Remote sensing monitoring method for typical diseases and insect pests in forestry
Through multi-band comprehensive analysis and feature zoning technology, the accuracy and timeliness problems of pest and disease monitoring in existing technologies have been solved, and early and sharp identification and dynamic tracking of pests and diseases have been achieved, providing a scientific basis for prevention and control.
Patent Information
- Application Number
- CN202510550819.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-28
- Publication Date
- 2025-09-12
AI Technical Summary
Existing remote sensing monitoring methods for forestry pests and diseases are unable to accurately capture early subtle spectral changes, resulting in large errors in monitoring results and a high missed detection rate. In addition, there is a lack of effective tracking of the dynamic spread of pests and diseases, making it difficult to achieve timely early warning.
A multi-band comprehensive analysis method is adopted to construct spectral feature indicators by selecting feature points in the blue, green and red light bands. The Sobel algorithm is used to divide the feature partitions. The vertical distance L between the low-order feature line and the high-order feature point is quantified, combined with infrared spectrum and verification features, to identify abnormal points and monitor their diffusion.
It achieves early and sharp identification of pests and diseases, reduces interference from environmental factors, improves monitoring accuracy and timeliness, and provides scientific prevention and control strategies for forestry management.
Smart Images
Figure CN120629038A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of forestry monitoring, in particular to a remote sensing monitoring method for typical forestry pests and diseases. Background Art
[0002] With growing global awareness of forest resource conservation, forest pest and disease monitoring has become a critical component in maintaining ecological balance and ensuring sustainable forestry development. Traditional forest pest and disease monitoring methods, such as manual inspections and plot surveys, suffer from limitations such as limited scope, poor timeliness, and high labor costs, making them inadequate for large-scale, dynamic monitoring.
[0003] In recent years, remote sensing technology, leveraging its advantages of macroscopic, rapid, and periodic observation, has gradually become a key tool for monitoring forest pests and diseases. Existing methods, including some that analyze changes in vegetation spectral reflectance or calculate vegetation indices, identify pests and diseases. However, these methods often rely on single-band or simple spectral feature analysis, making it difficult to accurately capture subtle spectral changes in the early stages of pests and diseases. This can lead to significant errors in monitoring results and a high rate of missed detections.
[0004] Furthermore, in remote sensing image processing, traditional zoning methods often use fixed thresholds or simple clustering algorithms, which fail to fully account for the complex boundary characteristics of forestry areas. This results in insufficient zoning accuracy, impacting the accuracy of subsequent monitoring results. Furthermore, most existing technologies lack effective tracking of the dynamic spread of pests and diseases, making it difficult to provide timely early warning of pest and disease spread trends.
[0005] Therefore, there is an urgent need for a remote sensing monitoring method that can achieve high-precision image partitioning, deep mining of spectral features, and accurate dynamic monitoring of pests and diseases to meet the actual needs of modern forestry pest and disease control. Summary of the Invention
[0006] In view of the shortcomings of the existing technology, the present invention provides a remote sensing monitoring method for typical forestry pests and diseases, which solves the problem of insufficient spectral feature mining.
[0007] To achieve the above objectives, the present invention is implemented through the following technical solutions: A remote sensing monitoring method for typical forestry pests and diseases comprises the following steps:
[0008] Step 1: Based on the remote sensing image associated with the corresponding forestry monitoring area, the image features associated with the remote sensing image are initially processed and the remote sensing image is divided into multiple feature partitions. The specific method is as follows:
[0009] Gray-scale the associated remote sensing image to confirm the gray-scale image associated with the corresponding remote sensing image;
[0010] Based on the different pixels associated with the grayscale image, the grayscale values associated with the different pixels are determined, and then the vertical gradient and vertical gradient associated with the corresponding pixels are determined based on the Sobel algorithm;
[0011] And adopt: Lock the integrated gradient associated with the corresponding pixel point, and mark the pixel point that satisfies the following conditions: integrated gradient > Y1 as a gradient pixel point, and its Y1 is the preset value. Otherwise, no calibration is performed;
[0012] Marking a number of gradient pixel points in the remote sensing image in sequence, and based on the continuously marked gradient pixel points, confirming the gradient profiles associated with the continuous gradient pixel points, and marking them in the remote sensing image, confirming the area included in each gradient profile, recording the corresponding included area as a single feature partition, and dividing the remote sensing image into multiple feature partitions;
[0013] Step 2: Based on the several characteristic partitions marked in the remote sensing image, confirm the RGB mean associated with each characteristic partition, and based on the specific confirmation results, lock the partition to be monitored. The specific method is as follows:
[0014] Based on the calibrated feature partition, confirm the RGB values associated with different points in the corresponding partition, then average the RGB values associated with several different points, and lock the feature value to be verified associated with the corresponding feature partition;
[0015] Compare the confirmed characteristic values to be verified with the preset numerical intervals, and the endpoint values of the numerical intervals are all preset values. If the characteristic values to be verified all belong to the associated numerical intervals, then the characteristic partition is calibrated as the partition to be monitored. Otherwise, no calibration is performed;
[0016] Step 3: Based on the calibrated monitored partitions, confirm the detection spectrum associated with different monitored partitions, lock the spectral features associated with different partition points and record them. The specific method is as follows:
[0017] Confirm the detection spectrum associated with different partition points in the monitored partition at the current moment, where the horizontal axis of the detection spectrum is wavelength data and the vertical axis is reflectivity data;
[0018] The specific bands associated with (435-495nm), (525-565nm) and (646-673nm) are locked from the corresponding detection spectrum. The point with the lowest reflectivity value is selected from the specific bands associated with (435-495nm) and (646-673nm), and the selected point with the lowest reflectivity value is calibrated as a low-level feature point. The point with the highest reflectivity value is selected from the specific band associated with (525-565nm) and calibrated as a high-level feature point.
[0019] = Connect the two groups of low-level feature points to determine a group of low-level feature lines, then determine the vertical distance L between the high-level feature point and the feature line, and record the determined vertical distance L as the spectral feature of the current partition point at the current moment;
[0020] Then, different spectral features associated with different partition points are calibrated and confirmed in turn;
[0021] Step 4: Based on the real-time detection spectrum generated by different partition points in the monitored partition, confirm the real-time associated spectral features, and calibrate the abnormal points based on the feature changes associated with the corresponding partition points. The specific method is as follows:
[0022] Based on the real-time detection spectrum generated by different partition points, confirm the real-time associated spectral features, and sort and confirm the real-time confirmed spectral features according to the time sequence of the feature relationship, and confirm the spectral feature sorting sequence associated with the corresponding partition point:
[0023] In the spectral feature sorting sequence, if the latter set of spectral features is lower than the former set of spectral features, a set of low-variable signals is generated. If the low-variable signals are generated three times in a row, the corresponding partition point is calibrated as the pending point. Otherwise, the pending point is not calibrated.
[0024] Then confirm the infrared spectrum associated with the pending point at the current moment: lock the high-level feature point from the confirmed infrared spectrum, and confirm the band turning points associated with both sides of the high-level feature point;
[0025] The band associated with the two groups of band turning points and the high-level feature points is recorded as the pending band. The band characteristics associated with adjacent points in the pending band are confirmed: its band characteristic = |reflectivity value difference between adjacent points| ÷ time interval. Based on the confirmed several band characteristics, the maximum value is selected as the verification feature of the pending point at the current moment;
[0026] Then, the calibration features associated with the next pending point are continuously monitored, and the calibration value difference is confirmed. The calibration value difference = |calibration feature at the current moment - calibration feature at the next moment|. If the calibration value difference is greater than Y2, the pending point is calibrated as an abnormal point, and its Y2 is the preset value; if the calibration value difference is ≤ Y2, no calibration is performed.
[0027] Preferably, the (435-495 nm) corresponds to the blue light band, the (525-565 nm) corresponds to the green light band, and the (646-673 nm) corresponds to the red light band.
[0028] Preferably, the step 4 further comprises:
[0029] Identify whether other points associated with the abnormal point in the subsequent process are marked as abnormal points. If so, it means that there is pest impact, and the abnormal point area associated with the pest image will be displayed, and timely response measures will be taken. If not, continue monitoring.
[0030] The present invention provides a remote sensing monitoring method for typical forestry pests and diseases. Compared with the existing technology, it has the following advantages:
[0031] This method innovatively selects characteristic points in the blue, green, and red light bands to construct spectral characteristic indicators. By quantifying the vertical distance L between the low-order characteristic line and the high-order characteristic point, it can keenly capture subtle changes in the vegetation spectral curve. Compared with traditional single-band analysis, this multi-band comprehensive analysis method can more comprehensively and accurately reflect the health of vegetation. Especially in the early stages of pests and diseases, when vegetation spectral changes are not yet obvious, it can also effectively identify potential pathological characteristics, providing strong support for early warning of pests and diseases.
[0032] By determining the pending points through three consecutive low-variable signals and combining the comprehensive analysis of infrared spectra and verification features, the system can effectively eliminate the spectral anomaly interference caused by environmental factors (such as weather changes and light intensity fluctuations) and accurately identify the abnormal areas actually caused by pests and diseases. At the same time, by monitoring the spread of abnormal points, the spread trend of pests and diseases can be grasped in a timely manner, providing an accurate basis for forestry management departments to formulate scientific and efficient prevention and control strategies, thereby minimizing the damage of pests and diseases to forest resources. BRIEF DESCRIPTION OF THE DRAWINGS
[0033] Figure 1 Schematic diagram of the process of the present invention;
[0034] Figure 2 Schematic diagram for determining the spectral characteristics of the present invention. DETAILED DESCRIPTION
[0035] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only 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 ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.
[0036] First embodiment
[0037] See also Figure 1 , this application provides a remote sensing monitoring method for typical forestry pests and diseases, comprising the following steps:
[0038] Step 1: Based on the remote sensing image associated with the corresponding forestry monitoring area, the image features associated with the remote sensing image are preliminarily processed, and the remote sensing image is divided into multiple feature partitions. Specifically, in combination with the specific verification of the grayscale feature value in the image, a number of gradient pixel points existing in the remote sensing image are calibrated in sequence, and then based on the confirmed number of gradient pixel points, the corresponding gradient profile is confirmed, and then based on the divided number of gradient profiles, the specific division of the area is performed, and the corresponding feature partition is locked;
[0039] The feature partitioning method specifically includes:
[0040] Grayscale the associated remote sensing image to confirm the grayscale image associated with the corresponding remote sensing image. The corresponding RGB value exists in the remote sensing image. After weighting different RGB values, the grayscale value associated with the corresponding point can be obtained. Then, grayscale adjustment is performed on the remote sensing image based on the grayscale value to confirm the grayscale image associated with the corresponding remote sensing image.
[0041] Based on the different pixels associated with the grayscale image, the grayscale values associated with the different pixels are confirmed. Then, the vertical gradient and vertical gradient associated with the corresponding pixel are confirmed based on the Sobel algorithm. For the corresponding vertical gradient and vertical gradient, the weights of the associated pixel and the surrounding pixels are confirmed. Based on the specific associated weights, the corresponding vertical gradient and vertical gradient can be locked.
[0042] And adopt: Lock the integrated gradient associated with the corresponding pixel point, and mark the pixel point that satisfies the following conditions: integrated gradient > Y1 as a gradient pixel point. Otherwise, no calibration is performed. Y1 is a preset value, and its specific value is determined by the operator based on experience.
[0043] Marking a number of gradient pixel points in the remote sensing image in sequence, and based on the continuously marked gradient pixel points, confirming the gradient profiles associated with the continuous gradient pixel points, and marking them in the remote sensing image (that is, after the corresponding gradient pixel points are confirmed, if the pixels are continuous, the continuous gradient pixel points can be regarded as a gradient profile, thereby sequentially determining a number of gradient profiles in the remote sensing image), confirming the area included in each gradient profile, recording the corresponding included area as a single feature partition, and dividing the remote sensing image into multiple feature partitions;
[0044] Among them, in the remote sensing image, for the existing forestry area, the edge points are all associated gradient pixel points. According to the confirmed gradient pixel points, several gradient pixel points can be combined for field confirmation, the corresponding gradient contour can be locked, and then different feature partitions can be calibrated in turn based on the locked gradient contour;
[0045] Step 2: Based on the several characteristic partitions calibrated in the remote sensing image, confirm the RGB mean associated with each characteristic partition, and based on the specific confirmation results, lock the partition to be monitored. Its RGB value can effectively show the greenness of the corresponding partition. According to the specific numerical monitoring process, the area with greenness can be quickly locked and subsequent analysis and confirmation can be carried out. The specific method for locking the partition to be monitored is as follows:
[0046] Based on the calibrated feature partition, confirm the RGB values associated with different points in the corresponding partition, then perform average processing on the RGB values associated with several different points, lock the feature value to be verified associated with the corresponding feature partition, and in the process of average processing, perform average feature processing on its R value, G value and B value separately in turn, so that the feature value to be verified is determined to be the average of three groups;
[0047] The confirmed feature value to be verified is compared with the preset numerical interval (different means are associated with different numerical intervals). The endpoint values of the numerical interval are all preset values, which are prepared in advance by the operator based on experience. If the feature values to be verified all belong to the associated numerical interval, then this feature partition is calibrated as the partition to be monitored. Otherwise, no calibration is performed, which means that the RGB feature associated with the corresponding feature partition meets the corresponding green feature, and feature recognition can be determined, the corresponding partition to be monitored is locked, and subsequent numerical verification is convenient;
[0048] Step 3: Based on the calibrated monitored partitions, confirm the detection spectrum graphs associated with different monitored partitions, lock the spectral features associated with different partition points and record them. The specific method of recording is as follows:
[0049] Combine Figure 2 , confirm the detection spectrum graph associated with different partition points in the partition to be monitored at the current moment, the horizontal coordinate axis of the detection spectrum graph is wavelength data, and the vertical coordinate axis is reflectivity data;
[0050] Lock the specific bands associated with (435-495nm), (525-565nm) and (646-673nm) from the corresponding detection spectrum, select the point with the lowest reflectivity from the specific bands associated with (435-495nm) and (646-673nm), and mark the selected point with the lowest reflectivity as a low-level feature point. Then, select the point with the highest reflectivity from the specific band associated with (525-565nm) and mark it as a high-level feature point. (435-495nm) corresponds to the blue light band, (525-565nm) corresponds to the green light band, and (646-673nm) corresponds to the red light band.
[0051] Connect the two groups of low-order feature points to determine a group of low-order feature lines, then determine the vertical distance L between the high-order feature point and the feature line, and record the determined vertical distance L as the spectral feature of the current partition point at the current moment;
[0052] Then, the different spectral features associated with different partition points are calibrated and confirmed in turn. Specifically, in the corresponding remote sensing monitoring data, different partition points have different spectral features, and different wavelengths of light bands have different reflectivities. Therefore, the specific band features associated with the corresponding points can be confirmed from the corresponding spectral graph, which is convenient for subsequent feature verification.
[0053] Step 4: Based on the real-time detection spectrum generated by different partition points of the monitored partition, confirm the real-time associated spectral features, and calibrate the abnormal points based on the feature changes associated with the corresponding partition points. The specific method of calibration is as follows:
[0054] Based on the real-time detection spectrum generated by different partition points, confirm the real-time associated spectral features, and sort and confirm the real-time confirmed spectral features according to the time sequence of the feature relationship, and confirm the spectral feature sorting sequence associated with the corresponding partition point:
[0055] In the spectral feature sorting sequence, if the latter set of spectral features is lower than the former set of spectral features, a set of low-variable signals is generated. If the low-variable signals are generated three times in a row, the corresponding partition point is calibrated as the pending point. Otherwise, the pending point is not calibrated.
[0056] Then confirm the infrared spectrum associated with the pending point at the current moment: lock the high-level feature point from the confirmed infrared spectrum, and confirm the band turning points associated with the high-level feature point on both sides. The band turning point is the change point of the reflectivity value. The trends of the two bands before and after the change point are opposite, so this type of point is the corresponding band turning point. For example: after the high-level feature point is determined, the front-end band trend is downward, and the back-end band trend is downward, then the two band turning points confirmed before and after are the trend change points of the corresponding band;
[0057] The band associated with the two groups of band turning points and the high-level feature points is recorded as the pending band. The band characteristics associated with adjacent points in the pending band are confirmed: its band characteristic = |reflectivity value difference between adjacent points| ÷ time interval. Based on the confirmed several band characteristics, the maximum value is selected as the verification feature of the pending point at the current moment;
[0058] Then, the calibration features associated with the next-to-be-determined point are continuously monitored, and the calibration value difference is confirmed. The calibration value difference = |calibration feature at the current moment - calibration feature at the next moment|. There is a time period between adjacent moments, which may be 24 hours or 48 hours. If the calibration value difference is greater than Y2, the point to be determined is marked as an abnormal point. Otherwise, no calibration is performed. Y2 is a preset value, and its specific value is determined by the operator based on experience.
[0059] The confirmed abnormal points are used as standard points. In the subsequent process, other points associated with the standard points are identified as abnormal points. If so, it means there is pest impact. The abnormal point area associated with the pest image is displayed for external personnel to review and take timely response measures. If not, continue monitoring;
[0060] When pests are present in the green area, the pests will spread, which will cause the corresponding abnormal points to spread, and cause other points associated with the corresponding points to have corresponding abnormal conditions.
[0061] Some of the data in the above formulas are dimensionless and numerically calculated. Meanwhile, the contents not described in detail in this specification belong to the prior art known to those skilled in the art.
[0062] The above embodiments are only used to illustrate the technical method of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.
Claims
1. A remote sensing monitoring method for typical forestry pests and diseases, characterized in that: The following steps are involved: Step 1: Based on the remote sensing image associated with the corresponding forestry monitoring area, the image features associated with the remote sensing image are preliminarily processed to divide the remote sensing image into multiple feature partitions; Step 2: Based on several characteristic partitions marked in the remote sensing image, the RGB mean associated with each characteristic partition is confirmed, and based on the specific confirmation results, the partition to be monitored is locked; Step 3: Based on the calibrated monitored partitions, confirm the detection spectrum associated with different monitored partitions, lock the spectral features associated with different partition points and record them; Step 4: Based on the real-time detection spectrum generated by different partition points of the monitored partition, confirm the real-time associated spectral features, and calibrate the abnormal points based on the feature changes associated with the corresponding partition points.
2. A remote sensing monitoring method for typical forestry pests and diseases according to claim 1, characterized in that: In step 1, the specific method of dividing the feature partitions is: Gray-scale the associated remote sensing image to confirm the gray-scale image associated with the corresponding remote sensing image; Based on the different pixels associated with the grayscale image, the grayscale values associated with the different pixels are determined, and then the vertical gradient and vertical gradient associated with the corresponding pixels are determined based on the Sobel algorithm; And adopt: Lock the integrated gradient associated with the corresponding pixel point, and mark the pixel point that satisfies the following conditions: integrated gradient > Y1 as a gradient pixel point, where Y1 is the preset value; Several gradient pixel points in the remote sensing image are marked in sequence, and based on the continuously marked gradient pixel points, the gradient profiles associated with the continuous gradient pixel points are confirmed and marked in the remote sensing image, and the area included in each gradient profile is confirmed. The corresponding included area is recorded as a single feature partition, and the remote sensing image is divided into multiple feature partitions.
3. A remote sensing monitoring method for typical forestry pests and diseases according to claim 2, characterized in that: If the integrated gradient associated with the corresponding pixel point does not satisfy: integrated gradient > Y1, no calibration is performed.
4. The remote sensing monitoring method for typical forestry pests and diseases according to claim 1, characterized in that: In step 2, the specific method of locking the monitored partition is: Based on the calibrated feature partition, confirm the RGB values associated with different points in the corresponding partition, then average the RGB values associated with several different points, and lock the feature value to be verified associated with the corresponding feature partition; The confirmed characteristic values to be verified are compared with the preset numerical intervals, and the endpoint values of the numerical intervals are all preset values. If the characteristic values to be verified all belong to the associated numerical intervals, then this characteristic partition is calibrated as the partition to be monitored. Otherwise, no calibration is performed.
5. The remote sensing monitoring method for typical forestry pests and diseases according to claim 1, characterized in that: In step 3, the specific method of recording the spectral features associated with different partition points is: Confirm the detection spectrum associated with different partition points in the monitored partition at the current moment, where the horizontal axis of the detection spectrum is wavelength data and the vertical axis is reflectivity data; The specific bands associated with (435-495nm), (525-565nm) and (646-673nm) are locked from the corresponding detection spectrum. The point with the lowest reflectivity value is selected from the specific bands associated with (435-495nm) and (646-673nm), and the selected point with the lowest reflectivity value is calibrated as a low-level feature point. The point with the highest reflectivity value is selected from the specific band associated with (525-565nm) and calibrated as a high-level feature point. = Connect the two groups of low-level feature points to determine a group of low-level feature lines, then determine the vertical distance L between the high-level feature point and the feature line, and record the determined vertical distance L as the spectral feature of the current partition point at the current moment; Then the different spectral features associated with different partition points are calibrated and confirmed in turn.
6. A remote sensing monitoring method for typical forestry pests and diseases according to claim 5, characterized in that: The (435-495 nm) corresponds to the blue light band, the (525-565 nm) corresponds to the green light band, and the (646-673 nm) corresponds to the red light band.
7. The remote sensing monitoring method for typical forestry pests and diseases according to claim 1, characterized in that: In step 4, the specific method of calibrating the abnormal points is: Based on the real-time detection spectrum generated by different partition points, confirm the real-time associated spectral features, and sort and confirm the real-time confirmed spectral features according to the time sequence of the feature relationship, and confirm the spectral feature sorting sequence associated with the corresponding partition point: In the spectral feature sorting sequence, if the latter set of spectral features is lower than the former set of spectral features, a set of low-variable signals is generated. If the low-variable signals are generated three times in a row, the corresponding partition point is calibrated as the pending point. Otherwise, the pending point is not calibrated. Then confirm the infrared spectrum associated with the pending point at the current moment: lock the high-level feature point from the confirmed infrared spectrum, and confirm the band turning points associated with both sides of the high-level feature point; The band associated with the two groups of band turning points and the high-level feature points is recorded as the pending band. The band characteristics associated with adjacent points in the pending band are confirmed: its band characteristic = |reflectivity value difference between adjacent points| ÷ time interval. Based on the confirmed several band characteristics, the maximum value is selected as the verification feature of the pending point at the current moment; Then, the calibration feature associated with the next pending point is continuously monitored, and the calibration value difference is confirmed. The calibration value difference = |calibration feature at the current moment - calibration feature at the next moment|. If the calibration value difference > Y2, the pending point is marked as an abnormal point, and its Y2 is the preset value.
8. A remote sensing monitoring method for typical forestry pests and diseases according to claim 7, characterized in that: If the calibration value difference is ≤ Y2, no calibration is performed.
9. The remote sensing monitoring method for typical forestry pests and diseases according to claim 7, characterized in that: The step 4 further includes: Identify whether other points associated with the abnormal point in the subsequent process are marked as abnormal points. If so, it means that there is pest impact, and the abnormal point area associated with the pest image will be displayed, and timely response measures will be taken. If not, continue monitoring.