Surface water body classification data quality evaluation method based on remote sensing image

Abstract

The invention discloses a surface water body classification data quality evaluation method based on a remote sensing image, and the method comprises the steps: grouping to-be-classified remote sensing image samples, and distributing different groups to different persons for classification; establishing a square grid; selecting a cross-group grid and a cross-group sample; calculating the percentage of various surface water bodies of each group in the cross-group grid; calculating correlation coefficients of percentages of various surface water bodies among each group in the cross-group grid; and judging whether the groups are abnormal or not according to the correlation coefficient. And based on the similarity principle in the first law of geography, carrying out anomaly detection and quality evaluation on massive classification results by utilizing the classification result correlation coefficients of different classification personnel.

Description

Technical Field

[0001] The present invention relates to the field of remote sensing technology for surface water extraction and classification, and particularly to a method for evaluating the quality of surface water classification data based on remote sensing images. Background Art

[0002] In recent years, significant progress has been made in the field of remote sensing technology for surface water extraction and classification. The mainstream methods include: (1) Threshold method and spectral index method. Threshold segmentation based on spectral features is a basic method. For example, the Normalized Difference Water Index (NDWI) extracts water bodies through the reflectance difference between the green and near-infrared bands, and is applicable to optical images (such as Landsat, Sentinel-2). Some researchers have adopted an improved MNDWI index, introducing the mid-infrared band to reduce the interference of building shadows. SAR data utilizes the characteristic of low backscattering coefficient of water bodies and sets a threshold through the gray histogram to extract water bodies, but shallow water areas are easily affected by signal attenuation. (2) Object-oriented classification method. Combining spatial features such as texture and shape, homogeneous pixels are aggregated into objects through multi-scale segmentation to improve the extraction accuracy of complex regions (such as fragmented wetlands). However, the selection of the segmentation scale depends on experience, and there is also the problem of reduced homogeneity caused by wind and wave disturbances. (3) Multi-source data fusion method. Combining the complementary advantages of optical, SAR, and LiDAR data. For example, Sentinel-1 SAR can penetrate clouds to monitor flood dynamics, and combining it with Sentinel-2 NDWI can improve the recognition ability of turbid water bodies. Generally speaking, the technology for extracting water bodies using remote sensing data is relatively mature.

[0003] Due to different application methods of water resources, in practical applications, on the basis of extracting water bodies, it is necessary to refine the classification of water bodies. For example, in the national land industry, surface water bodies are divided into reservoirs, rivers, lakes, ditches, etc. In terms of the refined classification of water bodies, the current remote sensing classification methods have the following problems: First, spectral similarity and different substances with the same spectrum, such as turbid water bodies (such as rivers containing sediment), eutrophic lakes, and some vegetation-covered areas with overlapping spectral characteristics, which can lead to misclassification. Second, morphological feature ambiguity and engineering interference. Small reservoirs and natural lakes have similar morphologies, and it is difficult to distinguish them when there are no obvious dam body features. In urban areas, channels intersect with linear features such as roads and drainage pipelines, and edge extraction is easily interfered. Mountain shadows and cloud cover cause the water body boundary to be blurred, affecting the accuracy of morphological analysis. Third, data resolution and timeliness limitations. Low-resolution images (such as 30-meter Landsat) cannot capture narrow channels (width < 10 meters) or fine shorelines. Although high-resolution data (such as GF series) can identify details, the coverage period is long and it is difficult to support dynamic monitoring.

[0004] The current focus is on classifying water bodies using various algorithms, which is efficient but prone to misclassification. If combined with high-resolution remote sensing images, visual interpretation can eliminate this misclassification, but due to issues such as the knowledge level and working status of the classification personnel, it is prone to systematic biases. Summary of the Invention

[0005] In view of the above problems, the present invention is proposed to provide a method for evaluating the quality of surface water body classification data based on remote sensing images that overcomes the above problems or at least partially solves the above problems.

[0006] According to one aspect of the present invention, there is provided a method for evaluating the quality of surface water body classification data based on remote sensing images, and the quality evaluation method includes: Grouping the remote sensing image samples to be classified and assigning different groups to different personnel for classification; Establishing a square grid; Selecting cross-group grids and cross-group samples; Calculating the percentage of each type of surface water body in each group within the cross-group grid; Calculating the correlation coefficient of the percentage of each type of surface water body between each group within the cross-group grid; Determining whether there are anomalies between groups according to the correlation coefficient.

[0007] Optionally, the grouping of the remote sensing image samples to be classified and assigning different groups to different personnel for classification specifically includes: Grouping the remote sensing image samples of the surface water body objects to be identified. When grouping, it is necessary to ensure that the samples in each group are geographically adjacent, and ensure that the objects in each group are adjacent to other groups but do not overlap; Assigning each group of samples to different classification personnel, and the classification personnel classify the surface water bodies into 5 categories: reservoir, pond, lake, river, and ditch.

[0008] Optionally, the establishment of the square grid specifically includes: Drawing a square grid according to the size of longitude 1°×latitude 1° and sequentially numbering it; Extracting the central point coordinates of the remote sensing image picture, converting them into a shp point layer using ArcGIS, and assigning the grid number where each sample's central point is located to the attribute field of the sample.

[0009] Optionally, the selection of cross-group grids and cross-group samples specifically includes: According to the grid number, counting the samples contained in each grid. If a grid contains more than two groups of samples, then this grid is considered a cross-group grid; Determine the cross-group samples within each cross-group grid. The determination method is as follows: If there are samples from other groups within a 10-km range of a certain sample, then the sample is considered a cross-group sample.

[0010] Optionally, the specific calculation of the percentage of various types of surface water bodies in each group within the cross-group grid includes: Calculate the proportion of various water body objects in the cross-group samples of different groups within each selected cross-group grid; The formula is as follows: In the formula, i represents the cross-group grid number, j represents the group number, t represents a certain type of surface water body type, and 1, 2, 3, 4, 5 represent reservoir, pond, lake, river, and ditch respectively. represents the number of reservoirs identified in the j-th group within grid i. represents the number of all remote sensing image samples in the j-th group within grid i.

[0011] Optionally, the specific calculation of the correlation coefficient of the percentage of various types of surface water bodies between each group within the cross-group grid includes: Calculate the correlation coefficient of the percentage of the identified land body water body types in different groups within each grid. The formula is as follows: In the formula, i represents the grid number, j represents group 1, and m represents group 2. represents the correlation coefficient between the j-th group and the m-th group in grid i; t represents the surface water body type, and t = 1, 2, 3, 4, 5 represent reservoir, pond, lake, river, and ditch respectively. represents the percentage of the t-type water body within the j-th group in grid i. represents the average value of the percentage of the 5-type water body within the j-th group in grid i. represents the percentage of the t-type water body within the m-th group in grid i. represents the average value of the percentage of the 5-type water body within the m-th group in grid i. If the grid contains more than 3 groups, calculate the correlation coefficient for each pair of groups.

[0012] Optionally, the specific determination of whether there are anomalies between groups based on the correlation coefficient includes: Determine whether there are anomalies in the classification results based on the correlation coefficient between different groups. If the correlation coefficient r between two certain groups is greater than 0.8, it is considered that the similarity of the two groups of samples is very high. If r is between 0.4 and 0.8, it is considered that the similarity is relatively high, and spot checks are required. If r is less than 0.4, it is considered that there is an obvious abnormality between these two groups, and the group of samples with problems is evaluated in combination with the actual situation.

[0013] A method for evaluating the quality of surface water body classification data based on remote sensing images provided by the present invention, the quality evaluation method comprising: grouping the remote sensing image samples to be classified and assigning different groups to different personnel for classification; establishing a square grid; selecting cross-group grids and cross-group samples; calculating the percentage of each type of surface water body in each group within the cross-group grid; calculating the correlation coefficient of the percentage of each type of surface water body between each group within the cross-group grid; and determining whether there is an abnormality between the groups according to the correlation coefficient. Based on the similarity principle in the first law of geography, the correlation coefficient of the classification results of different classifiers is used for anomaly detection and quality evaluation of a large number of classification results.

[0014] The above description is only an overview of the technical solution of the present invention. In order to be able to understand the technical means of the present invention more clearly, it can be implemented according to the content of the description. And in order to make the above and other purposes, features and advantages of the present invention more obvious and understandable, the specific embodiments of the present invention are specifically described below. Brief Description of the Drawings

[0015] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0016] Figure 1 It is a flowchart of a method for evaluating the quality of surface water body classification data based on remote sensing images provided by an embodiment of the present invention; Figure 2 It is a schematic diagram of sample grouping provided by an embodiment of the present invention; Figure 3 It is a schematic diagram of selecting cross-group grids and cross-group samples provided by an embodiment of the present invention; Figure 4 It is a schematic diagram of the correlation coefficient between different groups provided by an embodiment of the present invention; Figure 5 It is a schematic diagram of the sample grouping situation within the grid provided by an embodiment of the present invention; Figure 6 It is a schematic diagram of the reservoir distribution identified within the grid provided by an embodiment of the present invention. Detailed Description of the Embodiments

[0017] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present disclosure can be more thoroughly understood and the scope of the present disclosure can be fully communicated to those skilled in the art.

[0018] In the description of the embodiments of the present invention, the terms "including" and "having" and any variations thereof in the claims and the drawings are intended to cover non-exclusive inclusion. For example, including a series of steps or units.

[0019] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments.

[0020] Embodiment 1 As Figure 1 shown, a method for evaluating the quality of surface water body classification data based on remote sensing images includes: Step 1: Group a large number of remote sensing image samples of surface water body objects to be identified. When grouping, it is necessary to ensure that the samples in each group are geographically adjacent, that is, the objects in each group are adjacent to other groups but do not intersect. Assign each group of samples to different classifiers, and the classifiers classify the surface water bodies into 5 categories: reservoir, pond, lake, river, and ditch.

[0021] Step 2: Establish a square grid with a size of 1° longitude × 1° latitude and number it sequentially. Extract the central point coordinates of the remote sensing image, convert it into a shp point layer using ArcGIS, and assign the grid number where each sample's central point is located to the attribute field of the sample.

[0022] Step 3: Select cross-group grids and cross-group samples. According to the grid number, count the samples included in each grid. If a grid contains samples from more than two groups, then this grid is considered a cross-group grid. Then determine the cross-group samples within each cross-group grid. The determination method is: if there are samples from other groups within 10 km of a certain sample, then this sample is considered a cross-group sample.

[0023] Step 4: Calculate the proportion of various water body objects in the cross-group samples of different groups within each cross-group grid selected in Step 3. The formula is as follows:

[0024] In the formula, i represents the cross-group grid number, j represents the group number, t represents a certain type of surface water body type (1, 2, 3, 4, 5 represent reservoir, pond, lake, river, and ditch respectively), represents the number of reservoirs identified in the j-th group within grid i, Indicates the number of all remote sensing image samples in the j-th group within grid i.

[0025] Step 5: Calculate the correlation coefficient of the percentage of identified land and water body types in different groups within each grid. The formula is as follows:

[0026] In the formula, i represents the grid number, j represents group 1, and m represents group 2; represents the correlation coefficient between group j and group i in grid i; t represents the surface water body type, where t = 1, 2, 3, 4, 5 represent reservoir, pond, lake, river, and ditch respectively, represents the percentage of water body of type t within group j in grid i, represents the average value of the percentage of water body of type 5 within group j in grid i; represents the percentage of water body of type t within group m in grid i, represents the average value of the percentage of water body of type 5 within group m in grid i.

[0027] If there are more than 3 groups within the grid, calculate the correlation coefficient for each pair of groups.

[0028] Step 6: Determine whether there are anomalies in the classification results based on the correlation coefficient between different groups. If the correlation coefficient r between two groups is greater than 0.8, it can be considered that the similarity of the two groups of samples is very high; if r is between 0.4 - 0.8, it can be considered that the similarity is relatively high and spot checks can be carried out; if r is less than 0.4, it is considered that there are obvious anomalies between these two groups, and it is necessary to evaluate which group of samples has problems in combination with the actual situation.

[0029] Example 2 Step 1: Collected remote sensing image samples of surface water bodies within the scope of China, divided them into three groups, and assigned them to different personnel for classification, as Figure 2 shown.

[0030] Step 2: Within the sample annotation range, establish and draw grids with a size of 1° longitude × 1° latitude.

[0031] Step 3: Select 19 cross-group grids covering more than two groups of annotated samples. Using GIS software, select the cross-group samples within each cross-group grid. As Figure 3 shown.

[0032] Step 4: Calculate the percentage of each type of water body in each group within the cross-group grids selected in Step 3.

[0033] Step 5: Calculate the correlation coefficient of the percentage of each type of water body in different groups within each grid. As can be seen from Figure 4 , there are 3 with a correlation coefficient less than 0.4.

[0034] Step 6: Determine the rationality of the labeled samples in different groups according to the correlation coefficient. Take a grid with a correlation coefficient less than 0.4 as an example. Taking a reservoir as an example, by comparing Figure 5 and Figure 6 it can be found that the distribution of reservoirs identified in the two groups of samples (labeled by different labelers) within this grid is significantly different in the two groups of grids. The reservoir density in the first group is significantly higher than that in the second group. It is preliminarily determined that there is a problem with one of the groups of grids.

[0035] Combined with the judgment of the remote sensing image, the geographical environments of the two groups of samples are very similar. Under normal circumstances, the reservoir density should not change suddenly at the junction of the samples. By spot-checking the samples, it is found that many reservoirs in the second group of samples are not identified, indicating a problem.

[0036] Beneficial effects: Based on the similarity principle in the first law of geography, the correlation coefficient of the classification results of different classifiers is used for anomaly detection and quality evaluation of a large number of classification results.

[0037] The above specific implementation manners further elaborate on the purpose, technical solutions, and beneficial effects of the present invention. It should be understood that the above are only specific implementation manners of the present invention and are not used to limit the protection scope of the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for evaluating the quality of surface water body classification data based on remote sensing images, characterized in that, The quality evaluation method includes: Group the remote sensing image samples to be classified and assign different groups to different personnel for classification; Establish a square grid; Select cross-group grids and cross-group samples; Calculate the percentage of each type of surface water body in each group within the cross-group grid; Calculate the correlation coefficient of the percentage of each type of surface water body between each group within the cross-group grid; Determine whether there is an abnormality between groups according to the correlation coefficient.

2. The method for evaluating the quality of surface water body classification data based on remote sensing images according to claim 1, characterized in that, The step of grouping the remote sensing image samples to be classified and assigning different groups to different personnel for classification specifically includes: Group the remote sensing image samples of the surface water body objects to be identified. When grouping, ensure that the samples in each group are geographically adjacent, and ensure that the objects in each group are adjacent to other groups but do not overlap; Assign each group of samples to different classification personnel, and the classification personnel classify the surface water bodies into 5 categories: reservoir, pond, lake, river, and ditch.

3. A method for evaluating the quality of surface water body classification data based on remote sensing images according to claim 1, characterized in that The step of establishing a square grid specifically includes: Establish and draw a square grid according to the size of longitude 1° × latitude 1°, and carry out sequential numbering; Extract the central point coordinates of the remote sensing image picture, convert them into a shp point layer by using ArcGIS, and assign the grid number where each sample central point is located to the attribute field of the sample.

4. A method for evaluating the quality of surface water body classification data based on remote sensing images according to claim 1, characterized in that, The step of selecting cross-group grids and cross-group samples specifically includes: According to the grid number, count the samples contained in each grid. If a grid contains more than two groups of samples, it is considered that the grid is a cross-group grid; Determine the cross-group samples within each cross-group grid. The determination method is: if there are samples of other groups within 10 km of a certain sample, it is considered that the sample is a cross-group sample.

5. A method for evaluating the quality of surface water body classification data based on remote sensing images according to claim 1, characterized in that, The step of calculating the percentage of each type of surface water body in each group within the cross-group grid specifically includes: Calculate the proportion of each type of water body object in the cross-group samples of different groups within each selected cross-group grid; The formula is as follows: Wherein, i represents the cross-group grid number, j represents the group number, t represents a certain type of surface water body type, and 1, 2, 3, 4, and 5 respectively represent reservoirs, ponds, lakes, rivers, and ditches. represents the number of reservoirs identified in the j-th group within grid i. represents the number of all remote sensing image samples in the j-th group within grid i.

6. A method for evaluating the quality of surface water body classification data based on remote sensing images according to claim 1, characterized in that, The step of calculating the correlation coefficient of the percentage of each type of surface water body between each group within the cross-group grid specifically includes: Calculate the correlation coefficient of the percentage of the identified land body water types in different groups within each grid; The formula is as follows: Wherein, i represents the grid number, j represents Group 1, and m represents Group 2; represents the correlation coefficient between Group j and Group i in Grid i; t represents the type of surface water body, where t = 1, 2, 3, 4, 5 represent reservoir, pond, lake, river, and ditch respectively, represents the percentage of water body of type t in Group j within Grid i, represents the average value of the percentage of water body of type 5 in Group j within Grid i; represents the percentage of water body of type t in Group m within Grid i, represents the average value of the percentage of water body of type 5 in Group m within Grid i; If there are more than 3 groups within the grid, calculate the correlation coefficient by grouping two by two.

7. A method for evaluating the quality of surface water body classification data based on remote sensing images according to claim 1, characterized in that, The step of determining whether there is an abnormality between groups according to the correlation coefficient specifically includes: Determine whether there is an abnormality in the classification results according to the correlation coefficient between different groups; If the correlation coefficient r between two groups is greater than 0.8, it is considered that the similarity of the two groups of samples is very high; If r is between 0.4 and 0.8, it is considered that the similarity is relatively high, and spot checks are carried out; If r is less than 0.4, it is considered that there is an obvious abnormality between these two groups, and evaluate which group of samples has problems in combination with the actual situation.

Images