters. The environmental parameters are used as supplementary input variables of the model to correct the prediction results. Then, the suitability of the environmental parameters is inferred from the growth index output by the model. After the growth prediction model is associated with environmental parameters, the environmental parameters are time-aligned with the input and output data of the growth prediction model according to the timestamp of data collection. After time alignment, a three-dimensional dataset of environmental parameters, physiological characteristics and predicted growth indicators is obtained. The three-dimensional dataset is embedded into the growth prediction model, and the resulting logic for the linkage between seedling environment and growth is obtained. The prediction logic is evaluated based on the set evaluation indicators, which are growth quality indicators, environmental adaptability indicators, and stress risk indicators. Among them, the growth quality index is to compare the predicted growth index with the standard value under the same environment and calculate the compliance rate; the compliance rate is to analyze whether the environmental parameters are within the suitable range and to assess the matching degree between growth and environment; the stress risk index is to predict the probability of growth stagnation and leaf yellowing within the next 7 days based on the degree to which environmental parameters deviate from the threshold. The scope of the assessment is confirmed based on the assessment indicators, including individual plant assessment and regional assessment. Finally, the growth status of the seedlings within the assessment scope is classified into three levels based on the assessment indicators: high risk, medium risk, and low risk. After the assessment and grading are completed, the final seedling growth status assessment data is obtained.

10. The method for evaluating seedling planting and growth based on image recognition parameters according to claim 9, characterized in that, Based on the evaluation results, decision-making recommendations for seedling planting plans are generated, including: The obtained seedling growth status assessment data is broken down and the core data is identified. The core data includes the assessment level and problem type, and then the seedlings are prioritized according to the problem type and assessment level. A decision-making framework is constructed based on the growth cycle of the seedlings and the types of problems, including nutrient deficiency, water stress, and uneven growth. Based on the growth cycle and problem type in the decision-making recommendation framework, implementation measures are proposed, and the proposed implementation measures are integrated into a plan of action. After the measures and plans are integrated, we obtain the types of problems encountered in each growth cycle of the seedlings and the corresponding measures and plans based on the types of problems. The types of problems encountered in each growth cycle of the seedlings and the corresponding solutions are converted into reports and then transmitted to the display terminal for display.

Images