Coastline characteristic change analysis method based on remote sensing image

Abstract

The invention discloses a coastline characteristic change analysis method based on a remote sensing image, and relates to the technical field of coastline monitoring. Comprising the following steps: collecting Landsat TM / OLI data, preprocessing the data, carrying out sea-land classification and water body index extraction on a remote sensing image of a research area by adopting an object-oriented nearest neighbor algorithm, carrying out post-processing such as vector data conversion and shoreline smoothing processing on final shoreline information to obtain a shoreline data set, and carrying out spatial-temporal change analysis. And by combining a plurality of remote sensing image processing technologies and a statistical analysis method, the coastline feature change can be accurately extracted and analyzed, and the influence of human factors and natural environment in a traditional method is avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of coastline monitoring, and in particular to a method for analyzing coastline feature changes based on remote sensing images. Background Art

[0002] The coastline, the boundary where land meets water, is a crucial component of the coastal environment. As the most fundamental and important geographical feature of the Earth's surface, the coastline often undergoes significant changes in its characteristics due to processes such as sea level fluctuations, tidal fluctuations, waves, and storm surges. Coastline changes carry significant warning signs for shoreline erosion, coastal environmental changes, and the rise and fall of ecosystems. Since the 20th century, coastal countries have gradually shifted the focus of economic development to coastal areas. This shift in economic center has had a significant impact on the economic, social, and ecological environment of coastal areas. For example, large-scale land reclamation projects, excessive coastline exploitation, artificial correction of naturally curved coastlines, and unregulated aquaculture have led to the continuous deterioration of the coastal ecological environment. Therefore, rapid and accurate monitoring of coastline changes is crucial for coastal resource development and utilization, coastal management, maintaining coastal ecological balance, and combating marine disasters.

[0003] When monitoring coastline changes over long periods of time, traditional measurement methods often suffer from numerous limitations, including long cycle times, low measurement efficiency, and high costs, and are susceptible to objective environmental factors. However, remote sensing technology can improve efficiency while enabling large-scale operations and reducing labor costs. Visual interpretation and automatic extraction are the two main approaches for extracting coastlines from remote sensing images. Visual interpretation is susceptible to subjective factors and requires advanced coastline identification skills. Furthermore, the extraction process is time-consuming and difficult to achieve efficient processing. Automatic extraction, however, offers greater timeliness and universality, and has become a crucial tool for extracting coastline information.

[0004] However, due to the influence of human activities and natural factors on the coast, some coasts have large amounts of suspended sediment and accumulation of large amounts of silt and mud. As a result, current remote sensing image analysis methods still face many challenges in dealing with coastline changes under complex terrain and changeable climatic conditions. Summary of the Invention

[0005] The purpose of the present invention is to provide a method for analyzing changes in coastline characteristics based on remote sensing images, which solves the above-mentioned technical problems.

[0006] To achieve the above object, the present invention provides the following technical solution: a method for analyzing coastline feature changes based on remote sensing images, comprising the following steps:

[0007] S1. Collect Landsat TM / OLI data and obtain Landsat TM / OLI remote sensing image data covering the target study area;

[0008] S2. Data preprocessing: Screening out low-resolution remote sensing images from Landsat TM / OLI data. Specifically, this includes removing remote sensing images with large amounts of sea ice and those with cloud cover above 5%, to obtain remote sensing images of the study area that meet analysis requirements.

[0009] S3. Use the object-oriented nearest neighbor algorithm to classify the land and sea in the remote sensing images of the study area to obtain a land and sea classification binary map. At the same time, use Canny edge detection to extract the coastline information in the binary image to obtain the land and sea boundary line.

[0010] S4. Extract water bodies using water body index. Use five water body indices to extract coastline information from remote sensing images of the study area. Compare and analyze the five water body extraction results by establishing a confusion matrix, and select the water body index extraction result with the highest accuracy.

[0011] S5, combining S3 and S4 to obtain the final coastline information, and performing post-processing such as converting the final coastline information into vector data and performing coastline smoothing to obtain a coastline dataset;

[0012] S6, spatiotemporal change analysis, including net shoreline movement, endpoint change rate, and linear regression change rate;

[0013] S7. Get the result.

[0014] Preferably, the data preprocessing in S2 also includes radiation calibration, atmospheric correction, image stitching and cropping processing, so that the remote sensing image data of the study area is in the GCS_WGS_1984 geographic coordinate system, and the remote sensing image of the study area that meets the analysis requirements is obtained.

[0015] Preferably, the S3 specifically adopts a multi-scale segmentation algorithm to segment objects based on spectral and shape homogeneity, which can better segment homogeneous areas, has good segmentation effect, and the segmentation results are relatively smooth. The classification method is the nearest neighbor algorithm, and the selected sample system is executed using the nearest neighbor configuration. The classification is performed using a classifier, and the classifier used is classification. The classification results are subjected to Canny operator edge detection. The object-oriented nearest neighbor algorithm classification is more advantageous for extracting waters with high sediment content.

[0016] Preferably, in S4, the OA values, Kappa coefficients, production accuracies, usage accuracies, misclassification errors, and missed classification errors of each water body index are calculated respectively. Among them, the ANDWI water body index has the highest overall Kappa coefficient, reaching an average of more than 0.90. The misclassification errors from low to high are ANDWI < MNDWI < MBEI < MAWEI < NDWI; the missed classification errors from low to high are ANDWI < MAWEI < NDWI < MBWI < MNDWI, and the data results of the extracted water bodies of ANDWI are obtained.

[0017] Preferably, in S5, the water body extraction result of ANDWI is combined with the object-oriented nearest neighbor classification method. This can make the object-oriented nearest neighbor classification method compensate for the deficiency of the ANDWI water body extraction result in the area with more sediment-laden water bodies, and also make the ANDWI water body extraction method compensate for the deficiency of the slow process of the object-oriented nearest neighbor algorithm classification method. The combination of steps S3 and S4 greatly improves the accuracy of land-water classification and the efficiency of ocean water body extraction.

[0018] Preferably, in S6, the net coastline movement refers to the distance between the earliest coastline and the latest coastline within the research time range. The end point change rate refers to the ratio of the distance of coastline movement to the difference between the earliest and latest years. The calculation formula is as follows:

[0019]

[0020] In the formula: EPM is the end point change rate; NSM is the net coastline movement; Δy is the difference between the earliest and latest years;

[0021] The linear regression change rate is determined by fitting the profile line passing through all the sample points of the coastline by the least squares method to obtain the coastline change rate. The linear regression calculation method is used to calculate using all available data without considering the change of the trend and accuracy of the calculation data. The calculation formula is as follows:

[0022] y = a + bx

[0023]

[0024] In the formula, x represents the year as the independent variable, and y represents the position of the coastline in space; x i represents the i-th year, and y i represents the distance from the intersection point of the i-th year's profile and the coastline to the baseline; and respectively represent the averages of x i and y i ; a represents the intercept of the fitting constant; b represents the linear regression change rate, that is, LRR.

[0025] Preferably, in S7, conclusions and laws of changes in coastline characteristics are drawn based on the results of the spatiotemporal change analysis, providing a scientific basis for marine environmental protection, coastal zone planning and management, etc.

[0026] Compared with related technologies, the coastline feature change analysis method based on remote sensing images provided by the present invention has the following beneficial effects:

[0027] 1. Improved accuracy and reliability of coastline feature changes: By combining multiple remote sensing image processing technologies and statistical analysis methods, it is possible to accurately extract and analyze coastline feature changes, avoiding the influence of human factors and natural environment in traditional methods.

[0028] 2. Improved analysis efficiency: The use of automated and intelligent processing methods can quickly process large amounts of remote sensing image data, thereby improving analysis efficiency.

[0029] 3. Suitable for complex terrain and changeable climate conditions: By optimizing algorithms and parameter settings, it can adapt to the needs of analyzing changes in coastline characteristics under different terrain and climate conditions. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 This is a flowchart of the coastline extraction technology for the study area of ​​the present invention. DETAILED DESCRIPTION

[0031] See also Figure 1 , the present invention provides a technical solution, comprising the following steps:

[0032] S1. Collect Landsat TM / OLI data and obtain Landsat TM / OLI remote sensing image data covering the target study area;

[0033] S2. Data preprocessing: Screening out low-resolution remote sensing images from Landsat TM / OLI data. Specifically, this includes removing remote sensing images with large amounts of sea ice and those with cloud cover above 5%, to obtain remote sensing images of the study area that meet analysis requirements.

[0034] S3. Use the object-oriented nearest neighbor algorithm to classify the land and sea in the remote sensing images of the study area to obtain a land and sea classification binary map. At the same time, use Canny edge detection to extract the coastline information in the binary image to obtain the land and sea boundary line.

[0035] S4. Extract water bodies using water body index. Use five water body indices to extract coastline information from remote sensing images of the study area. Compare and analyze the five water body extraction results by establishing a confusion matrix, and select the water body index extraction result with the highest accuracy.

[0036] S5. Combine S3 and S4 to obtain the final coastline information. After post-processing such as converting the final coastline information into vector data and shoreline smoothing, a coastline dataset is obtained, including but not limited to using software such as ArcGIS and QGIS for processing;

[0037] S6. Temporal and spatial change analysis, including net coastline movement, terminus change rate, and linear regression change rate;

[0038] S7. Obtain the results.

[0039] The data preprocessing in S2 also includes radiometric calibration, atmospheric correction, image mosaicking and cropping processing, so that the remote sensing image data of the study area is in the GCS_WGS_1984 geographic coordinate system, and a remote sensing image of the study area that meets the analysis requirements is obtained.

[0040] In S3, the multi-scale segmentation algorithm is specifically adopted to segment objects based on spectral and shape homogeneity, which can better segment homogeneous regions, with good segmentation effect and relatively smooth segmentation results. The classification method is the nearest neighbor algorithm. The selected sample system is executed using the nearest neighbor configuration, and then classified using the classifier. The classifier used is classification. The classification results are subjected to Canny operator edge detection. The object-oriented nearest neighbor algorithm classification has more advantages in extracting waters with more sediment content.

[0041] In S4, the OA value, Kappa coefficient, production accuracy, use accuracy, misclassification error, and omission error of each water body index are calculated respectively. Among them, the Kappa coefficient of the ANDWI water body index is the highest overall, reaching above 0.90 on average. The misclassification error from low to high is ANDWI < MNDWI < MBEI < MAWEI < NDWI; the omission error from low to high is ANDWI < MAWEI < NDWI < MBWI < MNDWI. The data results of the water body extracted by ANDWI are obtained, and the fragmented patches in the land-sea classification results are removed, and the irregular or abnormal regions are removed to improve the accuracy of coastline extraction.

[0042] As shown in Table 1, the calculation formula of the water body index is

[0043]

[0044] Table 1 is as follows. The evaluation of the water body extraction accuracy of five water body indices in different years is

[0045] Water Index OA Kappa Misclassification Missing points NDWI 95.25％ 0.90 5.33％ 8.86％ MNDWI 95.78％ 0.92 0.83％ 7％ ANDWI 96.84％ 0.94 0.81％ 3.31％ MAWEI 96.73％ 0.93 1.01％ 5.06％ MBEI 94.72％ 0.89 0.21％ 9.53％

[0046] Year a

[0047] <0> Water Index OA Kappa Misclassification Missing points NDWI 93.70％ 0.87 5.21％ 6.83％ MNDWI 94.30％ 0.89 1.24％ 9.68％ ANDWI 95.20％ 0.91 1.22％ 3.32％ MAWEI 92.40％ 0.85 8.93％ 5.12％ MBEI 95.10％ 0.90 6.20％ 8.73％

[0048] bYear

[0049] Water Index OA Kappa Misclassification Missing points NDWI 95.90％ 0.92 2.97％ 5.04％ MNDWI 94.30％ 0.87 1.84％ 9.69％ ANDWI 96.30％ 0.95 1.26％ 3.10％ MAWEI 96.20％ 0.92 2.01％ 5.43％ MBEI 95.50％ 0.91 1.04％ 7.75％

[0050] c year

[0051] Water Index OA Kappa Misclassification Missing points NDWI 97.00％ 0.94 0.59％ 5.09％ MNDWI 96.30％ 0.93 0.40％ 6.60％ ANDWI 97.30％ 0.95 0.19％ 3.01％ MAWEI 96.40％ 0.93 0.30％ 3.77％ MBEI 96.70％ 0.93 0.20％ 6.04％

[0052] d year

[0053] Water Index OA Kappa Misclassification Missing points NDWI 93.10％ 0.86 5.89％ 7.63％ MNDWI 94.40％ 0.89 2.11％ 9.00％ ANDWI 94.70％ 0.90 2.00％ 3.25％ MAWEI 93.80％ 0.88 7.07％ 4.89％ MBEI 94.10％ 0.88 4.06％ 7.44％

[0054] eYear

[0055] Water Index OA Kappa Misclassification Missing points NDWI 96.90％ 0.93 0.75％ 4.83％ MNDWI 96.40％ 0.92 0.63％ 6.08％ ANDWI 96.60％ 0.93 0.56％ 3.21％ MAWEI 96.10％ 0.92 0.76％ 7.00％ MBEI 96.80％ 0.93 0.56％ 5.19％

[0056] f year

[0057] Water Index OA Kappa Misclassification Missing points NDWI 96.10％ 0.92 2.15％ 4.27％ MNDWI 95.00％ 0.91 0.38％ 7.34％ ANDWI 96.00％ 0.94 0.37％ 3.56％ MAWEI 96.70％ 0.93 0.55％ 5.24％ MBEI 96.00％ 0.92 0.56％ 6.47％

[0058] Year

[0059] The water body extraction results of ANDWI are combined with the object-oriented nearest neighbor classification method in S5. In this way, the object-oriented nearest neighbor classification method can make up for the shortcomings of the ANDWI water body extraction results in extracting water areas with more sediment, and also make up for the shortcomings of the slow process of the object-oriented nearest neighbor algorithm classification method. The combination of steps S3 and S4 greatly improves the accuracy of water and land classification and the efficiency of ocean water extraction.

[0060] The net coastline movement in S6 refers to the distance between the earliest coastline and the latest coastline within the study time range, and the terminal change rate refers to the distance of coastline movement compared with the difference between the earliest and latest years. The calculation formula is as follows:

[0061]

[0062] Where: EPM is the rate of change of the terminal point; NSM is the net shoreline movement; Δy is the difference between the earliest and latest years;

[0063] The linear regression change rate is determined by fitting the profile line through all sample points of the coastline using the least squares method. The coastline change rate is obtained by using the linear regression calculation method and all available data for calculation without considering the change in the trend and accuracy of the calculated data. The calculation formula is as follows:

[0064] y=a+bx

[0065]

[0066] Where x represents the year as the independent variable, y represents the spatial location of the coastline; i Indicates the i-th year, yi represents the distance from the intersection of the profile and the coastline in year i to the baseline; and Represents x i and y i The average value of ; a represents the intercept of the fitting constant; b represents the linear regression rate of change, i.e. LRR.

[0067] The Digital Coast Analysis System (DSAS) was used to analyze the spatiotemporal change rate of the coastline. The operation was performed in ArcGIS software. A baseline for auxiliary calculation was drawn based on the extracted coastline information. DSAS generated profile lines of fixed distance based on the drawn baseline. Finally, various coastline rates were calculated to obtain the length change and change rate of the study area. According to the remote sensing image data, it can be concluded that the marine area of ​​the study area was increasing in the early stage and decreased significantly in the later stage. Overall, the study area showed a decreasing trend. The main reasons affecting the above changes are the transformation of natural coasts into artificial coasts, the transformation of natural coasts into fishery coasts, and the accumulation of sediment at the estuary due to sediment transport from the estuary.

[0068] In S7, conclusions and laws of changes in coastline characteristics are drawn based on the results of spatiotemporal change analysis, which can effectively provide a scientific basis for marine environmental protection, coastal zone planning and management, and provide data support for reasonable planning, green management, and sustainable development of the study area.

Claims

1. A method for analyzing coastline feature changes based on remote sensing images, characterized in that: It includes the following steps: S1. Collect Landsat TM / OLI data to obtain Landsat TM / OLI remote sensing image data covering the target study area; S2. Data preprocessing, screening remote sensing images with low clarity in Landsat TM / OLI data, specifically including removing remote sensing images with a large amount of sea ice and removing remote sensing images with cloud cover higher than 5%, to obtain remote sensing images of the study area meeting the analysis requirements; S3. Use the object-oriented nearest neighbor algorithm to classify the land and sea of the remote sensing images of the study area to obtain a binary land-sea classification map. At the same time, use Canny edge detection to extract the coastline information in the binary image to obtain the land-sea boundary line; S4. Extract water bodies using water body indices. Use five water body indices to extract coastline information from the remote sensing images of the study area. Compare and analyze the extraction results of the five water bodies by establishing a confusion matrix, and select the extraction result of the water body index with the highest accuracy; S5. Combine S3 and S4 to obtain the final coastline information, and perform post-processing such as converting vector data and shoreline smoothing on the final coastline information to obtain a coastline data set; S6. Temporal and spatial change analysis, including net coastline movement, end point change rate, and linear regression change rate; S7. Obtain the results.

2. The method for analyzing coastline feature changes based on remote sensing images according to claim 1, characterized in that: The data preprocessing in S2 also includes radiometric calibration, atmospheric correction, image mosaicking and cropping processing, so that the remote sensing image data of the study area is in the GCS_WGS_1984 geographic coordinate system.

3. The method for analyzing coastline feature changes based on remote sensing images according to claim 2, characterized in that: In S3, specifically use the multi-scale segmentation algorithm to segment objects based on spectral and shape homogeneity, which can better segment homogeneous regions, with good segmentation effect and relatively smooth segmentation results. The classification method is the nearest neighbor algorithm. Use the nearest neighbor configuration to execute the selected sample system, and then execute classification using the classifier, and the classifier used is classification. Perform Canny operator edge detection on the classification results. The object-oriented nearest neighbor algorithm classification has more advantages in extracting water areas with more sediment content.

4. The method for analyzing coastline feature changes based on remote sensing images according to claim 3, characterized in that: In S4, calculate the OA value, Kappa coefficient, production accuracy, use accuracy, misclassification error, and omission error of each water body index respectively. Among them, the Kappa coefficient of the ANDWI water body index is the highest overall, reaching above 0.90 on average. The misclassification error from low to high is ANDWI < MNDWI < MBEI < MAWEI < NDWI; the omission error from low to high is ANDWI < MAWEI < NDWI < MBWI < MNDWI, and obtain the data result of water body extraction of ANDWI.

5. The method for analyzing coastline feature changes based on remote sensing images according to claim 4, characterized in that: In S5, use the water body extraction result of ANDWI combined with the object-oriented nearest neighbor classification method, which can make the object-oriented nearest neighbor classification method compensate for the deficiency of the ANDWI water body extraction result in extracting water areas with more sediment, and also make the ANDWI water body extraction method compensate for the deficiency of the slow process of the object-oriented nearest neighbor algorithm classification method. The combination of steps S3 and S4 greatly improves the accuracy of land-sea classification and the efficiency of ocean water body extraction.

6. The method for analyzing coastline feature changes based on remote sensing images according to claim 5, characterized in that: The net coastline movement in S6 refers to the distance between the earliest coastline and the latest coastline within the study time range, and the terminal change rate refers to the distance of coastline movement compared with the difference between the earliest and latest years. The calculation formula is as follows: Where: EPM is the rate of change of the terminal point; NSM is the net shoreline movement; Δy is the difference between the earliest and latest years; The linear regression change rate is determined by fitting the profile line through all sample points of the coastline using the least squares method. The coastline change rate is obtained by using the linear regression calculation method and all available data for calculation without considering the change in the trend and accuracy of the calculated data. The calculation formula is as follows: y=a+bx Where x represents the year as the independent variable, y represents the spatial location of the coastline; i Indicates the i-th year, y i represents the distance from the intersection of the profile and the coastline in year i to the baseline; and Represents x i and y i The average value of ; a represents the intercept of the fitting constant; b represents the linear regression rate of change, i.e. LRR.

7. The method for analyzing coastline feature changes based on remote sensing images according to claim 6, characterized in that: In S7, conclusions and laws of changes in coastline characteristics are drawn based on the results of temporal and spatial change analysis, providing a scientific basis for marine environmental protection, coastal zone planning and management, etc.

Images