Multispectral monitoring and optimizing method and system for growth of tea trees in tea garden

Abstract

The invention relates to the technical field of image data processing, in particular to a multispectral monitoring and optimizing method and system for tea tree growth in a tea garden. The method comprises the following steps: acquiring tea tree growth spectrum data; performing tea tree growth spectral feature extraction according to the tea tree growth spectral data to obtain tea tree growth spectral feature data; performing tea tree growth index calculation according to the tea tree growth spectral feature data to obtain tea tree growth index data; and performing growth state monitoring according to the tea tree growth index data to obtain tea tree growth state monitoring data so as to perform tea tree growth and cultivation optimization operation. By continuously monitoring the growth state of the tea trees and performing dynamic optimization, the tea trees can be ensured to be always in the optimal growth environment, so that the yield and quality of the tea leaves are improved to the maximum extent.

Description

Technical Field

[0001] The present invention relates to the technical field of image data processing, and in particular to a multi-spectral monitoring and optimization method and system for tea tree growth in a tea garden. Background Art

[0002] With the development of agricultural science and technology, the use of information technology to monitor and manage the growth of crops has gradually become an important part of modern agriculture. Traditional crop growth monitoring methods mainly rely on manual observation and simple sensor data collection, which have great limitations in the breadth, depth and real-time monitoring. Especially in large-scale crop planting environments, how to accurately monitor the growth status of plants (such as tea trees), warn of abnormal conditions and carry out scientific cultivation management has become a major problem in agricultural production. Summary of the invention

[0003] In order to solve the above technical problems, the present invention proposes a multi-spectral monitoring and optimization method and system for tea tree growth in a tea garden to solve at least one of the above technical problems.

[0004] The present application provides a multi-spectral monitoring and optimization method for tea tree growth in a tea garden, the method comprising:

[0005] S1. Obtaining tea tree growth spectrum data;

[0006] S2, extracting tea tree growth spectrum characteristics according to the tea tree growth spectrum data to obtain tea tree growth spectrum characteristic data;

[0007] S3, calculating the tea tree growth index according to the tea tree growth spectrum characteristic data to obtain tea tree growth index data;

[0008] S4. Monitor the growth status of the tea tree according to the tea tree growth index data to obtain tea tree growth status monitoring data to optimize tea tree growth and cultivation operations.

[0009] The present invention is based on the multi-spectral data (including visible light and near-infrared bands) of tea trees, which can more accurately capture the growth status of tea trees, reflect the health status of tea trees, important physiological indicators such as water content and chlorophyll content, and make monitoring more accurate. The spectral characteristic data of tea tree growth can help to discover potential problems of tea trees in advance, such as leaf diseases, lack of water, nutritional imbalance, etc. Compared with traditional manual observation, it can issue an early warning when the problem just occurs to prevent the problem from expanding. It is possible to achieve fine-grained monitoring of tea trees at different growth stages and in different regions. For example, individualized monitoring and adjustment can be performed on a specific tea tree or tea trees in a specific area to ensure that the problem can be accurately located and handled in a timely manner.

[0010] Optionally, the tea tree growth spectrum data includes tea tree growth visible light data and tea tree growth near infrared band data, and S1 includes:

[0011] Visible light images of tea trees are collected by visible light collection equipment to obtain visible light data of tea tree growth;

[0012] The near-infrared band data of tea trees are collected by near-infrared sensors to obtain the near-infrared band data of tea tree growth.

[0013] The present invention uses visible light equipment and near-infrared sensors to cover the visible light and near-infrared bands of tea tree growth, and obtains detailed growth information of tea trees from multiple bands. Visible light data reflects the color, morphology, chlorophyll content, etc. of tea trees, while near-infrared data provides deep physiological information such as the water content and cell structure inside the leaves. The combination of visible light and near-infrared band data allows the system to monitor the external and internal growth conditions of tea trees at the same time, helping to identify early physiological abnormalities (such as water shortage and disease), and making optimized decisions based on this information.

[0014] Optionally, S2 includes:

[0015] Extracting color and texture features of visible light data of tea tree growth to obtain visible light feature data of tea tree growth;

[0016] Extract near-infrared band features from the near-infrared band data of tea tree growth to obtain near-infrared band feature data of tea tree growth;

[0017] The visible light characteristic data of tea tree growth and the near-infrared band characteristic data of tea tree growth are integrated to obtain the spectral characteristic data of tea tree growth.

[0018] The present invention extracts the color, texture features and spectral features of the tea tree from visible light data and near infrared data respectively, so that the system can understand the growth status of the tea tree from multiple dimensions. The color features and texture features reflect the appearance and surface characteristics of the tea tree, and the near infrared features reveal the internal physiological information of the tea tree. By extracting features from data of different bands, information can be obtained from both visual and spectral perspectives to help reveal the state of the tea tree at different growth stages. The combination of multimodal feature data helps to build a more comprehensive tea tree growth model, identify potential problems and conduct more accurate management. Through color and texture feature extraction, the system can monitor the external manifestations of the color change, leaf yellowing, disease texture, etc. of the tea tree. At the same time, the feature extraction of near infrared band data can deeply analyze the internal physiological indicators of the tea tree such as water content and nitrogen. The combination of the two can provide a more accurate tea tree health assessment, so that users can make more accurate judgments on the growth status of the tea tree. Color and texture feature extraction can help identify the appearance problems of the tea tree (such as disease spots, leaf dryness, etc.), and near infrared band feature extraction can detect water loss and nutrient deficiency inside the tea tree. By combining these two feature extraction methods, pests and diseases or water shortage problems can be detected earlier and more accurately, and timely responses can be made to prevent the problem from further expanding.

[0019] Optionally, the color texture feature extraction includes:

[0020] Perform regional segmentation on the visible light data of tea tree growth to obtain tea tree regional segmentation data;

[0021] Performing color space conversion on the tea tree region segmentation data to obtain tea tree color space conversion data;

[0022] Performing color histogram calculation on the tea tree color space conversion data to obtain tea tree color histogram data;

[0023] Performing cluster calculation according to the tea tree color histogram data to obtain tea tree color histogram cluster data;

[0024] Calculating the difference of the tea tree color histogram clustering data to obtain the tea tree color histogram clustering difference data;

[0025] When it is determined that the cluster difference data of the tea tree color histogram is greater than the preset difference threshold data, a color feature vector is calculated for the tea tree color histogram data to obtain tea tree color feature data;

[0026] When it is determined that the tea tree color histogram clustering difference data is less than or equal to the preset difference threshold data, the tea tree color histogram data is regionally aggregated according to the tea tree color histogram clustering difference data to obtain tea tree color histogram regional aggregation data;

[0027] Calculate the color feature vector of the tea tree color histogram area aggregation data to obtain the tea tree color feature data;

[0028] Extracting texture features from visible light data of tea tree growth to obtain tea tree texture feature data;

[0029] Color and texture feature encoding is performed according to the color feature data and the texture feature data of the tea tree to obtain the visible light feature data of the tea tree growth.

[0030] In the present invention, by performing regional segmentation on the visible light data of tea tree growth, different regions of the tea tree (such as leaves, branches, background, etc.) can be separated to ensure that color and texture feature extraction is only concentrated in the region of interest of the tea tree, reduce the interference of background noise, and improve the accuracy and precision of feature extraction. By calculating the difference of color histogram clustering, the method dynamically selects the path of feature extraction. When the cluster difference is greater than the threshold, the system performs more detailed color feature vector calculations for each cluster to capture the color details in a specific area. On the contrary, when the difference is small, the system performs regional aggregation on the color histogram data to reduce complexity, dynamically adjust the depth of feature extraction and balance the calculation cost, and ensure efficiency and accuracy. The path selection driven by the difference enables the method to automatically adapt to different growth states and environmental changes, intelligently adjust the analysis strategy, and ensure that the system can make the best feature extraction decision when facing different tea varieties and different growth conditions.

[0031] Optionally, the color texture feature encoding includes:

[0032] Perform spatial position encoding according to the tea tree color feature data and the tea tree texture feature data to obtain tea tree color feature spatial encoding data and tea tree texture feature spatial encoding data respectively;

[0033] Perform multi-head self-attention calculation on the tea tree color feature data and the tea tree texture feature data according to the tea tree color feature space encoding data and the tea tree texture feature space encoding data to obtain tea tree color texture multi-head self-attention feature data;

[0034] According to the multi-head self-attention feature data of tea tree color and texture, feature weighted fusion is performed on the tea tree color feature data and the tea tree texture feature data to obtain the visible light feature data of tea tree growth.

[0035] In the present invention, by spatially encoding the color feature data and texture feature data of tea trees, the system can capture the spatial context information of each pixel or feature position. Spatial position encoding enhances the expressive power of features, ensuring that the system not only focuses on single color or texture features, but also considers the distribution and arrangement of these features in space. The morphology and texture of tea leaves often have certain spatial dependencies, which helps to retain and strengthen these dependencies, thereby improving the accuracy of monitoring the growth status of tea trees. Through spatial position encoding, the system can handle the complex growth structure of tea trees, such as the difference between the edge and center of the leaf, the change of local texture, etc., so that the system can better analyze the relationship between different areas of the tea tree and capture the subtle changes in the growth of the tea tree. Through feature weighted fusion, the system can intelligently assign weights to color features and texture features. The growth status of tea trees depends on different features at different stages. For example, in some cases, color features may be more important indicators (such as changes in leaf color indicate malnutrition or disease), while in other cases, texture features may better reflect the growth health of tea trees (such as subtle structural changes on the surface of the leaves). Through weighted fusion, the system can dynamically adjust the weights of color and texture features in comprehensive judgment to ensure the optimal combination of features.

[0036] Optionally, the near-infrared band feature extraction includes:

[0037] According to the texture feature data of tea trees, the near-infrared band data of tea tree growth is divided into regions to obtain the near-infrared band division data of tea trees;

[0038] Extract spectral feature points according to the near-infrared band division data of tea trees to obtain spectral feature point data of tea trees;

[0039] Band ratio feature extraction is performed based on the visible light data of tea tree growth and the near-infrared band division data of tea tree to obtain band ratio feature data;

[0040] The spectral envelope removal analysis is performed on the tea tree near-infrared band division data to obtain the envelope removal feature data;

[0041] The spectral feature point data, band ratio feature data and envelope removal feature data of tea trees are integrated to obtain the near-infrared band feature data of tea tree growth.

[0042] In the present invention, by utilizing the texture features of tea trees to divide the near-infrared band data into regions, different regions of tea trees, such as healthy leaves, damaged leaves, dry areas, etc., can be accurately located. Ensure the focus of feature extraction, so that the analysis is concentrated on the key parts of the tea tree, avoid the interference of background noise, not only improve the accuracy of feature extraction, but also more accurately identify the problem areas in the growth of tea trees. Through fine-grained regional division, the system can perform more detailed spectral analysis for different regions. For example, healthy areas and areas where diseases may occur can be processed separately to ensure that the system can perform targeted analysis and processing according to the characteristics of different regions. The extraction of spectral feature points (such as absorption peaks, valleys, slopes, etc.) can capture the most critical changes in the near-infrared spectrum, and the system can accurately locate the health status of tea leaves, such as changes in chemical components such as water and nitrogen in the leaves. Spectral envelope removal can eliminate the interference of spectral background on absorption features, highlight the absorption peaks in the spectrum, and more clearly reveal the physiological changes in tea leaves, such as the dynamic changes of water and nutrients.

[0043] Optionally, the spectral envelope removal analysis comprises:

[0044] Multi-resolution smoothing is performed on the near-infrared band division data of tea trees to obtain near-infrared band division smoothing data;

[0045] Multi-scale spline fitting is performed on the smoothed data divided in the near-infrared band to obtain envelope fitting data;

[0046] According to the tea tree near-infrared band division data and envelope fitting data, dynamic envelope removal is performed to obtain near-infrared band envelope removal data;

[0047] Perform multi-layer feature decomposition on the near-infrared band envelope removal data to obtain near-infrared band multi-layer feature data;

[0048] The feature fusion of multi-layer feature data in the near-infrared band is performed to obtain the envelope-removed feature data.

[0049] In the present invention, before the spectral envelope is removed, multi-resolution smoothing is performed on the tea tree near-infrared band division data, which can effectively remove the noise in the data while retaining the key features of the spectral curve. The multi-scale spline fitting method can fit the envelope at different scales, so that the fitting result can not only tightly envelop the local peak of the spectral curve, but also will not overfit, thereby improving the accuracy and flexibility of the envelope fitting. Compared with the traditional envelope removal method, the dynamic envelope removal can adaptively adjust the spectral envelope of different regions according to the characteristics of the tea tree near-infrared band division data. For example, for areas with deeper absorption peaks, the envelope removal intensity is weakened to avoid excessive feature removal; and for smooth areas, the envelope removal intensity is enhanced to eliminate background effects. By performing multi-layer feature decomposition on the spectral data after the envelope is removed, the spectral signal can be decomposed into multiple frequency levels, each level representing different degrees of feature information in the spectral curve. For example, the high-frequency layer can reveal subtle changes in the spectrum, while the low-frequency layer can show the overall trend of the spectrum. The feature fusion step can effectively integrate information of different scales and levels. By combining the high-frequency and low-frequency information after multi-layer decomposition, the system can more comprehensively analyze the spectral characteristics of the tea tree and provide accurate data support for monitoring the health status of the tea tree.

[0050] Optionally, S3 includes:

[0051] Calculate the leaf growth index based on the visible light characteristic data of tea tree growth to obtain leaf growth index data;

[0052] The water content of tea trees is calculated based on the near-infrared band characteristic data of tea tree growth to obtain the water content data of tea trees;

[0053] The spectral index is calculated based on the leaf growth index data and the tea tree water content data to obtain the tea tree growth index data.

[0054] The present invention enables the system to capture the appearance and internal physiological characteristics of tea trees at the same time, providing a more comprehensive assessment of the growth status of tea trees. By understanding the color, texture and other information of the leaves through visible light feature data, combined with the moisture and nutrient conditions in near-infrared data, the true health status of the tea trees can be better reflected. Visible light data can provide the external growth status of the leaves (such as color and morphological changes), while near-infrared data can reveal internal moisture and other physiological characteristics. Through independent analysis and integration of these data, the generated tea tree growth index can more accurately reflect the overall health status of the tea trees.

[0055] Optionally, S4 includes:

[0056] Classify the growth status of tea trees according to the tea tree growth index data to obtain the growth status data of tea trees;

[0057] Performing time series monitoring of the growth state according to the tea tree growth state data to obtain the tea tree growth state time series data;

[0058] Growth status anomaly detection is performed based on the tea tree growth status time series data to obtain tea tree growth status monitoring data for optimizing tea tree growth and cultivation operations.

[0059] In the present invention, by classifying the tea tree growth index data, the system can divide the growth status of the tea tree into different health levels or categories, such as healthy, sub-healthy, water shortage, disease, etc. The time series monitoring of the growth status can reveal the long-term growth trend of the tea tree by tracking the changes in the growth status of the tea tree at different time points. Anomaly detection can help the system identify abnormal phenomena in the growth of tea trees, such as sudden leaf yellowing, abnormal water loss, disease outbreaks, etc. By analyzing the time series data of the growth status of the tea tree, the system can promptly detect abnormal patterns and issue early warnings to help managers take intervention measures before the problem develops into a major problem.

[0060] Optionally, a multi-spectral monitoring and optimization system for tea tree growth in a tea garden is used to execute the multi-spectral monitoring and optimization method for tea tree growth in a tea garden as described above, and the multi-spectral monitoring and optimization system for tea tree growth in a tea garden comprises:

[0061] Tea tree growth spectrum data acquisition module, used to obtain tea tree growth spectrum data;

[0062] A tea tree growth spectrum feature extraction module is used to extract tea tree growth spectrum features according to tea tree growth spectrum data to obtain tea tree growth spectrum feature data;

[0063] A tea tree growth index calculation module is used to calculate the tea tree growth index according to the tea tree growth spectrum characteristic data to obtain tea tree growth index data;

[0064] The tea tree growth status monitoring module is used to monitor the growth status according to the tea tree growth index data, and obtain the tea tree growth status monitoring data to optimize the tea tree growth and cultivation operations.

[0065] The objects of the present invention are:

[0066] 1. Make full use of the multispectral data of tea trees, including data fusion of visible light and near-infrared bands. By acquiring and processing multispectral data, the system can fully capture the external morphological characteristics of tea trees (such as leaf color and texture) and internal physiological information (such as water content, nitrogen, etc.). Using multispectral data for precise growth monitoring can improve the overall health of tea trees by 15%-20%. By calculating the leaf growth index and water content data of tea trees, the tea tree growth index is generated, which combines external growth performance (such as leaf color changes, leaf thickness) and internal physiological status (such as water, nitrogen, etc.), thereby providing reliable data support for time series monitoring.

[0067] 2. The timing monitoring mechanism of the present invention allows the system to continuously track the growth status of tea trees, and can capture the state changes of tea trees at different growth stages in real time, which can not only help managers understand the dynamic trend of tea tree growth, but also quickly detect potential problems when there are subtle changes in the state of tea trees. By using visible light data to monitor the appearance characteristics of tea leaves (such as color changes, yellowing, etc.), combined with near-infrared band data to analyze physiological characteristics such as internal moisture and nitrogen, the system can detect potential signals of physiological problems 2-3 weeks earlier than traditional manual detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] Other features, objects and advantages of the present application will become more apparent by reading the detailed description of non-limiting implementations made with reference to the following drawings:

[0069] Figure 1 A flowchart showing a method for multi-spectral monitoring and optimization of tea tree growth in a tea garden according to an embodiment of the present invention is shown;

[0070] Figure 2 A flowchart of a method for collecting tea tree growth spectrum data according to an embodiment is shown;

[0071] Figure 3 A flowchart of a method for extracting tea tree growth spectrum characteristics according to an embodiment is shown;

[0072] Figure 4 A flowchart showing a method for calculating a tea tree growth index according to an embodiment of the present invention is provided;

[0073] Figure 5 A flowchart showing a method for monitoring the growth status of a tea tree according to an embodiment of the present invention is provided;

[0074] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0075] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0076] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0077] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0078] See also Figures 1 to 5 The present application provides a multi-spectral monitoring and optimization method for tea tree growth in a tea garden, the method comprising:

[0079] S1. Obtaining tea tree growth spectrum data;

[0080] Specifically, multispectral imaging equipment is used to collect spectral data of tea tree growth from different angles, including visible light, near infrared, far infrared and other bands. Data collection methods include multispectral sensors carried by drones, or ground-based multispectral camera systems for periodic data collection. The collected spectral data is pre-processed by denoising, normalization and other pre-processing operations through an automated data processing system to improve data quality.

[0081] Spectral reflectance calculation is performed on the data of each band to obtain standardized spectral reflectance data. Principal component analysis (PCA) is performed on the spectral data to reduce the data dimension and extract key spectral components. Spectral segmentation algorithm is applied to the spectral data to distinguish the spectral change areas under different growth conditions.

[0082] S2, extracting tea tree growth spectrum characteristics according to the tea tree growth spectrum data to obtain tea tree growth spectrum characteristic data;

[0083] Specifically, based on the spectral reflectance data of different bands, key spectral features are extracted, including vegetation index (such as NDVI, SAVI), moisture index (such as NDWI), chlorophyll content index, etc. The spectral data are feature screened through machine learning algorithms (such as random forest or support vector machine) to ensure that the extracted features are strongly correlated with the growth status of the tea trees.

[0084] Calculate common vegetation indices, such as the Normalized Difference Vegetation Index. Use the Recursive Feature Elimination (RFE) algorithm to reduce the dimensionality of the extracted spectral features and retain the features that contribute most to the growth status. Use clustering algorithms (such as K-means or DBSCAN) to classify the spectral features of tea trees to form different growth characteristic groups.

[0085] S3, calculating the tea tree growth index according to the tea tree growth spectrum characteristic data to obtain tea tree growth index data;

[0086] Specifically, based on the extracted spectral feature data, a mathematical model is used to calculate the growth index of the tea tree, such as the biomass index, leaf area index, etc. A tea tree growth index model is established by performing regression analysis on historical spectral data and tea tree growth data.

[0087] The tea tree growth index is fitted using a multivariate linear regression model, with spectral feature data as input and tea tree growth index data as output. If a nonlinear relationship exists, nonlinear regression methods such as support vector regression (SVR) or neural networks can be used for fitting. According to seasonal changes and environmental factors, the calculation weight of the growth index is adjusted to improve the accuracy of the calculation.

[0088] S4. Monitor the growth status of the tea tree according to the tea tree growth index data to obtain tea tree growth status monitoring data to optimize tea tree growth and cultivation operations.

[0089] Specifically, the growth status of tea trees is monitored using tea tree growth index data combined with environmental parameters (such as temperature, precipitation, soil moisture, etc.). Based on the monitoring data, tea tree cultivation measures such as watering, fertilizing or pruning are dynamically adjusted to optimize the growth status of tea trees.

[0090] Through time series analysis, the changing trend of tea tree growth index is monitored and abnormal growth is warned. Combined with the Bayesian optimization algorithm, the cultivation strategy is dynamically adjusted according to the real-time monitored growth data to minimize resource usage and maximize growth effects. Using the control algorithm, the frequency and intensity of cultivation operations are automatically adjusted according to real-time changes in environmental conditions, such as adjusting the amount of water and fertilizer application.

[0091] Optionally, the tea tree growth spectrum data includes tea tree growth visible light data and tea tree growth near infrared band data, and S1 includes:

[0092] S11, collecting visible light images of tea trees by using visible light collection equipment to obtain visible light data of tea tree growth;

[0093] Specifically, high-resolution visible light camera equipment is used, installed on drones or fixed cameras, to collect visible light images of tea trees according to set routes or fixed points. To ensure the accuracy of visible light images, parameters such as exposure and white balance are adjusted during image acquisition to cope with changes in image quality under different lighting conditions. After image acquisition is completed, the visible light images are preliminarily processed, including image denoising, color correction, geometric correction, etc.

[0094] The collected visible light images are denoised by image preprocessing algorithms, such as median filtering or Gaussian filtering, to eliminate environmental interference. Images collected at different times or angles are aligned using image registration technology to ensure the temporal and spatial consistency of the data. The preprocessed images are segmented, and the tea tree and background areas are distinguished by the watershed algorithm or K-means clustering method to extract the effective image area of ​​the tea tree.

[0095] S12. Collect near-infrared band data of tea trees through near-infrared sensors to obtain near-infrared band data of tea tree growth.

[0096] Specifically, professional multispectral or near-infrared sensors are installed on drones or fixed structures to regularly collect near-infrared band data of tea trees. During the collection process, the gain and exposure time of the sensor are adjusted to ensure that the near-infrared data obtained under different lighting conditions have sufficient contrast and details. The collected near-infrared images are calibrated to compensate for the response characteristics of the equipment and the impact of ambient light on the near-infrared signal.

[0097] The radiation correction algorithm is used to eliminate the band response differences caused by sensor characteristics and external environmental factors to ensure the accuracy of near-infrared data. The normalization algorithm is used to standardize the data in the near-infrared band so that data from different devices or acquisition conditions can be uniformly compared.

[0098] Optionally, S2 includes:

[0099] S21, extracting color and texture features from the visible light data of tea tree growth to obtain visible light feature data of tea tree growth;

[0100] Specifically, the color features of tea leaves are extracted from visible light images, and color feature data are obtained using methods such as color histogram and color space conversion (such as RGB to HSV, Lab). Texture analysis algorithms such as gray-level co-occurrence matrix (GLCM) are used to extract the texture features of the image, and the texture information of the tea leaves is evaluated, including indicators such as contrast, entropy, and uniformity, to obtain texture feature data. Edge detection algorithms (such as Canny and Sobel) are used to extract the edge features of tea leaves, and the shape of the leaves is analyzed in combination with shape descriptors (such as Fourier descriptors) to obtain shape feature data.

[0101] Convert the RGB image to the HSV color space, and calculate the mean and standard deviation of hue, saturation, and brightness to generate color distribution data. Construct a color histogram and normalize each channel to obtain standardized color features to obtain color feature data.

[0102] Gray-level co-occurrence matrix was calculated and features such as contrast, entropy, correlation and energy were extracted to describe the surface roughness and texture uniformity of tea leaves. Local texture features were extracted using the local binary pattern (LBP) algorithm to generate a tea leaf texture distribution map and obtain texture feature data.

[0103] The boundary information of tea leaves is obtained through the edge detection algorithm, and the geometric features of the leaves, such as aspect ratio, curvature, boundary complexity, etc., are extracted by combining the shape descriptor to obtain shape feature data.

[0104] The color feature data, texture feature data and shape feature data are subjected to spatial multi-head self-attention calculation and feature weighted fusion to obtain the visible light feature data of tea tree growth.

[0105] S22, extracting near-infrared band features from the near-infrared band data of tea tree growth to obtain near-infrared band feature data of tea tree growth;

[0106] Specifically, using data from the near-infrared band, the spectral reflectance of tea leaves at different wavelengths is calculated to analyze the health status of the plants. Combined with visible light data, common vegetation indices such as the Normalized Difference Vegetation Index (NDVI) and the Enhanced Vegetation Index (EVI) are calculated to assess the growth and health of tea trees. Water-related indices (such as NDWI, Normalized Difference Water Index) are calculated to assess the moisture content of tea trees.

[0107] The radiation correction algorithm is used to correct sensor response and ambient light interference to ensure consistency of reflectance data at different wavelengths. The corrected near-infrared band data is interpolated to obtain a complete spectral curve and extract reflectance characteristics of key wavelengths.

[0108] S23, integrating the visible light characteristic data of tea tree growth and the near-infrared band characteristic data of tea tree growth to obtain the spectral characteristic data of tea tree growth.

[0109] Specifically, data from different sources are normalized to ensure that the feature values ​​are within the same numerical range to avoid a certain feature having too much influence on the results. The feature data are processed using the Z-score normalization method or the minimum-maximum normalization method.

[0110] Apply multimodal feature fusion algorithm to weighted average the visible light and near infrared features according to different weights to form comprehensive features. Use principal component analysis (PCA) to extract the main components in the fused features to reduce data redundancy. Use PCA algorithm to select the first few principal components with the largest explained variance in the features to reduce feature dimensions. Nonlinear dimensionality reduction methods such as t-SNE can be used to further analyze the internal structure of high-dimensional data to ensure that important patterns and feature relationships are retained.

[0111] Optionally, the color texture feature extraction includes:

[0112] Perform regional segmentation on the visible light data of tea tree growth to obtain tea tree regional segmentation data;

[0113] Specifically,

[0114] Performing color space conversion on the tea tree region segmentation data to obtain tea tree color space conversion data;

[0115] Specifically,

[0116] Performing color histogram calculation on the tea tree color space conversion data to obtain tea tree color histogram data;

[0117] Specifically,

[0118] Performing cluster calculation according to the tea tree color histogram data to obtain tea tree color histogram cluster data;

[0119] Specifically,

[0120] Calculating the difference of the tea tree color histogram clustering data to obtain the tea tree color histogram clustering difference data;

[0121] Specifically,

[0122] When it is determined that the cluster difference data of the tea tree color histogram is greater than the preset difference threshold data, a color feature vector is calculated for the tea tree color histogram data to obtain tea tree color feature data;

[0123] Specifically,

[0124] When it is determined that the tea tree color histogram clustering difference data is less than or equal to the preset difference threshold data, the tea tree color histogram data is regionally aggregated according to the tea tree color histogram clustering difference data to obtain tea tree color histogram regional aggregation data;

[0125] Specifically,

[0126] Calculate the color feature vector of the tea tree color histogram area aggregation data to obtain the tea tree color feature data;

[0127] Specifically,

[0128] Extracting texture features from visible light data of tea tree growth to obtain tea tree texture feature data;

[0129] Specifically,

[0130] Color and texture feature encoding is performed according to the color feature data and the texture feature data of the tea tree to obtain the visible light feature data of the tea tree growth.

[0131] Specifically,

[0132] Optionally, the color texture feature encoding includes:

[0133] Perform spatial position encoding according to the tea tree color feature data and the tea tree texture feature data to obtain tea tree color feature spatial encoding data and tea tree texture feature spatial encoding data respectively;

[0134] Specifically, the tea tree color feature data and tea tree texture feature data are positionally encoded, and the position information of each pixel is embedded into the feature data so that the model can capture the spatial position information. The spatial position in the image is converted into a high-dimensional vector using sine and cosine position embedding functions and added to the color and texture features.

[0135] Perform multi-head self-attention calculation on the tea tree color feature data and the tea tree texture feature data according to the tea tree color feature space encoding data and the tea tree texture feature space encoding data to obtain tea tree color texture multi-head self-attention feature data;

[0136] Specifically, based on the spatially encoded data of color and texture, multi-head self-attention calculation is performed to learn the complex relationship between the color and texture features of the tea tree through multiple attention heads. Each attention head learns local features independently, and then the results of each head are combined to obtain stronger expression capabilities. Calculate the query matrix (Q), key matrix (K) and value matrix (V); use multiple independent attention heads to calculate the self-attention of the color and texture feature spatial encoding data respectively, and then splice the results of each head; weighted fusion is performed on the output results of each attention head to obtain the color and texture multi-head self-attention feature data.

[0137] According to the multi-head self-attention feature data of tea tree color and texture, feature weighted fusion is performed on the tea tree color feature data and the tea tree texture feature data to obtain the visible light feature data of tea tree growth.

[0138] Specifically, a weight coefficient is assigned to each color and texture feature, and weighted according to the weighting matrix generated by the multi-head self-attention feature. The color and texture features of the tea tree are weighted and fused, combined with the multi-head self-attention feature to enhance the overall expression ability. The weight coefficient is automatically generated through the learning model to ensure that the contribution of color and texture features is adaptable at different growth stages.

[0139] Optionally, the near-infrared band feature extraction includes:

[0140] According to the texture feature data of tea trees, the near-infrared band data of tea tree growth is divided into regions to obtain the near-infrared band division data of tea trees;

[0141] Specifically, the texture feature data of tea trees (such as the texture distribution of leaves) is used to guide the regional division of near-infrared band images. K-means, DBSCAN or texture-based segmentation algorithms (such as SLIC superpixel segmentation) are used to perform spatial regional division of near-infrared band data and extract different growth areas.

[0142] By calculating the local variance of the tea tree texture feature data, the texture salient areas are identified and used as a reference for segmentation of the near-infrared band data. The near-infrared band data points are assigned to different cluster centers (representing different areas) to generate near-infrared band partition data.

[0143] Extract spectral feature points according to the near-infrared band division data of tea trees to obtain spectral feature point data of tea trees;

[0144] Specifically, in each divided area, spectral feature points of key wavelengths are identified using spectral analysis technology. The feature points of a specific band, such as the wavelength points with the highest and lowest reflectivity, are selected through the peak and trough detection algorithm. The spectral curve of the divided area is smoothed, such as using a Savitzky-Golay filter.

[0145] Detect peaks and troughs by zeros of the first derivative or extreme values ​​of the second derivative. Smooth the spectral curve of the divided area, such as using a Savitzky-Golay filter. Detect peaks and troughs by zeros of the first derivative or extreme values ​​of the second derivative.

[0146] Band ratio feature extraction is performed based on the visible light data of tea tree growth and the near-infrared band division data of tea tree to obtain band ratio feature data;

[0147] Specifically, the ratio characteristics of different bands are calculated by combining visible light data (such as red light band) and near infrared data to analyze the health of vegetation growth. Vegetation index extraction, such as the Normalized Difference Vegetation Index (NDVI) and the Enhanced Vegetation Index (EVI), is extracted through band ratios.

[0148] The spectral envelope removal analysis is performed on the tea tree near-infrared band division data to obtain the envelope removal feature data;

[0149] Specifically, the spectral continuum removal technique is used to remove background interference in the spectral reflectance curve to highlight the specific spectral characteristics of the tea tree. After the continuum removal, the spectral characteristics, such as the depth and width of the absorption band, are enhanced for further analysis.

[0150] By constructing an envelope, the local maximum points on the spectrum curve are connected to form an envelope curve. The spectrum curve is divided by the envelope to obtain the characteristic curve after removing the envelope. On the spectrum curve after removing the envelope, the depth, position, width and other characteristics of the absorption band are calculated as the spectrum envelope removal features.

[0151] The spectral feature point data, band ratio feature data and envelope removal feature data of tea trees are integrated to obtain the near-infrared band feature data of tea tree growth.

[0152] Specifically, the spectral feature points, band ratio features and envelope removal features are integrated (or vectorized) to form a comprehensive feature vector containing multiple spectral information.

[0153] Optionally, the spectral envelope removal analysis comprises:

[0154] Multi-resolution smoothing is performed on the tea tree near-infrared band division data to obtain near-infrared band division smoothing data;

[0155] Specifically, a multi-resolution filter is used to smooth the near-infrared band data to eliminate noise and retain the main spectral features. Wavelet transform or Gaussian smoothing filter is used to smooth the spectral data at different scales to improve the stability of the spectral curve.

[0156] Through wavelet decomposition, the spectral data is decomposed into details and approximations of different scales. The details are soft-thresholded or hard-thresholded to remove noise and retain the main features. Through wavelet decomposition, the spectral data is decomposed into details and approximations of different scales. The details are soft-thresholded or hard-thresholded to remove noise and retain the main features.

[0157] Multi-scale spline fitting is performed on the smoothed data divided in the near-infrared band to obtain envelope fitting data;

[0158] Specifically, the multi-scale spline fitting technology is used to fit the smooth data of the near-infrared band to obtain the envelope. The spectral envelope is established through multi-scale spline interpolation, which can accurately capture the local extreme values ​​of the spectral curve. The spline curve is interpolated between the local extreme points of the spectral data to form an envelope. By adjusting the scale parameters of the spline, the envelope fitting of spectral features of different scales is achieved, and the multi-scale envelope fitting data is obtained.

[0159] According to the tea tree near-infrared band division data and envelope fitting data, dynamic envelope removal is performed to obtain near-infrared band envelope removal data;

[0160] Specifically, according to the near-infrared band division data of the tea tree and the fitted envelope data, the envelope is dynamically removed to obtain the processed spectral data. The spectral curve is divided by the corresponding envelope to eliminate the baseline drift of the spectral curve and retain the absorption band characteristics. The fitting parameters of the envelope are dynamically adjusted, and the spectral curve is de-enveloped according to different spectral regions and envelope fitting errors. According to the characteristics of the spectral region, the fitting strength of the envelope is dynamically adjusted to avoid overfitting or underfitting. By comparing the changes in the spectral curve before and after removal, the effect of removal is evaluated, and the parameter settings for de-envelope are further optimized.

[0161] Perform multi-layer feature decomposition on the near-infrared band envelope removal data to obtain near-infrared band multi-layer feature data;

[0162] Specifically, multi-layer feature decomposition is performed on the near-infrared band data after envelope removal to extract various spectral features, such as absorption band intensity, width and position. Using methods such as empirical mode decomposition (EMD) and wavelet transform, the spectral signal is decomposed into different components to extract key features.

[0163] The spectral data after envelope removal is decomposed into a series of intrinsic mode functions (IMFs), each of which contains different scale features of the spectrum. Features such as local extreme values ​​and band energy are extracted from each IMF to characterize different levels of features in the near-infrared band. Wavelet decomposition is applied to the data after envelope removal to extract spectral features of different frequency components and generate multi-layer feature data.

[0164] The feature fusion of multi-layer feature data in the near-infrared band is performed to obtain the envelope-removed feature data.

[0165] Specifically, different weights are assigned according to the importance of each layer of features, and the multi-layer features are weighted summed to generate an envelope to remove feature data.

[0166] Optionally, S3 includes:

[0167] S31, calculating the leaf growth index of the tea tree growth visible light characteristic data to obtain leaf growth index data;

[0168] Specifically, the color, texture, shape and other information of tea leaves are extracted from the visible light feature data to calculate the leaf growth index. Based on the color, texture and other characteristics of tea trees, a biomass index (such as leaf area index LAI) is used as a measure of tea leaf growth.

[0169] The color and shape characteristics of tea leaves are used to estimate the leaf coverage area; the edge of each leaf is detected to estimate the area of ​​each leaf and calculate the total area. Combined with leaf texture characteristics (such as surface roughness), it supplements the assessment information of leaf growth status, especially for assessing health and growth quality.

[0170] S32, calculating the water content of the tea trees based on the near-infrared band characteristic data of tea tree growth to obtain the water content data of the tea trees;

[0171] Specifically, water absorption bands (such as the features at 950nm and 1450nm) are extracted from the spectral feature data of the near-infrared band to calculate the water content of the tea tree. The water content of the tea tree is calculated by the water index (such as the normalized water index NDWI) and evaluated in combination with the band reflectance data. According to the changes in the water index, the water content of the tea tree is estimated in combination with the empirical model, and the water content of the tea tree is inferred using the band ratio and the intensity of the characteristic absorption band.

[0172] S33, performing spectral index calculation according to the leaf growth index data and the tea tree water content data to obtain tea tree growth index data.

[0173] Specifically, the leaf growth index (such as LAI) and the tea tree water content index (such as NDWI) are integrated to form a tea tree growth index, which is used to describe the overall growth status of the tea tree. The leaf growth index and water content data are integrated through a multiple regression model or a machine learning algorithm to generate a tea tree growth index.

[0174] Optionally, S4 includes:

[0175] S41, classifying the growth status of the tea trees according to the tea tree growth index data to obtain tea tree growth status data;

[0176] Specifically, according to the tea tree growth index data, the growth status of the tea tree is divided into different categories (such as normal growth, accelerated growth, slow growth or abnormal growth). A supervised learning classification algorithm (such as support vector machine SVM, random forest or KNN) is applied to classify the growth status of the tea tree based on the historical growth index data.

[0177] Extract key features (such as leaf growth index, water content index, etc.) from the tea tree growth index data and train them using a classification algorithm. Make predictions on the trained model to classify the growth status of the tea tree. Use cross-validation or holdout methods to evaluate the accuracy of the classification model to ensure the reliability of the growth status classification results.

[0178] S42, performing time series monitoring of the growth state according to the tea tree growth state data to obtain the tea tree growth state time series data;

[0179] Specifically, based on the growth status data of tea trees, a time series model is constructed to conduct continuous status monitoring and record the changing trend of the growth status. Time series prediction algorithms (such as ARIMA, LSTM, etc.) are applied to analyze the changing rules of the growth status of tea trees in the time dimension.

[0180] Based on the growth status data, a time series model is constructed to capture the state changes of tea trees at different time points. Use the ARIMA (autoregressive moving average) model or the long short-term memory network (LSTM) for time series forecasting. Perform trend analysis on the growth status of tea trees to identify whether there is acceleration, slowdown or abnormal fluctuation. Smooth the time series data through moving average or exponential smoothing methods to reduce the impact of short-term fluctuations on monitoring results.

[0181] S43. Perform growth status abnormality detection based on the tea tree growth status time series data to obtain tea tree growth status monitoring data to optimize tea tree growth and cultivation operations.

[0182] Specifically, based on time series data, we identify anomalies in the growth status of tea trees, such as growing too fast or too slow, or drastic changes in growth index. We use statistical methods or machine learning algorithms (such as Isolation Forest, LOF, PCA) to detect anomalies and discover potential problems that may occur during the growth of tea trees.

[0183] Use threshold-based anomaly detection methods to monitor the growth status of tea trees and detect whether they deviate from the normal range. Alternatively, use Isolation Forest or Local Anomaly Factor (LOF) to model the time series data of tea tree growth status and find anomalies that do not conform to the expected pattern. After the anomaly is detected, feedback is provided through visualization and alarm mechanisms to optimize the growth status.

[0184] Optionally, a multi-spectral monitoring and optimization system for tea tree growth in a tea garden is used to execute the multi-spectral monitoring and optimization method for tea tree growth in a tea garden as described above, and the multi-spectral monitoring and optimization system for tea tree growth in a tea garden comprises:

[0185] Tea tree growth spectrum data acquisition module, used to obtain tea tree growth spectrum data;

[0186] A tea tree growth spectrum feature extraction module is used to extract tea tree growth spectrum features according to tea tree growth spectrum data to obtain tea tree growth spectrum feature data;

[0187] A tea tree growth index calculation module is used to calculate the tea tree growth index according to the tea tree growth spectrum characteristic data to obtain tea tree growth index data;

[0188] The tea tree growth status monitoring module is used to monitor the growth status according to the tea tree growth index data, and obtain the tea tree growth status monitoring data to optimize the tea tree growth and cultivation operations.

[0189] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A multi-spectral monitoring and optimization method for tea tree growth in a tea garden, characterized in that: The method comprises: S1. Obtaining tea tree growth spectrum data; S2, extracting tea tree growth spectrum characteristics according to the tea tree growth spectrum data to obtain tea tree growth spectrum characteristic data; S3, calculating the tea tree growth index according to the tea tree growth spectrum characteristic data to obtain tea tree growth index data; S4. Monitor the growth status of the tea tree according to the tea tree growth index data to obtain tea tree growth status monitoring data to optimize tea tree growth and cultivation operations.

2. The method according to claim 1, characterized in that: The tea tree growth spectrum data includes the tea tree growth visible light data and the tea tree growth near infrared band data. S1 includes: Visible light images of tea trees are collected by visible light collection equipment to obtain visible light data of tea tree growth; The near-infrared band data of tea trees are collected by near-infrared sensors to obtain the near-infrared band data of tea tree growth.

3. The method according to claim 2, characterized in that S2 include: Extracting color and texture features of visible light data of tea tree growth to obtain visible light feature data of tea tree growth; Extract near-infrared band features from the near-infrared band data of tea tree growth to obtain near-infrared band feature data of tea tree growth; The visible light characteristic data of tea tree growth and the near-infrared band characteristic data of tea tree growth are integrated to obtain the spectral characteristic data of tea tree growth.

4. The method according to claim 3, characterized in that The color texture feature extraction includes: Perform regional segmentation on the visible light data of tea tree growth to obtain tea tree regional segmentation data; Performing color space conversion on the tea tree region segmentation data to obtain tea tree color space conversion data; Performing color histogram calculation on the tea tree color space conversion data to obtain tea tree color histogram data; Performing cluster calculation according to the tea tree color histogram data to obtain tea tree color histogram cluster data; Calculating the difference of the tea tree color histogram clustering data to obtain the tea tree color histogram clustering difference data; When it is determined that the cluster difference data of the tea tree color histogram is greater than the preset difference threshold data, a color feature vector is calculated for the tea tree color histogram data to obtain tea tree color feature data; When it is determined that the tea tree color histogram clustering difference data is less than or equal to the preset difference threshold data, the tea tree color histogram data is regionally aggregated according to the tea tree color histogram clustering difference data to obtain tea tree color histogram regional aggregation data; Calculate the color feature vector of the tea tree color histogram area aggregation data to obtain the tea tree color feature data; Extracting texture features from visible light data of tea tree growth to obtain tea tree texture feature data; Color and texture feature encoding is performed according to the color feature data and the texture feature data of the tea tree to obtain the visible light feature data of the tea tree growth.

5. The method according to claim 4, characterized in that The color texture feature encoding includes: Perform spatial position encoding according to the tea tree color feature data and the tea tree texture feature data to obtain tea tree color feature spatial encoding data and tea tree texture feature spatial encoding data respectively; Perform multi-head self-attention calculation on the tea tree color feature data and the tea tree texture feature data according to the tea tree color feature space encoding data and the tea tree texture feature space encoding data to obtain tea tree color texture multi-head self-attention feature data; According to the multi-head self-attention feature data of tea tree color and texture, feature weighted fusion is performed on the tea tree color feature data and the tea tree texture feature data to obtain the visible light feature data of tea tree growth.

6. The method according to claim 4, characterized in that The near-infrared band feature extraction includes: According to the texture feature data of tea trees, the near-infrared band data of tea tree growth is divided into regions to obtain the near-infrared band division data of tea trees; Extract spectral feature points according to the near-infrared band division data of tea trees to obtain spectral feature point data of tea trees; Band ratio feature extraction is performed based on the visible light data of tea tree growth and the near-infrared band division data of tea tree to obtain band ratio feature data; The spectral envelope removal analysis is performed on the tea tree near-infrared band division data to obtain the envelope removal feature data; The spectral feature point data, band ratio feature data and envelope removal feature data of tea trees are integrated to obtain the near-infrared band feature data of tea tree growth.

7. The method according to claim 6, characterized in that The spectrum envelope removal analysis includes: Multi-resolution smoothing is performed on the tea tree near-infrared band division data to obtain near-infrared band division smoothing data; Multi-scale spline fitting is performed on the smoothed data divided in the near-infrared band to obtain envelope fitting data; According to the tea tree near-infrared band division data and envelope fitting data, dynamic envelope removal is performed to obtain near-infrared band envelope removal data; Perform multi-layer feature decomposition on the near-infrared band envelope removal data to obtain near-infrared band multi-layer feature data; The feature fusion of multi-layer feature data in the near-infrared band is performed to obtain the envelope-removed feature data.

8. The method according to claim 3, characterized in that S3 include: Calculate the leaf growth index based on the visible light characteristic data of tea tree growth to obtain leaf growth index data; The water content of tea trees is calculated based on the near-infrared band characteristic data of tea tree growth to obtain the water content data of tea trees; The spectral index is calculated based on the leaf growth index data and the tea tree water content data to obtain the tea tree growth index data.

9. The method according to claim 1, characterized in that S4 include: Classify the growth status of tea trees according to the tea tree growth index data to obtain the growth status data of tea trees; Performing time series monitoring of the growth state according to the tea tree growth state data to obtain the tea tree growth state time series data; Growth status anomaly detection is performed based on the tea tree growth status time series data to obtain tea tree growth status monitoring data for optimizing tea tree growth and cultivation operations.

10. A multi-spectral monitoring and optimization system for tea tree growth in a tea garden, characterized in that: Used to implement the multi-spectral monitoring and optimization method for tea tree growth in a tea garden as claimed in claim 1, the multi-spectral monitoring and optimization system for tea tree growth in a tea garden comprises: Tea tree growth spectrum data acquisition module, used to obtain tea tree growth spectrum data; A tea tree growth spectrum feature extraction module is used to extract tea tree growth spectrum features according to tea tree growth spectrum data to obtain tea tree growth spectrum feature data; A tea tree growth index calculation module is used to calculate the tea tree growth index according to the tea tree growth spectrum characteristic data to obtain tea tree growth index data; The tea tree growth status monitoring module is used to monitor the growth status according to the tea tree growth index data, and obtain the tea tree growth status monitoring data to optimize the tea tree growth and cultivation operations.

Images