Rubber garden establishment year correction method based on neighborhood analysis

Abstract

The invention discloses a rubber plantation establishment year correction method based on neighborhood analysis, and the method optimizes the estimation of establishment years based on the spatial continuity characteristics of a rubber plantation through a two-stage correction strategy. In the first stage, abnormal pixels are corrected through the difference value and the standard deviation; in the second stage, boundaries and abnormal pixels are further corrected through neighborhood mode frequency and a hole filling strategy. According to the method, the problem of rubber plantation establishment year estimation deviation under the conditions of mixed pixels and complex landscapes is effectively solved, and the precision and reliability of rubber plantation dynamic monitoring are improved.

Description

Technical Field

[0001] The invention relates to the technical field of remote sensing image processing, and in particular to a method for correcting the year of rubber plantation establishment based on neighborhood analysis. Background Art

[0002] When using remote sensing technology to estimate the establishment year of rubber plantations, existing technologies face several key challenges. The primary issue is the accurate identification of plantation boundary areas. Due to the spatial resolution limitations of remote sensing images, mixed pixel phenomena often occur in the edge areas of plantations, that is, a single pixel may contain spectral information of plantations and other types of landforms at the same time. This mixed pixel effect not only affects the accurate demarcation of boundaries, but also leads to incorrect estimates of the establishment year. In addition, different land use types in neighboring areas will interfere with the spectral characteristics of boundary pixels, and the shadow effect caused by terrain undulations further increases the difficulty of identifying boundary areas.

[0003] Maintaining temporal continuity is another important challenge facing existing technologies. In theory, the establishment year within the same rubber plantation should show strong spatial continuity, but in actual processing, discontinuous year estimates often occur. This discontinuity is partly due to the uneven quality of historical remote sensing images. Images from some historical periods may have quality problems such as cloud cover and strip noise, which increases the uncertainty of time series analysis results. At the same time, the seasonal variation characteristics of rubber will also interfere with the judgment of the establishment year, especially in the transition period between the growing season and the non-growing season, where this impact is more significant.

[0004] Existing correction methods have obvious limitations in dealing with the above problems. Commonly used simple statistical methods (such as mean filtering, median filtering, etc.) can reduce the impact of outliers to a certain extent, but they are often unable to effectively handle complex boundary conditions and abnormal areas. These methods usually only consider the statistical characteristics of pixel values, while ignoring the importance of spatial continuity, resulting in "patchy" correction results. More importantly, when making corrections, these methods may over-smooth and destroy the true temporal variation characteristics of the years when rubber plantations were established. Especially in areas where there are indeed differences in years, incorrect corrections will reduce the reliability of the results.

[0005] Therefore, the present invention provides a method that can solve the deviation problem of estimating the establishment year of rubber plantations under mixed pixels and complex landscape conditions, and improve the accuracy and reliability of dynamic monitoring of rubber plantations. Summary of the invention

[0006] In view of this, the present invention provides a method for correcting the year of rubber plantation establishment based on neighborhood analysis.

[0007] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0008] A method for correcting the year of rubber plantation establishment based on neighborhood analysis, the method comprising a first-stage correction and a second-stage correction.

[0009] The first stage of calibration includes the following steps:

[0010] Step 1: Determine the neighborhood:

[0011] Select a target pixel in the rubber plantation, whose pixel value index is EY (i,j) ;

[0012] Get the 8-neighborhood of the target pixel, that is, 8 adjacent pixels; the index coordinates are: (i-1, j-1), (i, j-1), (i+1, j-1); (i-1, j), (i+1, j); (i-1, j+1), (i, j+1), (i+1, j+1);

[0013] Step 2: Collect neighborhood values:

[0014] For each neighborhood index, check whether it is within the matrix boundary. If the index value exceeds the row and column range of the matrix, skip the index;

[0015] Extract the neighboring pixel values ​​that meet the conditions and put them into a list;

[0016] Step 3, calculate the mode and standard deviation of non-masked neighborhood pixels;

[0017] Step 4, calculate the absolute difference between the central pixel and the mode;

[0018] Step 5, setting the difference threshold T1 and the standard deviation threshold T2;

[0019] Step 6: Calibration test to see if |EY is satisfied at the same time (i,j) -EY mode_8NB |≤T1 and EY STDDEV_8NB ≤T2;

[0020] Step 7: Satisfy both |EY (i,j) -EY mode_8NB |≤T1 and EY STDDEV_8NB The establishment year of the central pixel ≤T2 is updated to the neighborhood mode value;

[0021] Step 8: This process is performed twice to ensure sufficient correction effect;

[0022] The second stage calibration includes the following steps:

[0023] Step a, re-obtaining the 8-neighborhood pixel values ​​of each plantation pixel;

[0024] Step b, calculating the frequency of occurrence of the mode in the 8 neighborhood pixels; when the frequency of occurrence of the mode exceeds 50%, updating the establishment year of the central pixel to the neighborhood mode value;

[0025] Step c: For isolated non-rubber plantation pixels in the rubber plantation area, when more than N pixels in its 8-neighborhood are classified as rubber plantations, the pixel is filled with rubber plantations and assigned a corresponding establishment year value.

[0026] Preferably, in step 3, the method for calculating the mode is: counting the frequency of occurrence of each value in the list, finding the value with the highest frequency of occurrence, which is the mode of the neighborhood; if there are multiple values ​​with the same frequency of occurrence, select any one of them as the mode; the formula for calculating the standard deviation is:

[0027]

[0028] Where: N is the number of unmasked pixels in the neighborhood, with a maximum value of 8; X K is the value of each pixel; μ is the mean of the neighborhood values.

[0029] Preferably, in step 4, the method for calculating the absolute difference between the central pixel and the mode is: subtract the neighborhood mode from the central pixel value and take the absolute value. The formula is as follows:

[0030] Absolute Difference=│EY (i,j) -EY mode_8NB │

[0031] Among them, EY (i,j) is the center pixel at row and column position i, j, EY Mode_8NB It is the majority of the pixels in the neighborhood of this pixel.

[0032] Preferably, in step 6, correction is performed when the absolute difference between the central pixel and the mode is less than a threshold value T1, and the neighborhood standard deviation is less than a threshold value T2.

[0033] Preferably, in step b, the method for calculating the frequency of occurrence of the mode is Where N is the number of unmasked pixels, and the maximum value is 8.

[0034] Preferably, the window size used in the first stage correction and the second stage correction is 3×3 pixels.

[0035] Preferably, the difference threshold T1 and the standard deviation threshold T2 of the first stage correction are both set to 2.

[0036] Preferably, the mode frequency threshold of the second stage correction is set to 50%.

[0037] Compared with the prior art, the present invention has achieved the following technical effects:

[0038] The method of the present invention effectively solves the problem of deviation in estimating the establishment year of rubber plantations under mixed pixel and complex landscape conditions, and improves the accuracy and reliability of dynamic monitoring of rubber plantations. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] Figure 1 A rubber plantation establishment year correction map of neighborhood information of a rubber plantation establishment year correction method based on neighborhood analysis of the present invention;

[0040] Figure 2 Schematic diagram of plantation establishment year estimation based on neighborhood information of a rubber plantation establishment year correction method based on neighborhood analysis of the present invention; a) original output of rubber plantation establishment year, b) result after correction based on numerical difference and standard deviation, c) final result after correction using mode percentage (>50%);

[0041] Figure 3 It is a precision assessment diagram (scatter plot) of the observed planting years of random sampling points of a rubber plantation establishment year correction method based on neighborhood analysis of the present invention and the planting years estimated by the present invention. DETAILED DESCRIPTION

[0042] The technical scheme in the embodiment of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiment of the present invention. Obviously, the described embodiment is only a part of the embodiment of the present invention, not all of the embodiments. Based on the embodiment of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0043] The invention discloses a method for correcting the establishment year of a rubber plantation based on neighborhood analysis, the method comprising a first-stage correction and a second-stage correction, the window size used in the first-stage correction and the second-stage correction is 3×3 pixels;

[0044] The first stage of calibration includes the following steps:

[0045] Step 1: Determine the neighborhood:

[0046] Select a target pixel and represent it by its row and column index, denoted as (i, j);

[0047] Get the 8-neighborhood of the target pixel, that is, 8 adjacent pixels;

[0048] The neighborhood indexes are: (i-1,j-1), (i,j-1), (i+1,j-1); (i-1,j), (i+1,j); (i-1,j+1), (i,j+1), (i+1,j+1);

[0049] Step 2: Collect neighborhood values:

[0050] For each neighborhood index, check whether it is within the matrix boundary. If it exceeds the row and column range of the matrix, skip the index;

[0051] Extract the neighboring pixel values ​​that meet the conditions and put them into a list;

[0052] Step 3, calculate the mode and standard deviation of the neighborhood pixels;

[0053] The method for calculating the mode is: count the frequency of occurrence of each value in the list, find the value with the highest frequency, which is the mode of the neighborhood; if there are multiple values ​​with the same frequency of occurrence, choose any one of them as the mode;

[0054] The formula for calculating standard deviation is:

[0055]

[0056] Where: N is the number of unmasked pixels in the neighborhood, with a maximum value of 8; X K is the value of each pixel; μ is the mean of the neighborhood values.

[0057] Step 4, calculate the absolute difference between the central pixel and the mode;

[0058] The method to calculate the absolute difference between the central pixel value and the majority is to subtract the neighborhood majority from the central pixel value and take the absolute value. The formula is as follows:

[0059] AbsoluteDifference=│EY (i,j) -EY mode_8NB │

[0060] Among them: EY (i,j) is the center pixel at row and column position i, j, EY Mode_8NB is the majority of the pixels in the neighborhood of this pixel;

[0061] Step 5, setting the difference threshold T1 and the standard deviation threshold T2;

[0062] Step 6: Perform calibration;

[0063] When the absolute difference between the central pixel and the mode is less than the threshold value T1, and the neighborhood standard deviation is less than the threshold value T2, correction is performed; the difference threshold value T1 and the standard deviation threshold value T2 of the first stage correction are both set to 2;

[0064] Step 7, updating the establishment year of the central pixel that meets the conditions to the neighborhood mode value;

[0065] Step 8: This process is performed twice to ensure sufficient correction effect;

[0066] The second stage calibration includes the following steps:

[0067] Step a, re-obtaining the 8-neighborhood pixel values ​​of each plantation pixel;

[0068] Step b, calculate the frequency of occurrence of the majority in the 8-neighborhood pixels; when the frequency of occurrence of the majority exceeds 50%, (such as Figure 1 Correction of the rubber plantation establishment year (EY) using neighborhood information (shown in 18 cases, M / N represents the majority percentage, N / A represents the masked non-rubber pixel), and the establishment year of the central pixel is updated to the neighborhood majority value; the mode frequency threshold of the second stage correction is set to 50%;

[0069] The frequency of occurrence of the mode is calculated as Where N is the number of unmasked pixels, and the maximum value is 8.

[0070] Step c: For isolated non-rubber plantation pixels in the rubber plantation area, when more than N pixels in its 8-neighborhood are classified as rubber plantations, the pixel is filled with rubber plantations and assigned a corresponding establishment year value.

[0071] Among them, considering the existence of masks in surrounding pixels, there are 18 situations, such as Figure 1 shown.

[0072] Embodiment 1:

[0073] This embodiment takes Vietnam rubber plantation as an example. Figure 2 As shown in FIG. 1 , the estimation result of the establishment year of the rubber plantation is gradually improved by the neighborhood correction method proposed in the present invention. Figure 2 Figure a shows the initial output of the remote sensing monitoring algorithm for the establishment time of rubber plantations, where different colors represent different establishment years. It can be seen that there is obvious variability in the year estimation within a single plantation block, which is inconsistent with the actual situation that plantations are usually planted at the same time.

[0074] After applying the first stage correction, if Figure 2 As shown in (b), the correction method based on numerical difference and standard deviation significantly reduces the year variability within the plantation, and the block presents a more uniform color distribution. However, there are still some abnormal situations where different establishment years intersect in the edge area of ​​the plantation.

[0075] After the second stage correction Figure 2c shows that the estimation results are further optimized after the mode percentage correction method is adopted. The plantation blocks show more coherent and clear characteristics, and the boundaries between areas with different establishment years are clearer. This step-by-step optimization process effectively reduces the "salt and pepper" effect that appears in the initial estimation, making the results more consistent with the actual plantation establishment pattern.

[0076] It can be seen that this embodiment fully demonstrates the effectiveness of the present invention in practical applications. It can not only correct the year estimation deviation within the plantation, but also effectively deal with the mixed pixel problem in the boundary area, and finally obtain a more accurate and reliable year estimation result.

[0077] The present invention generates 100 random sample points on the classified rubber distribution map through the "ee.Image.sample" method in the GEE platform and obtains the estimated value. The observed value comes from the historical image of Google Earth to intuitively explain the establishment year of the rubber plantation. There is a strong correlation between the estimated establishment year of the Vietnamese rubber plantation and the observed establishment year of the plantation, such as Figure 3 As shown. Most of the points are tightly clustered on the 1:1 line, indicating good agreement between the estimated and observed values. The linear fit (red line) almost overlaps with the 1:1 line (grey dashed line), further demonstrating the high accuracy of the estimates. Excluding the 5 points (5.1%) that are significantly different, the linear fit of the remaining points shows an R2 of 0.99 and an RMSE of 0.25 years. The estimates seem to be particularly accurate for the years between 1990 and 2015, with most of the data points concentrated in these years. The scatter is slightly larger for the earliest years (before 1990) and the most recent years (after 2015), probably because there are fewer samples in these ranges.

[0078] The above description is only a preferred embodiment of the present invention and does not limit the technical scope of the present invention. Therefore, any slight modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A method for correcting the year of rubber plantation establishment based on neighborhood analysis, characterized in that: The correction method includes a first-stage correction and a second-stage correction. The first stage of calibration includes the following steps: Step 1: Select a target pixel in the rubber plantation, whose pixel value is EY (i,j) ; Step 2: Take out the 8 neighboring pixel values ​​of the target pixel, and the index coordinates are: (i-1, j-1), (i, j-1), (i+1, j-1); (i-1, j), (i+1, j); (i-1, j+1), (i, j+1), (i+1, j+1); discard the masked pixels and put them into a list; Step 3, calculate the mode and standard deviation of non-masked neighborhood pixels; Step 4, calculate the absolute difference between the target pixel and the mode; Step 5, setting the difference threshold T1 and the standard deviation threshold T2; Step 6: Calibration test to test whether |EY is satisfied at the same time (i,j) -EY mode_8NB |≤T1 and EY STDDEV_8NB ≤T2; Step 7: Update the target pixel value that meets the conditions to the neighborhood mode value; Step 8: This process is performed twice to ensure sufficient correction effect; The second stage calibration includes the following steps: Step a, re-obtain the 8-neighborhood pixel values ​​of each plantation pixel and discard the masked pixels; Step b, calculating the frequency of occurrence of the mode in the 8 neighborhood pixels; when the frequency of occurrence of the mode exceeds 50%, updating the central pixel value to the neighborhood mode value; Step c: For isolated non-rubber plantation pixels in the rubber plantation area, when more than N pixels in its 8-neighborhood are classified as rubber plantations, the pixel is filled with rubber plantations and assigned a corresponding establishment year value.

2. A method for correcting the establishment year of a rubber plantation based on neighborhood analysis according to claim 1, characterized in that: In step 3, the method for calculating the mode is: count the frequency of occurrence of each value in the list, find the value with the highest frequency of occurrence, which is the mode of the neighborhood; if there are multiple values ​​with the same frequency of occurrence, select any one of them as the mode; the formula for calculating the standard deviation is: Where: N is the number of unmasked pixels in the neighborhood, with a maximum value of 8; X K is the value of each pixel; μ is the mean of the neighborhood values.

3. A method for correcting the establishment year of a rubber plantation based on neighborhood analysis according to claim 1, characterized in that: In step 4, the method for calculating the absolute difference between the central pixel and the mode is: subtract the neighborhood mode from the central pixel value and take the absolute value. The formula is as follows: AbsoluteDifference=│EY (i,j) -EY Mode_8NB │ Among them, EY (i,j) is the center pixel at row and column position i, j, EY Mode_8NB It is the majority of the pixels in the neighborhood of this pixel.

4. A method for correcting the establishment year of a rubber plantation based on neighborhood analysis according to claim 1, characterized in that: In step 6, correction is performed when the absolute difference between the central pixel and the mode is less than the threshold value T1, and the neighborhood standard deviation is less than the threshold value T2. That is, the test is whether |EY (i,j) -EY mode_8NB |≤T1 and EY STDDEV_8NB ≤T2.

5. A method for correcting the establishment year of a rubber plantation based on neighborhood analysis according to claim 1, characterized in that: In step b, the frequency of occurrence of the mode is calculated as Where N is the number of unmasked pixels, and the maximum value is 8.

6. A method for correcting the establishment year of a rubber plantation based on neighborhood analysis according to claim 1, characterized in that: The window size used in the first stage correction and the second stage correction is 3×3 pixels.

7. A method for correcting the establishment year of a rubber plantation based on neighborhood analysis according to claim 1, characterized in that: The difference threshold T1 and the standard deviation threshold T2 of the first stage correction are both set to 2.

8. The method for correcting the rubber plantation establishment year based on neighborhood analysis according to claim 1, characterized in that: The mode frequency threshold for the second stage correction was set to 50%.

Images