Citrus huanglongbing tree health assessment method based on image processing and algorithm

Abstract

The invention provides a citrus tree health assessment method based on image processing and an algorithm, and aims to judge citrus tree Huanglongbing infection conditions. According to the method, the growth state of the citrus tree is captured by utilizing a high-resolution image, and plant diseases are identified and analyzed by adopting an advanced image processing technology. By establishing the health assessment model, the citrus trees are monitored and assessed in time, so that a scientific decision basis is provided for farmers. Compared with a traditional manual evaluation method, the method can adapt to citrus trees of different types and growth environments. Meanwhile, through long-term accumulation and analysis of data, valuable scientific research data support is provided for the citrus industry, and development of ecological agriculture is promoted. The implementation of the method provides an effective technical means for health management of citrus trees, promotes the scientific and intelligent process of the citrus industry, and assists sustainable agricultural development.

Description

Technical Field

[0001] The invention relates to the field of agricultural disease detection, and in particular to a method for evaluating the health of citrus Huanglongbing trees based on image processing and algorithms. Background Art

[0002] Huanglongbing is a highly contagious and devastating citrus plant disease that is primarily spread by the Asian citrus psyllid. The disease can cause yellowing of citrus leaves, deformed fruits, reduced yields, and ultimately tree death. Due to its complex symptoms and slow progression, the disease is usually already in a serious stage when obvious lesions appear on the appearance of the fruit tree.

[0003] In the prior art, the infection of Huanglongbing is judged by manually observing the external symptoms of citrus trees (such as yellowing leaves, sparse crowns, etc.). This method is time-consuming and laborious, and the evaluation results are easily disturbed by external factors (such as light, seasonal leaf changes, etc.), and the accuracy is low. In addition, the current orchard management relies on nutritional management guidelines, plant growth regulators, and trunk injections, which lack quantitative basis and make it difficult to judge whether the management measures are effective. Therefore, there is a need for an intelligent and precise technical means to help growers accurately judge the extent to which trees are affected by Huanglongbing and their health status through the combination of image processing and data analysis. Summary of the invention

[0004] The present invention proposes a citrus Huanglongbing tree health assessment method based on the combination of image processing technology and algorithm. By regularly photographing the tree crowns, using image processing software to analyze the changes in the crown density, and combining algorithms to judge the tree health status, it also has a certain effect on judging whether citrus trees are infected with Huanglongbing in the early stage. This provides a scientific management basis to help orchard managers determine whether to continue to retain or remove diseased citrus trees.

[0005] The purpose of the present invention is achieved by the following technical solutions:

[0006] 1. Data Collection

[0007] 1.1 Tree selection and marking

[0008] Select several representative citrus trees in the orchard for testing. These trees should be representative. For each tree, take a crown image in four quadrants (east, south, west, and north) at the middle position between the trunk and the crown edge.

[0009] 1.2 Shooting Standards

[0010] Image acquisition using a smartphone or other device equipped with a high-quality camera (all conditions being equal) requires high resolution and good imaging capabilities to obtain clear and realistic image data. The phone is mounted on a selfie stick with the camera facing above the tree canopy, shooting in the area between the trunk and the edge of the canopy. Before each shot, ensure that the camera lens is clean and the picture is clear to avoid blurry images or excessive light that affects data analysis. Each tree should be photographed regularly (for example, once a month or three months) to ensure dynamic monitoring of health status and establish a long-term data set for analysis; record the time of each shot, weather conditions, soil moisture and other environmental factors so that these influencing factors can be considered in subsequent analysis. Before shooting, ensure that the settings of the equipment used are consistent, including exposure time, focal length, ISO value, etc., to ensure that images taken at different times and conditions are comparable. Use a standardized shooting process, including choosing appropriate shooting angles, lighting conditions and shooting heights to reduce the impact of external factors on image quality; calibrate the shooting equipment before each shot to ensure consistency of shooting results.

[0011] 2. Image Processing and Analysis

[0012] 2.1 Image Preprocessing

[0013] Image preprocessing is an essential step before image analysis. Its purpose is to optimize the original image into a format suitable for algorithm processing to improve the accuracy and reliability of subsequent analysis. Specific steps include denoising, color standardization, contrast enhancement, image cropping, etc.

[0014] 2.1.1 Image Denoising

[0015] When shooting an image, changes in ambient light, dust, and noise from the camera itself will affect the image quality, so it is necessary to remove these noise interferences first. Common image denoising methods include Gaussian filtering, median filtering, and bilateral filtering.

[0016] Gaussian filtering: By applying a Gaussian function filter to the image, small noise points in the image are smoothed while keeping the larger image structure intact.

[0017] Median filtering: For each pixel in the image, take the median of the pixel values ​​in its neighborhood, which is suitable for removing salt and pepper noise.

[0018] Bilateral filtering: not only considers the spatial proximity of pixels, but also the color similarity of pixels, thereby retaining edge information while removing noise. Common parameter settings for Gaussian filtering:

[0019]

[0020] Here, σ is the standard deviation of the filter, which is usually set between 1 and 2 to ensure that it can remove noise without over-smoothing details.

[0021] 2.1.2 Color Standardization

[0022] Due to different lighting conditions at different time periods, the color of tree leaves may show large deviations in the images. Therefore, it is necessary to color standardize the images so that the images taken at different times and under different lighting conditions have relatively consistent color performance to ensure the comparability of subsequent analysis.

[0023] White balance adjustment: Detects the white reference point in the image and adjusts the color temperature of the image to make it close to the real leaf color.

[0024] Histogram equalization: stretches the color distribution of the image to enhance the contrast and maintain color consistency under various lighting conditions.

[0025] 2.1.3 Contrast Enhancement

[0026] The purpose of contrast enhancement is to highlight the difference between the tree crown and the background for subsequent image segmentation and edge detection. A commonly used contrast enhancement method is histogram stretching, which stretches the brightness range of the image to the maximum dynamic range to make the difference between the dark and bright areas more obvious. The mathematical expression of contrast enhancement is:

[0027]

[0028] Among them, I(x,y) is the pixel value of the original image, I min and I max are the minimum and maximum brightness values ​​in the image, respectively.

[0029] 2.1.4 Image Cropping and Adjustment

[0030] According to the shooting standards, the image is cropped to retain only the tree crown part, ensuring that irrelevant background parts are not brought into the analysis.

[0031] 2.2 Calculation of crown density

[0032] Crown density is one of the important parameters that reflect the health of trees. It quantifies the density of the crown by calculating the pixel ratio of leaves per unit area. The calculation of crown density goes through the following steps:

[0033] 2.2.1 Image Segmentation

[0034] The purpose of image segmentation is to separate the crown area from the background and extract the pixel area occupied by the leaves. Commonly used image segmentation methods include edge detection, region growing method, threshold segmentation method and deep learning-based segmentation models (such as U-Net).

[0035] Edge detection (Canny algorithm): This algorithm finds the edge in the image by detecting the gradient change of the grayscale value of pixels in the image. It is suitable for images with obvious contrast between the crown and the background. The parameter settings of Canny edge detection include:

[0036] Low Threshold and High Threshold: These two thresholds are used to control the sensitivity of edge detection and are usually adjusted according to the brightness and contrast of the image. Typical values ​​are 30 and 100.

[0037] The standard deviation of the Gaussian filter is usually between 1 and 2 to smooth out tiny noise in the image while preserving edge information.

[0038] Region growing method: This algorithm starts from a certain pixel in the image and gradually expands the region around it until no pixels similar to the seed point can be found. It is suitable for situations where the crown area is relatively coherent.

[0039] 2.2.2 Binarization

[0040] After the segmentation is completed, the crown area is separated from the background, the background pixels are set to 0, and the crown pixels are set to 1. The key to binarization is to choose a suitable threshold. The global threshold can be set according to the brightness histogram of the image, or an adaptive threshold method can be used to enable different parts of the image to be adaptively adjusted according to the local contrast.

[0041] 2.2.3 Calculation formula for crown density

[0042] The calculation of tree crown density is based on the number of pixels in the tree crown area per unit area. The specific formula is as follows:

[0043]

[0044] Where: N tree is the number of canopy pixels; N total The total number of pixels in the image; D is the crown density, in percentage.

[0045] 2.3 Health score generation

[0046] The health score is a numerical indicator generated by combining multiple image features (including crown density, leaf color, shape, edges, etc.) to quantify the overall health of the tree. The score range is usually 0 to 100, where 0 means the tree is completely withered and 100 means the tree is in optimal health. The generation of the health score involves the following steps:

[0047] 2.3.1 Leaf color analysis

[0048] Leaf color is one of the important characteristics for judging the early symptoms of Huanglongbing. Healthy citrus tree leaves are dark green, while leaves infected with Huanglongbing may have yellow spots or turn yellow overall. By analyzing the changes in leaf color, it can be used to make a preliminary early judgment on whether citrus trees have symptoms of Huanglongbing.

[0049] Color analysis is based on color space conversion and color histogram statistics:

[0050] Color space conversion: Convert the image from RGB color space to HSV (hue, saturation, brightness) space. HSV space is more in line with the visual perception characteristics of the human eye, where hue can be used to distinguish the color changes of leaves, and saturation can be used to judge the brightness of the color.

[0051] Tone analysis: Extract the hue distribution histogram of the leaf area and count the proportion of hues in the yellow-green range (usually between 60° and 120°). If the hue tends to be yellow and the proportion increases, it may indicate that the leaf is affected by Huanglongbing.

[0052] 2.3.2 Leaf shape and curl detection

[0053] Huanglongbing may also cause changes in leaf shape, such as curling and edge deformation. Shape analysis can be used to detect whether leaves have abnormal curling. Shape analysis can use Hough transform to detect leaf edges, or calculate leaf shape parameters (such as edge curvature, aspect ratio, etc.) through morphological processing (such as erosion and expansion operations).

[0054] 2.3.3 Health score calculation formula

[0055] The calculation of the health score takes into account multiple image features and generates the final health score through a weighted formula. The typical formula is as follows:

[0056] S=w1D+w2C+w3L+w4E

[0057] Among them: S is the health score; D is the crown density score (accounting for 30% of the total score, that is, w1=0.3); C is the leaf color score (accounting for 40% of the total score, that is, w2=0.4); L is the leaf shape score (accounting for 20% of the total score, that is, w3=0.2); E is the shape abnormality score such as edge curvature (accounting for 10% of the total score, that is, w4=0.1).

[0058] 3. Interpretation of results

[0059] 3.1 Data preprocessing and baseline establishment

[0060] 3.1.1 Data Preprocessing

[0061] Before interpreting the results, you first need to ensure that the captured image data has been adequately preprocessed. Preprocessing steps usually include image denoising, color standardization, and contrast adjustment. Through these preprocessing steps, all images are ensured to have consistent quality and comparability. The purpose of this process is to eliminate interference caused by differences in environmental conditions (such as lighting changes, dust, etc.) so that the subsequent analysis process is more accurate.

[0062] 3.1.2 Baseline establishment

[0063] Establishing a baseline is the first step in the interpretation of results. When taking photos for the first time, managers need to create a first set of canopy images of the citrus tree. This set of images is used to create a baseline of the health of the tree, which will record the health of the tree when there are no obvious lesions or early stages of disease. Baseline data includes:

[0064] Initial value of tree crown density: Based on the image processing algorithm, the initial value of tree crown density is calculated and stored. This value usually represents the pixel ratio occupied by leaves per unit area.

[0065] Health score: Through a comprehensive evaluation of multiple parameters such as crown density, leaf color, leaf shape, etc., an initial health score is generated, usually between 0 and 100. The higher the health score, the better the health of the tree.

[0066] The key parameters in this step include:

[0067] Crown density: the number of pixels occupied by tree crowns per unit area, usually expressed in %.

[0068] Health score: A comprehensive indicator, usually ranging from 0-100, with 100 indicating the healthiest state.

[0069] 3.2. Health trend analysis

[0070] 3.2.1 Multiple image data acquisition

[0071] In the subsequent health assessment process, managers need to collect new image data regularly (it is recommended to collect it every 1-3 months, or adjust the frequency according to the actual situation of the orchard). Each new image collected will be compared with the previous baseline data and the previously collected image data to monitor changes in tree health. This process is completed by calculating changes in crown density and health score changes.

[0072] 3.2.2 Analysis of tree crown density changes

[0073] By comparing crown density at different time points, we can determine the health trend of the trees. If the crown density shows a decreasing trend, especially a significant decrease, it usually indicates that the trees are affected by Huanglongbing or other external factors and their health is deteriorating. On the contrary, if the crown density remains stable or shows an upward trend, it indicates that the health of the trees is good or improving. The formula for calculating the change in crown density is:

[0074]

[0075] Where D t is the crown density at the current time point; D0 is the baseline crown density; ΔD is the crown density change rate (expressed as a percentage).

[0076] 3.2.3 Health score trend analysis

[0077] The health score is a comprehensive score calculated by combining multiple indicators such as crown density, leaf color, and leaf shape. The changing trend of the health score can more intuitively reflect the overall health of the tree. Compared with the crown density, the health score not only takes into account the coverage area of ​​the crown, but also other visual characteristics of the leaves (such as color changes, curling, etc.), which are important for the early detection of Huanglongbing and the evaluation of the health status of trees with Huanglongbing. The calculation formula for the health score trend is:

[0078] ΔS=S t -S0

[0079] Where: S t is the health score at the current time point; S0 is the baseline health score; ΔS is the change value of the health score.

[0080] 3.2.4 Trend graph drawing

[0081] To help managers understand health changes more intuitively, trend graphs are drawn based on changes in crown density and health scores. Trend graphs will show the health status of trees at different time points, and show the dynamic changes in tree health in the form of graphs. The following are two possible forms of trend graphs:

[0082] Health Upward Trend: A gradual increase in crown density and health score in the trend graph indicates that tree health is improving, possibly due to management practices such as trunk injections or nutrient management having a positive effect.

[0083] Declining health trend: A consistent decline in crown density and health scores in the trend graph may indicate a deterioration in tree health, especially as Huanglongbing may be eroding tree health.

[0084] 4. Management suggestion generation

[0085] Based on the health trend analysis, management recommendations are generated for each tree. The generation of management recommendations is based on the setting of management recommendation rules, which are set based on health scores and crown density changes.

[0086] The present invention also relates to the application of a citrus Huanglongbing tree health assessment method based on image processing and algorithms in identifying whether citrus trees are suffering from Huanglongbing. The application of this method to detecting and identifying whether citrus trees are suffering from Huanglongbing also requires the use of PCR testing to verify the results. At the same time, the method of this patent application is aimed at evergreen plants and can also be used for growth tracking of non-green plants.

[0087] The citrus tree health assessment method of the present invention is based on the combination of image processing and algorithm, and has the following beneficial effects:

[0088] 1. Timely monitoring: This method can realize timely monitoring of the growth status of citrus trees, obtain health information in a timely manner, provide scientific basis for farmers, and support quick decision-making.

[0089] 2. Enhanced adaptability: This method is applicable to citrus trees of different species and different growing environments. It can flexibly respond to diverse agricultural production needs and provide strong support for farmers in different regions.

[0090] 3. Sustainable development: By accurately assessing the health of citrus trees, it can help farmers to rationally apply fertilizers and control pests and diseases, promote the development of ecological agriculture, reduce the use of chemical agents, and promote sustainable agricultural practices.

[0091] 4. Data accumulation and analysis: The present invention can also form a health assessment database by accumulating and analyzing long-term data, provide data support for subsequent scientific research, and promote technological progress in the citrus industry.

[0092] Through the above beneficial effects, the present invention provides an effective technical means for the health management of citrus trees and promotes the scientific and intelligent development of the citrus industry. BRIEF DESCRIPTION OF THE DRAWINGS

[0093] Contents expressed in the drawings of the specification:

[0094] Figure 1 This is a bird's-eye view and quadrant zoning diagram of citrus trees affected by Huanglongbing in the embodiment.

[0095] Figure 2 This is a state diagram of the embodiment when photographing the tree crown.

[0096] Figure 3 These are the pre-processed images of the crowns of the four quadrants of the trees infected with Huanglongbing collected for the first time.

[0097] Figure 4 The pre-processed images of the crowns of the four quadrants of the trees infected with Huanglongbing were collected for the second time.

[0098] Figure 5 These are pre-processed images of the crowns of four quadrants of trees infected with Huanglongbing collected three times.

[0099] Figure 6 The linear graph of the health status of citrus trees suffering from Huanglongbing is shown in the example.

[0100] Figure 7 This is a graph showing the crown density changes of citrus trees suffering from Huanglongbing in Example.

[0101] Figure 8 This is a graph showing the health trend changes of citrus trees affected by Huanglongbing in Example 1. DETAILED DESCRIPTION

[0102] The present invention will be further described below with reference to a non-limiting embodiment of the present invention. However, it should be understood that these descriptions are merely illustrative and are not intended to limit the scope of the present invention. In addition, in the following description, the description of the known technology is omitted to avoid unnecessary confusion of the concept of the present invention.

[0103] Example

[0104] 1. Data Collection

[0105] Select 5 or more trees in the orchard that have citrus greening disease for testing. Figure 1 , 2 As shown, Figure 1 For citrus trees with Huanglongbing, images of the crown were taken in four quadrants (east, south, west, and north) between the trunk and the crown edge. Figure 2 As shown, a smartphone was used to take photos of the tree crown in four quadrants of each tree. The phone was mounted on a selfie stick and the shooting position was selected in the middle of the trunk and the edge of the crown. After the shooting position was selected in each quadrant, it was a fixed position for future shooting. The phone settings for each shooting were the same. Images should be saved in JPEG, PNG, TIFF, or RAW format, with a resolution of 704x358 pixels or more and true color. Ensure that each image to be processed is clear and complete for feature extraction in the subsequent image processing steps.

[0106] 2. Image Processing and Analysis

[0107] 2.1 Image Preprocessing

[0108] 2.1.1 Image Denoising

[0109] When shooting an image, changes in ambient light, dust, and noise from the camera itself will affect the image quality, so it is necessary to first remove these noise interferences. This embodiment uses Gaussian filtering to remove noise: by applying a Gaussian function filter to the image, the tiny noise points in the image are smoothed while keeping the larger image structure intact.

[0110] 2.1.2 Color Standardization

[0111] Due to different lighting conditions at different time periods, the color of tree leaves may have large deviations in the image. Therefore, it is necessary to perform color standardization on the image so that images taken at different times and under different lighting conditions have relatively consistent color performance to ensure the comparability of subsequent analysis. This embodiment uses white balance adjustment and histogram equalization to standardize the crown images in the four quadrants of each tree.

[0112] 2.1.3 Contrast Enhancement

[0113] The purpose of contrast enhancement is to highlight the difference between the tree crown and the background for subsequent image segmentation and edge detection. A commonly used contrast enhancement method is histogram stretching, which stretches the brightness range of the image to the maximum dynamic range to make the difference between the dark and bright areas more obvious.

[0114] 2.1.4 Image Cropping and Adjustment

[0115] According to the shooting standards, the image is cropped to retain only the crown part and ensure that irrelevant background parts are not included in the analysis. Figure 3 Initial photographs of the crown treatments in four quadrants of a citrus tree infected with Huanglongbing are shown.

[0116] 2.2 Calculation of crown density

[0117] Crown density is one of the important parameters that reflect the health of trees. It quantifies the density of the crown by calculating the pixel ratio of leaves per unit area. The calculation of crown density goes through the following steps:

[0118] 2.2.1 Image Segmentation

[0119] The purpose of image segmentation is to separate the crown area from the background and extract the pixel area occupied by the leaves. The image segmentation method used in this embodiment includes an edge detection algorithm, which finds the edge in the image by detecting the gradient change of the pixel gray value in the image, and is suitable for images with obvious contrast between the crown and the background.

[0120] 2.2.2 Binarization

[0121] After the segmentation is completed, the crown area is separated from the background, the background pixels are set to 0, and the crown pixels are set to 1. The key to binarization is to choose a suitable threshold. The global threshold can be set according to the brightness histogram of the image, or an adaptive threshold method can be used to enable different parts of the image to be adaptively adjusted according to the local contrast.

[0122] 2.2.3 Calculation formula for crown density

[0123] like Figure 3 As shown, the total number of pixels in the image is 249,000, of which the number of pixels in the crown area is 1,743,000. The crown density is:

[0124] D0=(1743000 / 2490000)*100%=70%

[0125] This means that the canopy of this tree covers 70% of the image.

[0126] like Figure 4 As shown, the total number of pixels in the image is 249,000, of which the number of pixels in the crown area is 1,619,000. The crown density is:

[0127] D1=(1619000 / 2490000)*100%=65%

[0128] This means that the canopy of this tree covers 65% of the image.

[0129] like Figure 3 As shown, the total number of pixels in the image is 250,000, of which the number of pixels in the crown area is 1,525,000. The crown density is:

[0130] D2=(1525000 / 2500000)*100%=61%

[0131] This means that the canopy of this tree covers 61% of the image.

[0132] 2.3 Health score generation

[0133] The health score is a numerical indicator generated by combining multiple image features (including crown density, leaf color, shape, edges, etc.) to quantify the overall health of the tree. The score range is usually 0 to 100, where 0 means the tree is completely withered and 100 means the tree is in optimal health. The generation of the health score involves the following steps:

[0134] 2.3.1 Leaf color analysis

[0135] Leaf color is one of the important characteristics for judging the early symptoms of Huanglongbing. Healthy citrus tree leaves are dark green, while leaves infected with Huanglongbing may have yellow spots or turn yellow overall. By analyzing the changes in leaf color, it can be used to make a preliminary early judgment on whether citrus trees have symptoms of Huanglongbing.

[0136] Color analysis is based on color space conversion and color histogram statistics:

[0137] Color space conversion: Convert the image from RGB color space to HSV (hue, saturation, brightness) space. HSV space is more in line with the visual perception characteristics of the human eye, where hue can be used to distinguish the color changes of leaves, and saturation can be used to judge the brightness of the color.

[0138] Tone analysis: Extract the hue distribution histogram of the leaf area and count the proportion of hues in the yellow-green range (usually between 60° and 120°). If the hue tends to be yellow and the proportion increases, it may indicate that the leaf is affected by Huanglongbing.

[0139] 2.3.2 Leaf shape and curl detection

[0140] Huanglongbing may also cause changes in leaf shape, such as curling and edge deformation. Shape analysis can be used to detect whether leaves have abnormal curling. Shape analysis can use Hough transform to detect leaf edges, or calculate leaf shape parameters (such as edge curvature, aspect ratio, etc.) through morphological processing (such as erosion and expansion operations).

[0141] 2.3.3 Health score calculation formula

[0142] The calculation of health score comprehensively considers multiple image features and generates the final health score through weighted formula. Through the above analysis of the first, second and third collected images and image processing and analysis, the first collection is the base point, and the first, second and third collection images are collected 3 months apart. After image processing, the following is obtained Figure 3 , Figure 4 , Figure 5 The images and corresponding data of the three time nodes shown in the figure finally obtained the analysis data as shown in Table 1, which is the shape abnormality score of the crown density, leaf color, leaf shape and leaf edge curvature of one of the trees infected with Huanglongbing at the three time nodes, as follows:

[0143] Table 1 The results of scoring the crown image features of the same Huanglongbing tree collected three times

[0144] Initial collection Second collection The third collection Crown density 70 65 61 Leaf color 70 63 62 Blade shape 80 75 71 Leaf edge curvature 80 76 72

[0145] When the data in Table 1 above were calculated separately, some were counted as percentages, which are converted here into values ​​of 0-100 for the convenience of calculating the health score. The following is the health score S0 obtained when the tree was first collected, the health score S1 obtained when the tree was second collected, and the health score S2 obtained when the tree was third collected, which are finally obtained through the weighted formula based on the data in Table 1.

[0146] S0=0.3*70+0.4*70+0.2*80+0.1*80=73

[0147] S1=0.3*65+0.4*63+0.2*75+0.1*76=67.3

[0148] S2=0.3*61+0.4*62+0.2*71+0.1*72=64.5

[0149] 3. Interpretation of results

[0150] 3.1 Data preprocessing and baseline establishment

[0151] 3.1.1 Data Preprocessing

[0152] Before interpreting the results, you first need to ensure that the captured image data has been adequately preprocessed. Preprocessing steps usually include image denoising, color standardization, and contrast adjustment. Through these preprocessing steps, all images are ensured to have consistent quality and comparability. The purpose of this process is to eliminate interference caused by differences in environmental conditions (such as lighting changes, dust, etc.) so that the subsequent analysis process is more accurate.

[0153] 3.1.2 Baseline establishment

[0154] like Figure 6 As shown, baseline establishment is the first step in the interpretation of results. When the first photo is taken, managers create the first set of canopy images of the citrus tree. This set of images is used to create a baseline of the health of the tree, which records the health of the tree when there are no obvious lesions or early stages of disease. Baseline data includes:

[0155] Initial value of tree crown density: Based on the image processing algorithm, calculate and store the initial value of tree crown density. This value usually represents the pixel ratio occupied by leaves per unit area. The number of pixels occupied by the tree crown per unit area is usually expressed in %.

[0156] Health score: An initial health score is generated by comprehensively evaluating multiple parameters such as crown density, leaf color, leaf shape, etc., usually between 0 and 100. The higher the health score, the better the health of the tree. The health score is a comprehensive indicator, usually ranging from 0-100, where 100 represents the healthiest state.

[0157] like Figure 6 As shown, the horizontal line is the scoring line. According to the different time nodes of the photos, the vertical line at 0 on the bisector is the baseline of the tree health status, and 1 and 2 on the bisector represent the data corresponding to the image processing and analysis after the second and third photos, respectively. From the time nodes, the citrus trees in this example develop in the direction of decreasing scores, which shows that Huanglongbing has a significant adverse effect on the citrus trees.

[0158] 3.2. Health trend analysis

[0159] 3.2.1 Calculation of crown density changes

[0160] Based on the data collected and analyzed above, the change rate of crown density during the second and third data collection can be calculated according to the calculation formula of crown density change.

[0161] ΔD1=

(D1-D0) / D0

(65-70) / 70

[0162] ΔD2=

(D2-D0) / D0

(61-70) / 70

[0163] Where D1 and D2 are the crown densities at the second and third shooting time points; D0 is the baseline crown density; ΔD is the crown density change rate (expressed as a percentage).

[0164] The above calculations show that the crown density of the tree decreased by 7% in the second shot and by 13% in the third shot. Figure 7 As shown, the tree's health may be deteriorating.

[0165] 3.2.3 Health score trend analysis

[0166] According to the data collected and analyzed above, the tree health score trends at the second and third data collection times can be calculated respectively according to the tree health scores obtained at the three shooting times, including S0, S1, and S2, using the calculation formula for the health score trend.

[0167] ΔS1=S1-S0=67.3-73=-5.7

[0168] ΔS2=S2-S0=64.5-73=-8.7

[0169] Among them: S1 and S2 are the health scores at the second and third shooting time points; S0 is the baseline health score; ΔS is the change value of the health score.

[0170] The above calculations show that the tree's health score decreased by 5.7 in the second shot and by 8.7 in the third shot. Figure 8 shown, indicating that the tree's health may be deteriorating.

[0171] 3.2.4 Trend graph drawing

[0172] like Figure 7 , Figure 8 As shown in the figure, in order to help managers understand the health changes more intuitively, a trend chart is drawn based on the changes in crown density and health score. The trend chart will show the health status of trees at different time points and show the dynamic changes of tree health in the form of charts. Figure 7 and Figure 8 Both show a downward trend in the health of the trees. The crown density and health score in the trend chart continue to decline, which may indicate that the health of the trees is deteriorating, especially Huanglongbing may be damaging the health of the trees.

[0173] 4. Management suggestion generation

[0174] Based on the health trend analysis, management recommendations are generated for each tree. Management recommendations are generated through the following steps:

[0175] 4.1 Management suggestion rule setting

[0176] A set of management rules based on health scores and crown density changes are set, which combine years of orchard management experience and Huanglongbing prevention and control programs. The management rules are as follows:

[0177] Continue to observe: If the tree health score remains above 80 and the canopy density remains near the baseline, managers are advised to continue observation and take new images again in three months.

[0178] Increase prevention and control efforts: If the tree health score drops to between 60-80 points and the crown density decreases between 5%-10%, it is recommended that managers increase prevention and control measures, such as increasing the frequency of trunk injections or using more efficient nutrient management programs.

[0179] Removing trees: If a tree's health score drops below 60 points and its crown density continues to decline by more than 10%, managers are advised to remove the tree to prevent the disease from spreading further to other healthy trees.

[0180] 4.2 Dynamic Management Adjustment

[0181] Management recommendations are dynamically adjusted based on each new health assessment result. For example, during a season when disease is most prevalent, managers are advised to increase the frequency of image acquisition to monitor tree health in real time. For trees whose health scores have improved, it is recommended to reduce intervention measures to save resources.

[0182] 4.3 Handling special situations

[0183] During the result interpretation process, if the health score or crown density of a tree drops sharply in a short period of time, this abnormality may be due to the rapid spread of Huanglongbing or other uncontrollable external factors (such as insect pests, extreme weather, etc.). Anomaly detection can help orchard managers detect potential problems early and take protective measures in time.

[0184] Parameter standards for anomaly detection:

[0185] Sharp decline threshold: Set the sharp decline threshold of crown density change to 10% or higher. If the crown density drops by more than 10% within a detection cycle, the tree will be marked as abnormal.

[0186] Sharp decline threshold: Trees whose health score drops by more than 15 points will be considered to have a sharp decline in health, marked as extremely abnormal, and recommended for further inspection.

[0187] The above embodiments should be understood to be only used to illustrate the present invention and not to limit the protection scope of the present invention. After reading the contents of the present invention, technicians can make various changes or modifications to the present invention, and these equivalent changes and modifications also fall within the scope defined by the claims of the present invention.

Claims

1. A method for assessing the health of citrus Huanglongbing trees based on image processing and algorithms, characterized in that: The following steps are involved: St1. Data collection: 1) Select several representative citrus trees with Huanglongbing in the orchard, and take crown images in several quadrants between the trunk and the crown edge of each tree to ensure that the collected data is comprehensive and representative; 2) Use high-quality camera equipment for image acquisition, ensure that the settings of the equipment used are consistent before shooting, and adopt a standardized shooting process, including choosing appropriate shooting angles, lighting conditions, and shooting heights to reduce the impact of external factors on image quality; Before each shooting, calibrate the shooting equipment to ensure the consistency of shooting results; 3) Regularly monitor and photograph the selected trees once every 1-3 months to track the dynamic changes in the health of the trees and establish a long-term data set for analysis; record the time of each photograph, weather conditions, soil moisture and other environmental factors so that these influencing factors can be considered in subsequent analysis; St2. Image processing: 1) Perform image denoising on the collected crown images; 2) Perform color standardization on the denoised image; 3) Performing contrast enhancement on the standardized image; 4) Image cropping and adjustment: crop the image and keep the crown part; St3. Crown density calculation 1) Image segmentation: Separate the crown area from the background area and extract the outline and morphological features of the crown; 2) Binarization: After segmentation, separate the crown area from the background and select a suitable threshold; 3) Calculation of tree crown density; St4. Leaf color analysis Leaf color analysis is based on color space conversion and color histogram statistics: 1) Color space conversion: convert the image from RGB color space to HSV space; 2) Hue analysis: extract the hue distribution histogram of the leaf area and count the proportion of hues in the yellow-green range; St5, leaf shape and curl detection Huanglongbing can cause changes in leaf shape. Shape analysis can be used to detect whether leaves are curling abnormally. St6. Health score calculation The calculation of the health score takes into account multiple image features and obtains the health score of the citrus tree through calculation; St7, baseline establishment and data processing Baseline establishment: When taking photos for the first time, you need to create the first set of citrus tree crown images. This set of images is used to create a baseline for the health of the tree. The baseline will record the health of the tree when there are no obvious lesions or diseases. The baseline data includes: Initial crown density value: Calculate and store the initial value of the tree crown density; Health score: Through comprehensive evaluation of multiple parameters such as crown density, leaf color, leaf shape, etc., an initial health score S0 is generated. The higher the health score, the better the health of the tree; St8. Generate management recommendations for each tree based on health trend analysis.

2. The citrus Huanglongbing tree health assessment method based on image processing and algorithm as claimed in claim 1, characterized in that: The image denoising method in step St2 includes Gaussian filtering, median filtering or bilateral filtering; wherein the Gaussian filtering parameters are set to: Here, σ is the standard deviation of the filter, which is usually set between 1 and 2 to ensure that it can remove noise without over-smoothing details.

3. The citrus Huanglongbing tree health assessment method based on image processing and algorithm as claimed in claim 1, characterized in that: The contrast enhancement processing method in step St2 is histogram stretching, and the mathematical formula is: Among them, I(x,y) is the pixel value of the original image, I min and I max are the minimum and maximum brightness values ​​in the image, respectively.

4. The citrus Huanglongbing tree health assessment method based on image processing and algorithm as claimed in claim 1, characterized in that: The calculation of the crown density in step St3 is based on the number of pixels in the crown area per unit area. The specific formula is as follows: Where: N tree is the number of tree crown pixels, N total The total number of pixels in the image, D is the tree crown density, in percentage.

5. The citrus Huanglongbing tree health assessment method based on image processing and algorithm as claimed in claim 1, characterized in that: In step St6, the final health score is generated by a weighted formula, which is as follows: S=w1D+w2C+w3L+w4E Among them: S is the health score; D is the crown density score (accounting for 30% of the total score, that is, w1=0.3); C is the leaf color score (accounting for 40% of the total score, that is, w2=0.4); L is the leaf shape score (accounting for 20% of the total score, that is, w3=0.2); E is the shape abnormality score such as edge curvature (accounting for 10% of the total score, that is, w4=0.1).

6. The citrus Huanglongbing tree health assessment method based on image processing and algorithm as claimed in claim 1, characterized in that: Calculating the crown density change in step St7 includes the following steps: By comparing the crown density at different time points, we can determine the health trend of the trees. If the crown density shows a decreasing trend, it indicates that the trees may be affected by Huanglongbing or other external factors and their health is deteriorating. On the contrary, if the crown density remains stable or shows an upward trend, it indicates that the health of the trees is good or improving. The calculation formula for crown density changes is: Where: D t is the crown density at the current time point, D0 is the baseline crown density, and ΔD is the crown density change rate.

7. The method for evaluating the health of citrus Huanglongbing trees based on image processing and algorithm as claimed in claim 1, characterized in that: Calculating the health score trend in step St7 includes the following steps: The health score is a comprehensive score calculated by combining multiple indicators such as crown density, leaf color, leaf shape, etc. The calculation formula for the health score trend is: △S=S t -S0 Where: S t is the health score at the current time point, S0 is the baseline health score, and ΔS is the change in health score.

8. The method for assessing the health of citrus Huanglongbing trees based on image processing and algorithm according to claim 1 or 7, characterized in that: Calculating the health score trend in step St7 also includes drawing a health trend graph and analyzing: Trend charts show the health of trees at different points in time, and show the dynamic changes in tree health in the form of charts. The following are two forms of trend charts: Health rising trend: The crown density and health score in the trend graph gradually increase, indicating that the health of the trees is improving; Declining health trend: The canopy density and health score in the trend graph continue to decrease, indicating that the health of the trees is deteriorating; During subsequent health assessments, managers need to compare new images with previous baseline data and image data regularly, that is, based on the frequency of capture, to monitor changes in tree health. This process is done by calculating changes in crown density and health score trends.

9. The method for evaluating the health of citrus Huanglongbing trees based on image processing and algorithm as claimed in claim 1, characterized in that: Step St8 also includes the setting of management recommendation rules, which are set based on the health score and the crown density change.

Images